1 <!DOCTYPE Article PUBLIC "-//OASIS//DTD DocBook V3.1//EN">
3 <article id="building-guide">
7 <title>Building the Glasgow Functional Programming Tools Suite</title>
8 <author><othername>The GHC Team</othername></author>
9 <address><email>glasgow-haskell-{users,bugs}@haskell.org</email></address>
10 <pubdate>November 2001</pubdate>
13 <para>The Glasgow fptools suite is a collection of Functional
14 Programming related tools, including the Glasgow Haskell
15 Compiler (GHC). The source code for the whole suite is kept in
16 a single CVS repository and shares a common build and
17 installation system.</para>
19 <para>This guide is intended for people who want to build or
20 modify programs from the Glasgow <literal>fptools</literal>
21 suite (as distinct from those who merely want to
22 <emphasis>run</emphasis> them). Installation instructions are
23 now provided in the user guide.</para>
25 <para>The bulk of this guide applies to building on Unix
26 systems; see <xref linkend="winbuild"/> for Windows notes.</para>
32 <sect1 id="sec-getting">
33 <title>Getting the sources</title>
35 <para>You can get your hands on the <literal>fptools</literal>
41 <term><indexterm><primary>Source
42 distributions</primary></indexterm>Source distributions</term>
44 <para>You have a supported platform, but (a) you like
45 the warm fuzzy feeling of compiling things yourself;
46 (b) you want to build something ``extra”—e.g., a
47 set of libraries with strictness-analysis turned off; or
48 (c) you want to hack on GHC yourself.</para>
50 <para>A source distribution contains complete sources for
51 one or more projects in the <literal>fptools</literal>
52 suite. Not only that, but the more awkward
53 machine-independent steps are done for you. For example, if
55 <command>happy</command><indexterm><primary>happy</primary></indexterm>
56 you'll find it convenient that the source distribution
57 contains the result of running <command>happy</command> on
58 the parser specifications. If you don't want to alter the
59 parser then this saves you having to find and install
60 <command>happy</command>. You will still need a working
61 version of GHC (version 5.x or later) on your machine in
62 order to compile (most of) the sources, however.</para>
67 <term>The CVS repository.<indexterm><primary>CVS repository</primary></indexterm></term>
69 <para>We make releases infrequently. If you want more
70 up-to-the minute (but less tested) source code then you need
71 to get access to our CVS repository.</para>
73 <para>All the <literal>fptools</literal> source code is held
74 in a CVS repository. CVS is a pretty good source-code
75 control system, and best of all it works over the
78 <para>The repository holds source code only. It holds no
79 mechanically generated files at all. So if you check out a
80 source tree from CVS you will need to install every utility
81 so that you can build all the derived files from
84 <para>More information about our CVS repository can be found
85 in <xref linkend="sec-cvs"/>.</para>
90 <para>If you are going to do any building from sources (either
91 from a source distribution or the CVS repository) then you need to
92 read all of this manual in detail.</para>
96 <title>Using the CVS repository</title>
98 <para>We use <ulink url="http://www.cvshome.org/">CVS</ulink> (Concurrent Version System) to keep track of our
99 sources for various software projects. CVS lets several people
100 work on the same software at the same time, allowing changes to be
101 checked in incrementally. </para>
103 <para>This section is a set of guidelines for how to use our CVS
104 repository, and will probably evolve in time. The main thing to
105 remember is that most mistakes can be undone, but if there's
106 anything you're not sure about feel free to bug the local CVS
107 meister (namely Jeff Lewis
108 <email>jlewis@galois.com</email>). </para>
110 <sect2 id="cvs-access">
111 <title>Getting access to the CVS Repository</title>
113 <para>You can access the repository in one of two ways:
114 read-only (<xref linkend="cvs-read-only"/>), or read-write (<xref
115 linkend="cvs-read-write"/>).</para>
117 <sect3 id="cvs-read-only">
118 <title>Remote Read-only CVS Access</title>
120 <para>Read-only access is available to anyone - there's no
121 need to ask us first. With read-only CVS access you can do
122 anything except commit changes to the repository. You can
123 make changes to your local tree, and still use CVS's merge
124 facility to keep your tree up to date, and you can generate
125 patches using 'cvs diff' in order to send to us for
128 <para>To get read-only access to the repository:</para>
132 <para>Make sure that <application>cvs</application> is
133 installed on your machine.</para>
136 <para>Set your <literal>$CVSROOT</literal> environment variable to
137 <literal>:pserver:anoncvs@glass.cse.ogi.edu:/cvs</literal></para>
138 <para>If you set <literal>$CVSROOT</literal> in a shell script, be sure not to
139 have any trailing spaces on that line, otherwise CVS will respond with
140 a perplexing message like
141 <screen>/cvs : no such repository</screen></para>
144 <para>Run the command</para>
145 <screen>$ cvs login</screen>
146 <para>The password is simply <literal>cvs</literal>. This
147 sets up a file in your home directory called
148 <literal>.cvspass</literal>, which squirrels away the
149 dummy password, so you only need to do this step once.</para>
153 <para>Now go to <xref linkend="cvs-first"/>.</para>
158 <sect3 id="cvs-read-write">
159 <title>Remote Read-Write CVS Access</title>
161 <para>We generally supply read-write access to folk doing
162 serious development on some part of the source tree, when
163 going through us would be a pain. If you're developing some
164 feature, or think you have the time and inclination to fix
165 bugs in our sources, feel free to ask for read-write
166 access. There is a certain amount of responsibility that goes
167 with commit privileges; we are more likely to grant you access
168 if you've demonstrated your competence by sending us patches
169 via mail in the past.</para>
171 <para>To get remote read-write CVS access, you need to do the
172 following steps.</para>
176 <para>Make sure that <literal>cvs</literal> and
177 <literal>ssh</literal> are both installed on your
182 <para>Generate a DSA private-key/public-key pair, thus:</para>
183 <screen>$ ssh-keygen -d</screen>
184 <para>(<literal>ssh-keygen</literal> comes with
185 <literal>ssh</literal>.) Running <literal>ssh-keygen
186 -d</literal> creates the private and public keys in
187 <literal>$HOME/.ssh/id_dsa</literal> and
188 <literal>$HOME/.ssh/id_dsa.pub</literal> respectively
189 (assuming you accept the standard defaults).</para>
191 <para><literal>ssh-keygen -d</literal> will only work if
192 you have Version 2 <literal>ssh</literal> installed; it
193 will fail harmlessly otherwise. If you only have Version
194 1 you can instead generate an RSA key pair using plain</para>
195 <screen>$ ssh-keygen</screen>
197 <para>Doing so creates the private and public RSA keys in
198 <literal>$HOME/.ssh/identity</literal> and
199 <literal>$HOME/.ssh/identity.pub</literal>
202 <para>[Deprecated.] Incidentally, you can force a Version
203 2 <literal>ssh</literal> to use the Version 1 protocol by
204 creating <literal>$HOME/config</literal> with the
205 following in it:</para>
206 <programlisting>BatchMode Yes
209 Protocol 1</programlisting>
211 <para>In both cases, <literal>ssh-keygen</literal> will
212 ask for a <firstterm>passphrase</firstterm>. The
213 passphrase is a password that protects your private key.
214 In response to the 'Enter passphrase' question, you can
218 <para>[Recommended.] Enter a passphrase, which you
219 will quote each time you use CVS.
220 <literal>ssh-agent</literal> makes this entirely
224 <para>[Deprecated.] Just hit return (i.e. use an empty
225 passphrase); then you won't need to quote the
226 passphrase when using CVS. The downside is that
227 anyone who can see into your <literal>.ssh</literal>
228 directory, and thereby get your private key, can mess
229 up the repository. So you must keep the
230 <literal>.ssh</literal> directory with draconian
231 no-access permissions.</para>
237 <emphasis>Windows users: see the notes in <xref linkend="configure-ssh"/> about <command>ssh</command> wrinkles!</emphasis>
244 <para>Send a message to to the CVS repository
245 administrator (currently Jeff Lewis
246 <email>jeff@galois.com</email>), containing:</para>
249 <para>Your desired user-name.</para>
252 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
253 <literal>.ssh/identity.pub</literal>).</para>
256 <para>He will set up your account.</para>
260 <para>Set the following environment variables:</para>
264 <constant>$HOME</constant>: points to your home directory. This is where CVS
265 will look for its <filename>.cvsrc</filename> file.
271 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
273 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
274 <literal>ssh</literal> assumes that your CVS client
275 understands how to execute shell script
276 ("#!"s,really), which is what
277 <literal>ssh</literal> is. This may not be the case on
278 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
279 <literal>ssh1</literal>.</para>
283 <para><literal>$CVSROOT</literal> to
284 <literal>:ext:</literal><replaceable>your-username</replaceable>
285 <literal>@cvs.haskell.org:/home/cvs/root</literal>
286 where <replaceable>your-username</replaceable> is your user name on
287 <literal>cvs.haskell.org</literal>.
289 <para>The <literal>CVSROOT</literal> environment variable will
290 be recorded in the checked-out tree, so you don't need to set
291 this every time. </para>
297 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
298 if you want to use an Emacs buffer for typing in those long commit messages.
304 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
305 set this to point to <filename>bash.exe</filename>.
316 Put the following in <filename>$HOME/.cvsrc</filename>:
319 <programlisting>checkout -P
322 diff -u</programlisting>
325 These are the default options for the specified CVS commands,
326 and represent better defaults than the usual ones. (Feel
327 free to change them.)
331 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
332 the 8.3 DOS filesystem, but that restriction should have
333 been lifted by now (i.e., you're using VFAT or later filesystems.) If
334 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
342 <para>[Experts.] Once your account is set up, you can get
343 access from other machines without bothering Jeff, thus:</para>
346 <para>Generate a public/private key pair on the new
350 <para>Use ssh to log in to
351 <literal>cvs.haskell.org</literal>, from your old
355 <para>Add the public key for the new machine to the file
356 <literal>$HOME/ssh/authorized_keys</literal> on
357 <literal>cvs.haskell.org</literal>.
358 (<literal>authorized_keys2</literal>, I think, for Version
362 <para>Make sure that the new version of
363 <literal>authorized_keys</literal> still has 600 file
372 <sect2 id="cvs-first">
373 <title>Checking Out a Source Tree</title>
377 <para>Make sure you set your <literal>CVSROOT</literal>
378 environment variable according to either of the remote
379 methods above. The Approved Way to check out a source tree
380 is as follows:</para>
382 <screen>$ cvs checkout fpconfig</screen>
384 <para>At this point you have a new directory called
385 <literal>fptools</literal> which contains the basic stuff
386 for the fptools suite, including the configuration files and
387 some other junk. </para>
389 <para>[Windows users.] The following messages appear to be harmless:
390 <screen>setsockopt IPTOS_LOWDELAY: Invalid argument
391 setsockopt IPTOS_THROUGHPUT: Invalid argument</screen>
395 <para>You can call the fptools directory whatever you like,
396 CVS won't mind: </para>
398 <screen>$ mv fptools <replaceable>directory</replaceable></screen>
400 <para> NB: after you've read the CVS manual you might be
401 tempted to try</para>
402 <screen>$ cvs checkout -d <replaceable>directory</replaceable> fpconfig</screen>
404 <para>instead of checking out <literal>fpconfig</literal>
405 and then renaming it. But this doesn't work, and will
406 result in checking out the entire repository instead of just
407 the <literal>fpconfig</literal> bit.</para>
408 <screen>$ cd <replaceable>directory</replaceable>
409 $ cvs checkout ghc hslibs libraries</screen>
411 <para>The second command here checks out the relevant
412 modules you want to work on. For a GHC build, for instance,
413 you need at least the <literal>ghc</literal>,
414 <literal>hslibs</literal> and <literal>libraries</literal>
415 modules (for a full list of the projects available, see
416 <xref linkend="projects"/>).</para>
418 <para>Remember that if you do not have
419 <literal>happy</literal> and/or <literal>Alex</literal>
420 installed, you need to check them out as well.</para>
425 <sect2 id="cvs-committing">
426 <title>Committing Changes</title>
428 <para>This is only if you have read-write access to the
429 repository. For anoncvs users, CVS will issue a "read-only
430 repository" error if you try to commit changes.</para>
434 <para>Build the software, if necessary. Unless you're just
435 working on documentation, you'll probably want to build the
436 software in order to test any changes you make.</para>
440 <para>Make changes. Preferably small ones first.</para>
444 <para>Test them. You can see exactly what changes you've
445 made by using the <literal>cvs diff</literal> command:</para>
446 <screen>$ cvs diff</screen>
447 <para>lists all the changes (using the
448 <literal>diff</literal> command) in and below the current
449 directory. In emacs, <literal>C-c C-v =</literal> runs
450 <literal>cvs diff</literal> on the current buffer and shows
451 you the results.</para>
455 <para>If you changed something in the
456 <literal>fptools/libraries</literal> subdirectories, also run
457 <literal>make html</literal> to check if the documentation can
458 be generated successfully, too.</para>
462 <para>Before checking in a change, you need to update your
466 $ cvs update</screen>
467 <para>This pulls in any changes that other people have made,
468 and merges them with yours. If there are any conflicts, CVS
469 will tell you, and you'll have to resolve them before you
470 can check your changes in. The documentation describes what
471 to do in the event of a conflict.</para>
473 <para>It's not always necessary to do a full cvs update
474 before checking in a change, since CVS will always tell you
475 if you try to check in a file that someone else has changed.
476 However, you should still update at regular intervals to
477 avoid making changes that don't work in conjuction with
478 changes that someone else made. Keeping an eye on what goes
479 by on the mailing list can help here.</para>
483 <para>When you're happy that your change isn't going to
484 break anything, check it in. For a one-file change:</para>
486 <screen>$ cvs commit <replaceable>filename</replaceable></screen>
488 <para>CVS will then pop up an editor for you to enter a
489 "commit message", this is just a short description
490 of what your change does, and will be kept in the history of
493 <para>If you're using emacs, simply load up the file into a
494 buffer and type <literal>C-x C-q</literal>, and emacs will
495 prompt for a commit message and then check in the file for
498 <para>For a multiple-file change, things are a bit
499 trickier. There are several ways to do this, but this is the
500 way I find easiest. First type the commit message into a
501 temporary file. Then either</para>
503 <screen>$ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable></screen>
505 <para>or, if nothing else has changed in this part of the
508 <screen>$ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable></screen>
510 <para>where <replaceable>directory</replaceable> is a common
511 parent directory for all your changes, and
512 <replaceable>commit-message</replaceable> is the name of the
513 file containing the commit message.</para>
515 <para>Shortly afterwards, you'll get some mail from the
516 relevant mailing list saying which files changed, and giving
517 the commit message. For a multiple-file change, you should
518 still get only <emphasis>one</emphasis> message.</para>
523 <sect2 id="cvs-update">
524 <title>Updating Your Source Tree</title>
526 <para>It can be tempting to cvs update just part of a source
527 tree to bring in some changes that someone else has made, or
528 before committing your own changes. This is NOT RECOMMENDED!
529 Quite often changes in one part of the tree are dependent on
530 changes in another part of the tree (the
531 <literal>mk/*.mk</literal> files are a good example where
532 problems crop up quite often). Having an inconsistent tree is a
533 major cause of headaches. </para>
535 <para>So, to avoid a lot of hassle, follow this recipe for
536 updating your tree:</para>
539 $ cvs update -P 2>&1 | tee log</screen>
541 <para>Look at the log file, and fix any conflicts (denoted by a
542 <quote>C</quote> in the first column). New directories may have
543 appeared in the repository; CVS doesn't check these out by
544 default, so to get new directories you have to explicitly do
545 <screen>$ cvs update -d</screen>
546 in each project subdirectory. Don't do this at the top level,
547 because then <emphasis>all</emphasis> the projects will be
550 <para>If you're using multiple build trees, then for every build
551 tree you have pointing at this source tree, you need to update
552 the links in case any new files have appeared: </para>
554 <screen>$ cd <replaceable>build-tree</replaceable>
555 $ lndir <replaceable>source-tree</replaceable></screen>
557 <para>Some files might have been removed, so you need to remove
558 the links pointing to these non-existent files:</para>
560 <screen>$ find . -xtype l -exec rm '{}' \;</screen>
562 <para>To be <emphasis>really</emphasis> safe, you should do
565 <screen>$ gmake all</screen>
567 <para>from the top-level, to update the dependencies and build
568 any changed files. </para>
571 <sect2 id="cvs-tags">
572 <title>GHC Tag Policy</title>
574 <para>If you want to check out a particular version of GHC,
575 you'll need to know how we tag versions in the repository. The
576 policy (as of 4.04) is:</para>
580 <para>The tree is branched before every major release. The
581 branch tag is <literal>ghc-x-xx-branch</literal>, where
582 <literal>x-xx</literal> is the version number of the release
583 with the <literal>'.'</literal> replaced by a
584 <literal>'-'</literal>. For example, the 4.04 release lives
585 on <literal>ghc-4-04-branch</literal>.</para>
589 <para>The release itself is tagged with
590 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
591 called <literal>ghc-4-06</literal>.</para>
595 <para>We didn't always follow these guidelines, so to see
596 what tags there are for previous versions, do <literal>cvs
597 log</literal> on a file that's been around for a while (like
598 <literal>fptools/ghc/README</literal>).</para>
602 <para>So, to check out a fresh GHC 4.06 tree you would
605 <screen>$ cvs co -r ghc-4-06 fpconfig
607 $ cvs co -r ghc-4-06 ghc hslibs</screen>
610 <sect2 id="cvs-hints">
611 <title>General Hints</title>
615 <para>As a general rule: commit changes in small units,
616 preferably addressing one issue or implementing a single
617 feature. Provide a descriptive log message so that the
618 repository records exactly which changes were required to
619 implement a given feature/fix a bug. I've found this
620 <emphasis>very</emphasis> useful in the past for finding out
621 when a particular bug was introduced: you can just wind back
622 the CVS tree until the bug disappears.</para>
626 <para>Keep the sources at least *buildable* at any given
627 time. No doubt bugs will creep in, but it's quite easy to
628 ensure that any change made at least leaves the tree in a
629 buildable state. We do nightly builds of GHC to keep an eye
630 on what things work/don't work each day and how we're doing
631 in relation to previous verions. This idea is truely wrecked
632 if the compiler won't build in the first place!</para>
636 <para>To check out extra bits into an already-checked-out
637 tree, use the following procedure. Suppose you have a
638 checked-out fptools tree containing just ghc, and you want
639 to add nofib to it:</para>
642 $ cvs checkout nofib</screen>
647 $ cvs update -d nofib</screen>
649 <para>(the -d flag tells update to create a new
650 directory). If you just want part of the nofib suite, you
654 $ cvs checkout nofib/spectral</screen>
656 <para>This works because <literal>nofib</literal> is a
657 module in its own right, and spectral is a subdirectory of
658 the nofib module. The path argument to checkout must always
659 start with a module name. There's no equivalent form of this
660 command using <literal>update</literal>.</para>
666 <sect1 id="projects">
667 <title>What projects are there?</title>
669 <para>The <literal>fptools</literal> suite consists of several
670 <firstterm>projects</firstterm>, most of which can be downloaded,
671 built and installed individually. Each project corresponds to a
672 subdirectory in the source tree, and if checking out from CVS then
673 each project can be checked out individually by sitting in the top
674 level of your source tree and typing <command>cvs checkout
675 <replaceable>project</replaceable></command>.</para>
677 <para>Here is a list of the projects currently available:</para>
682 <literal>alex</literal>
683 <indexterm><primary><literal>alex</literal></primary><secondary>project</secondary></indexterm>
687 url="http://www.haskell.org/alex/">Alex</ulink> lexical
688 analyser generator for Haskell.</para>
694 <literal>ghc</literal>
695 <indexterm><primary><literal>ghc</literal></primary>
696 <secondary>project</secondary></indexterm>
699 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
700 Haskell Compiler</ulink> (minus libraries). Absolutely
701 required for building GHC.</para>
707 <literal>glafp-utils</literal>
708 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
711 <para>Utility programs, some of which are used by the
712 build/installation system. Required for pretty much
719 <literal>greencard</literal>
720 <indexterm><primary><literal>greencard</literal></primary><secondary>project</secondary></indexterm>
724 url="http://www.haskell.org/greencard/">GreenCard</ulink>
725 system for generating Haskell foreign function
732 <literal>haggis</literal>
733 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
737 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
738 Haskell GUI framework.</para>
744 <literal>haddock</literal>
745 <indexterm><primary><literal>haddock</literal></primary><secondary>project</secondary></indexterm>
749 url="http://www.haskell.org/haddock/">Haddock</ulink>
750 documentation tool.</para>
756 <literal>happy</literal>
757 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
761 url="http://www.haskell.org/happy/">Happy</ulink> Parser
768 <literal>hdirect</literal>
769 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
773 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
774 Haskell interoperability tool.</para>
780 <literal>hood</literal>
781 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
784 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
785 Object Observation Debugger</ulink>.</para>
791 <literal>hslibs</literal>
792 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
795 <para>Supplemental libraries for GHC
796 (<emphasis>required</emphasis> for building GHC).</para>
802 <literal>libraries</literal>
803 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
806 <para>Hierarchical Haskell library suite
807 (<emphasis>required</emphasis> for building GHC).</para>
813 <literal>mhms</literal>
814 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
817 <para>The Modular Haskell Metric System.</para>
823 <literal>nofib</literal>
824 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
827 <para>The NoFib suite: A collection of Haskell programs used
828 primarily for benchmarking.</para>
834 <literal>testsuite</literal>
835 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
838 <para>A testing framework, including GHC's regression test
844 <para>So, to build GHC you need at least the
845 <literal>ghc</literal>, <literal>libraries</literal> and
846 <literal>hslibs</literal> projects (a GHC source distribution will
847 already include the bits you need).</para>
850 <sect1 id="sec-build-checks">
851 <title>Things to check before you start</title>
853 <para>Here's a list of things to check before you get
858 <listitem><para><indexterm><primary>Disk space needed</primary></indexterm>Disk
859 space needed: from about 100Mb for a basic GHC
860 build, up to probably 500Mb for a GHC build with everything
861 included (libraries built several different ways,
866 <para>Use an appropriate machine / operating system. <xref
867 linkend="sec-port-info"/> lists the supported platforms; if
868 yours isn't amongst these then you can try porting GHC (see
869 <xref linkend="sec-porting-ghc"/>).</para>
873 <para>Be sure that the “pre-supposed” utilities are
874 installed. <xref linkend="sec-pre-supposed"/>
879 <para>If you have any problem when building or installing the
880 Glasgow tools, please check the “known pitfalls” (<xref
881 linkend="sec-build-pitfalls"/>). Also check the FAQ for the
882 version you're building, which is part of the User's Guide and
883 available on the <ulink url="http://www.haskell.org/ghc/" >GHC web
886 <indexterm><primary>bugs</primary><secondary>known</secondary></indexterm>
888 <para>If you feel there is still some shortcoming in our
889 procedure or instructions, please report it.</para>
891 <para>For GHC, please see the <ulink
892 url="http://www.haskell.org/ghc/docs/latest/set/bug-reporting.html">bug-reporting
893 section of the GHC Users' Guide</ulink>, to maximise the
894 usefulness of your report.</para>
896 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
897 <para>If in doubt, please send a message to
898 <email>glasgow-haskell-bugs@haskell.org</email>.
899 <indexterm><primary>bugs</primary><secondary>mailing
900 list</secondary></indexterm></para>
905 <sect1 id="sec-port-info">
906 <title>What machines the Glasgow tools run on</title>
908 <indexterm><primary>ports</primary><secondary>GHC</secondary></indexterm>
909 <indexterm><primary>GHC</primary><secondary>ports</secondary></indexterm>
910 <indexterm><primary>platforms</primary><secondary>supported</secondary></indexterm>
912 <para>The main question is whether or not the Haskell compiler
913 (GHC) runs on your platform.</para>
915 <para>A “platform” is a
916 architecture/manufacturer/operating-system combination, such as
917 <literal>sparc-sun-solaris2</literal>. Other common ones are
918 <literal>alpha-dec-osf2</literal>,
919 <literal>hppa1.1-hp-hpux9</literal>,
920 <literal>i386-unknown-linux</literal>,
921 <literal>i386-unknown-solaris2</literal>,
922 <literal>i386-unknown-freebsd</literal>,
923 <literal>i386-unknown-cygwin32</literal>,
924 <literal>m68k-sun-sunos4</literal>,
925 <literal>mips-sgi-irix5</literal>,
926 <literal>sparc-sun-sunos4</literal>,
927 <literal>sparc-sun-solaris2</literal>,
928 <literal>powerpc-ibm-aix</literal>.</para>
930 <para>Some libraries may only work on a limited number of
931 platforms; for example, a sockets library is of no use unless the
932 operating system supports the underlying BSDisms.</para>
935 <title>What platforms the Haskell compiler (GHC) runs on</title>
937 <indexterm><primary>fully-supported platforms</primary></indexterm>
938 <indexterm><primary>native-code generator</primary></indexterm>
939 <indexterm><primary>registerised ports</primary></indexterm>
940 <indexterm><primary>unregisterised ports</primary></indexterm>
942 <para>The GHC hierarchy of Porting Goodness: (a) Best is a
943 native-code generator; (b) next best is a
944 “registerised” port; (c) the bare minimum is an
945 “unregisterised” port.
946 (“Unregisterised” is so terrible that we won't say
947 more about it).</para>
949 <para>We use Sparcs running Solaris 2.7 and x86 boxes running
950 FreeBSD and Linux, so those are the best supported platforms,
951 unsurprisingly.</para>
953 <para>Here's everything that's known about GHC ports. We
954 identify platforms by their “canonical”
955 CPU/Manufacturer/OS triple.</para>
959 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:
960 <indexterm><primary>alpha-dec-osf</primary></indexterm>
961 <indexterm><primary>alpha-dec-linux</primary></indexterm>
962 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
963 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
964 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
967 <para>The OSF port is currently working (as of GHC version
968 5.02.1) and well supported. The native code generator is
969 currently non-working. Other operating systems will
970 require some minor porting.</para>
975 <term>sparc-sun-sunos4
976 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
979 <para>Probably works with minor tweaks, hasn't been tested
985 <term>sparc-sun-solaris2
986 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
989 <para>Fully supported (at least for Solaris 2.7 and 2.6),
990 including native-code generator.</para>
995 <term>sparc-unknown-openbsd
996 <indexterm><primary>sparc-unknown-openbsd</primary></indexterm>
999 <para>Supported, including native-code generator. The
1000 same should also be true of NetBSD</para>
1005 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)
1006 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1009 <para>A registerised port is available for version 4.08,
1010 but GHC hasn't been built on that platform since (as far
1011 as we know). No native-code generator.</para>
1016 <term>i386-unknown-linux (PCs running Linux, ELF binary format)
1017 <indexterm><primary>i386-*-linux</primary></indexterm>
1020 <para>GHC works registerised and has a native code
1021 generator. You <emphasis>must</emphasis> have GCC 2.7.x
1022 or later. NOTE about <literal>glibc</literal> versions:
1023 GHC binaries built on a system running <literal>glibc
1024 2.0</literal> won't work on a system running
1025 <literal>glibc 2.1</literal>, and vice versa. In general,
1026 don't expect compatibility between
1027 <literal>glibc</literal> versions, even if the shared
1028 library version hasn't changed.</para>
1033 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or higher)
1034 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1037 <para>GHC works registerised. Pre-built packages are
1038 available in the native package format, so if you just
1039 need binaries you're better off just installing the
1040 package (it might even be on your installation
1046 <term>i386-unknown-openbsd (PCs running OpenBSD)
1047 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1050 <para>Supported, with native code generator. Packages are
1051 available through the ports system in the native package
1057 <term>i386-unknown-netbsd (PCs running NetBSD)
1058 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1061 <para>Will require some minor porting effort, but should
1062 work registerised.</para>
1067 <term>i386-unknown-mingw32 (PCs running Windows)
1068 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1071 <para>Fully supported under Win9x, WinNT, Win2k, and
1072 WinXP. Includes a native code generator. Building from
1073 source requires a recent <ulink
1074 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1075 to be installed.</para>
1080 <term>ia64-unknown-linux
1081 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1084 <para>Supported, except there is no native code
1090 <term>x86_64-unknown-linux
1091 <indexterm><primary>x86_64-unknown-linux</primary></indexterm>
1094 <para>GHC currently works unregisterised. A registerised
1095 port is in progress.</para>
1100 <term>amd64-unknown-openbsd
1101 <indexterm><primary>amd64-unknown-linux</primary></indexterm>
1104 <para>(This is the same as x86_64-unknown-openbsd). GHC
1105 currently works unregisterised. A registerised port is in
1111 <term>mips-sgi-irix5
1112 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1115 <para>Port has worked in the past, but hasn't been tested
1116 for some time (and will certainly have rotted in various
1117 ways). As usual, we don't have access to machines and
1118 there hasn't been an overwhelming demand for this port,
1119 but feel free to get in touch.</para>
1124 <term>mips64-sgi-irix6
1125 <indexterm><primary>mips-sgi-irix6</primary></indexterm>
1128 <para>GHC currently works unregisterised.</para>
1133 <term>powerpc-ibm-aix
1134 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1137 <para>Port currently doesn't work, needs some minimal
1138 porting effort. As usual, we don't have access to
1139 machines and there hasn't been an overwhelming demand for
1140 this port, but feel free to get in touch.</para>
1145 <term>powerpc-apple-darwin
1146 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1149 <para>Supported registerised. Native code generator is
1150 almost working.</para>
1155 <term>powerpc-apple-linux
1156 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1159 <para>Not supported (yet).</para>
1164 <para>Various other systems have had GHC ported to them in the
1165 distant past, including various Motorola 68k boxes. The 68k
1166 support still remains, but porting to one of these systems will
1167 certainly be a non-trivial task.</para>
1171 <title>What machines the other tools run on</title>
1173 <para>Unless you hear otherwise, the other tools work if GHC
1179 <sect1 id="sec-pre-supposed">
1180 <title>Installing pre-supposed utilities</title>
1182 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1183 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1185 <para>Here are the gory details about some utility programs you
1186 may need; <command>perl</command>, <command>gcc</command> and
1187 <command>happy</command> are the only important
1188 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1189 important if you're going for Parallel Haskell.) The
1190 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1191 script will tell you if you are missing something.</para>
1197 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1198 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1201 <para>GHC is required to build many of the tools, including
1202 GHC itself. If you need to port GHC to your platform
1203 because there isn't a binary distribution of GHC available,
1204 then see <xref linkend="sec-porting-ghc"/>.</para>
1206 <para>Which version of GHC you need will depend on the
1207 packages you intend to build. GHC itself will normally
1208 build using one of several older versions of itself - check
1209 the announcement or release notes for details.</para>
1215 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1216 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1219 <para><emphasis>You have to have Perl to proceed!</emphasis>
1220 Perl version 5 at least is required. GHC has been known to
1221 tickle bugs in Perl, so if you find that Perl crashes when
1222 running GHC try updating (or downgrading) your Perl
1223 installation. Versions of Perl that we use and are known to
1224 be fairly stable are 5.005 and 5.6.1.</para>
1226 <para>For Win32 platforms, you should use the binary
1227 supplied in the InstallShield (copy it to
1228 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1231 <para>Perl should be put somewhere so that it can be invoked
1232 by the <literal>#!</literal> script-invoking
1233 mechanism. The full pathname may need to be less than 32
1234 characters long on some systems.</para>
1239 <term>GNU C (<command>gcc</command>)
1240 <indexterm><primary>pre-supposed: GCC (GNU C compiler)</primary></indexterm>
1241 <indexterm><primary>GCC (GNU C compiler), pre-supposed</primary></indexterm>
1244 <para>We recommend using GCC version 2.95.2 on all
1245 platforms. Failing that, version 2.7.2 is stable on most
1246 platforms. Earlier versions of GCC can be assumed not to
1247 work, and versions in between 2.7.2 and 2.95.2 (including
1248 <command>egcs</command>) have varying degrees of stability
1249 depending on the platform.</para>
1251 <para>GCC 3.2 is currently known to have problems building
1252 GHC on Sparc, but is stable on x86.</para>
1254 <para>If your GCC dies with “internal error” on
1255 some GHC source file, please let us know, so we can report
1256 it and get things improved. (Exception: on x86
1257 boxes—you may need to fiddle with GHC's
1258 <option>-monly-N-regs</option> option; see the User's
1265 <indexterm><primary>make</primary><secondary>GNU</secondary></indexterm>
1268 <para>The fptools build system makes heavy use of features
1269 specific to GNU <command>make</command>, so you must have
1270 this installed in order to build any of the fptools
1277 <indexterm><primary>Happy</primary></indexterm>
1280 <para>Happy is a parser generator tool for Haskell, and is
1281 used to generate GHC's parsers. Happy is written in
1282 Haskell, and is a project in the CVS repository
1283 (<literal>fptools/happy</literal>). It can be built from
1284 source, but bear in mind that you'll need GHC installed in
1285 order to build it. To avoid the chicken/egg problem,
1286 install a binary distribution of either Happy or GHC to get
1287 started. Happy distributions are available from <ulink
1288 url="http://www.haskell.org/happy/">Happy's Web
1289 Page</ulink>.</para>
1295 <indexterm><primary>Alex</primary></indexterm>
1298 <para>Alex is a lexical-analyser generator for Haskell,
1299 which GHC uses to generate its lexer. Like Happy, Alex is
1300 written in Haskell and is a project in the CVS repository.
1301 Alex distributions are available from <ulink
1302 url="http://www.haskell.org/alex/">Alex's Web
1303 Page</ulink>.</para>
1309 <indexterm><primary>pre-supposed: autoconf</primary></indexterm>
1310 <indexterm><primary>autoconf, pre-supposed</primary></indexterm>
1313 <para>GNU autoconf is needed if you intend to build from the
1314 CVS sources, it is <emphasis>not</emphasis> needed if you
1315 just intend to build a standard source distribution.</para>
1317 <para>Version 2.52 or later of the autoconf package is required.
1318 NB. version 2.13 will no longer work, as of GHC version
1321 <para><command>autoreconf</command> (from the autoconf package)
1322 recursively builds <command>configure</command> scripts from
1323 the corresponding <filename>configure.ac</filename> and
1324 <filename>aclocal.m4</filename> files. If you modify one of
1325 the latter files, you'll need <command>autoreconf</command> to
1326 rebuild the corresponding <filename>configure</filename>.</para>
1331 <term><command>sed</command>
1332 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1333 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1336 <para>You need a working <command>sed</command> if you are
1337 going to build from sources. The build-configuration stuff
1338 needs it. GNU sed version 2.0.4 is no good! It has a bug
1339 in it that is tickled by the build-configuration. 2.0.5 is
1340 OK. Others are probably OK too (assuming we don't create too
1341 elaborate configure scripts.)</para>
1346 <para>One <literal>fptools</literal> project is worth a quick note
1347 at this point, because it is useful for all the others:
1348 <literal>glafp-utils</literal> contains several utilities which
1349 aren't particularly Glasgow-ish, but Occasionally Indispensable.
1350 Like <command>lndir</command> for creating symbolic link
1353 <sect2 id="pre-supposed-gph-tools">
1354 <title>Tools for building parallel GHC (GPH)</title>
1358 <term>PVM version 3:
1359 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1360 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1363 <para>PVM is the Parallel Virtual Machine on which
1364 Parallel Haskell programs run. (You only need this if you
1365 plan to run Parallel Haskell. Concurrent Haskell, which
1366 runs concurrent threads on a uniprocessor doesn't need
1367 it.) Underneath PVM, you can have (for example) a network
1368 of workstations (slow) or a multiprocessor box
1371 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1372 It is readily available on the net; I think I got it from
1373 <literal>research.att.com</literal>, in
1374 <filename>netlib</filename>.</para>
1376 <para>A PVM installation is slightly quirky, but easy to
1377 do. Just follow the <filename>Readme</filename>
1378 instructions.</para>
1383 <term><command>bash</command>:
1384 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1387 <para>Sadly, the <command>gr2ps</command> script, used to
1388 convert “parallelism profiles” to PostScript,
1389 is written in Bash (GNU's Bourne Again shell). This bug
1390 will be fixed (someday).</para>
1396 <sect2 id="pre-supposed-other-tools">
1397 <title>Other useful tools</title>
1402 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1403 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1406 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1407 to build a couple of utilities in
1408 <literal>glafp-utils</literal>. Depending on your
1409 operating system, the supplied <command>lex</command> may
1410 or may not work; you should get the GNU version.</para>
1415 <para>More tools are required if you want to format the documentation
1416 that comes with GHC and other fptools projects. See <xref
1417 linkend="building-docs"/>.</para>
1421 <sect1 id="sec-building-from-source">
1422 <title>Building from source</title>
1424 <indexterm><primary>Building from source</primary></indexterm>
1425 <indexterm><primary>Source, building from</primary></indexterm>
1427 <para>You've been rash enough to want to build some of the Glasgow
1428 Functional Programming tools (GHC, Happy, nofib, etc.) from
1429 source. You've slurped the source, from the CVS repository or
1430 from a source distribution, and now you're sitting looking at a
1431 huge mound of bits, wondering what to do next.</para>
1433 <para>Gingerly, you type <command>make</command>. Wrong
1436 <para>This rest of this guide is intended for duffers like me, who
1437 aren't really interested in Makefiles and systems configurations,
1438 but who need a mental model of the interlocking pieces so that
1439 they can make them work, extend them consistently when adding new
1440 software, and lay hands on them gently when they don't
1443 <sect2 id="quick-start">
1444 <title>Quick Start</title>
1446 <para>If you are starting from a source distribution, and just
1447 want a completely standard build, then the following should
1450 <screen>$ autoreconf
1453 $ make install</screen>
1455 <para>For GHC, this will do a 2-stage bootstrap build of the
1456 compiler, with profiling libraries, and install the
1459 <para>If you want to do anything at all non-standard, or you
1460 want to do some development, read on...</para>
1463 <sect2 id="sec-source-tree">
1464 <title>Your source tree</title>
1466 <para>The source code is held in your <emphasis>source
1467 tree</emphasis>. The root directory of your source tree
1468 <emphasis>must</emphasis> contain the following directories and
1473 <para><filename>Makefile</filename>: the root
1478 <para><filename>mk/</filename>: the directory that contains
1479 the main Makefile code, shared by all the
1480 <literal>fptools</literal> software.</para>
1484 <para><filename>configure.ac</filename>,
1485 <filename>config.sub</filename>,
1486 <filename>config.guess</filename>: these files support the
1487 configuration process.</para>
1491 <para><filename>install-sh</filename>.</para>
1495 <para>All the other directories are individual
1496 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1497 system—for example, the Glasgow Haskell Compiler
1498 (<literal>ghc</literal>), the Happy parser generator
1499 (<literal>happy</literal>), the <literal>nofib</literal>
1500 benchmark suite, and so on. You can have zero or more of these.
1501 Needless to say, some of them are needed to build others.</para>
1503 <para>The important thing to remember is that even if you want
1504 only one project (<literal>happy</literal>, say), you must have
1505 a source tree whose root directory contains
1506 <filename>Makefile</filename>, <filename>mk/</filename>,
1507 <filename>configure.ac</filename>, and the project(s) you want
1508 (<filename>happy/</filename> in this case). You cannot get by
1509 with just the <filename>happy/</filename> directory.</para>
1513 <title>Build trees</title>
1514 <indexterm><primary>build trees</primary></indexterm>
1515 <indexterm><primary>link trees, for building</primary></indexterm>
1517 <para>If you just want to build the software once on a single
1518 platform, then your source tree can also be your build tree, and
1519 you can skip the rest of this section.</para>
1521 <para>We often want to build multiple versions of our software
1522 for different architectures, or with different options
1523 (e.g. profiling). It's very desirable to share a single copy of
1524 the source code among all these builds.</para>
1526 <para>So for every source tree we have zero or more
1527 <emphasis>build trees</emphasis>. Each build tree is initially
1528 an exact copy of the source tree, except that each file is a
1529 symbolic link to the source file, rather than being a copy of
1530 the source file. There are “standard” Unix
1531 utilities that make such copies, so standard that they go by
1533 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1534 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1535 are two (If you don't have either, the source distribution
1536 includes sources for the X11
1537 <command>lndir</command>—check out
1538 <filename>fptools/glafp-utils/lndir</filename>). See <xref
1539 linkend="sec-storysofar"/> for a typical invocation.</para>
1541 <para>The build tree does not need to be anywhere near the
1542 source tree in the file system. Indeed, one advantage of
1543 separating the build tree from the source is that the build tree
1544 can be placed in a non-backed-up partition, saving your systems
1545 support people from backing up untold megabytes of
1546 easily-regenerated, and rapidly-changing, gubbins. The golden
1547 rule is that (with a single exception—<xref
1548 linkend="sec-build-config"/>) <emphasis>absolutely everything in
1549 the build tree is either a symbolic link to the source tree, or
1550 else is mechanically generated</emphasis>. It should be
1551 perfectly OK for your build tree to vanish overnight; an hour or
1552 two compiling and you're on the road again.</para>
1554 <para>You need to be a bit careful, though, that any new files
1555 you create (if you do any development work) are in the source
1556 tree, not a build tree!</para>
1558 <para>Remember, that the source files in the build tree are
1559 <emphasis>symbolic links</emphasis> to the files in the source
1560 tree. (The build tree soon accumulates lots of built files like
1561 <filename>Foo.o</filename>, as well.) You can
1562 <emphasis>delete</emphasis> a source file from the build tree
1563 without affecting the source tree (though it's an odd thing to
1564 do). On the other hand, if you <emphasis>edit</emphasis> a
1565 source file from the build tree, you'll edit the source-tree
1566 file directly. (You can set up Emacs so that if you edit a
1567 source file from the build tree, Emacs will silently create an
1568 edited copy of the source file in the build tree, leaving the
1569 source file unchanged; but the danger is that you think you've
1570 edited the source file whereas actually all you've done is edit
1571 the build-tree copy. More commonly you do want to edit the
1572 source file.)</para>
1574 <para>Like the source tree, the top level of your build tree
1575 must be (a linked copy of) the root directory of the
1576 <literal>fptools</literal> suite. Inside Makefiles, the root of
1577 your build tree is called
1578 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1579 In the rest of this document path names are relative to
1580 <constant>$(FPTOOLS_TOP)</constant> unless
1581 otherwise stated. For example, the file
1582 <filename>ghc/mk/target.mk</filename> is actually
1583 <filename>$(FPTOOLS_TOP)/ghc/mk/target.mk</filename>.</para>
1586 <sect2 id="sec-build-config">
1587 <title>Getting the build you want</title>
1589 <para>When you build <literal>fptools</literal> you will be
1590 compiling code on a particular <emphasis>host
1591 platform</emphasis>, to run on a particular <emphasis>target
1592 platform</emphasis> (usually the same as the host
1593 platform)<indexterm><primary>platform</primary></indexterm>.
1594 The difficulty is that there are minor differences between
1595 different platforms; minor, but enough that the code needs to be
1596 a bit different for each. There are some big differences too:
1597 for a different architecture we need to build GHC with a
1598 different native-code generator.</para>
1600 <para>There are also knobs you can turn to control how the
1601 <literal>fptools</literal> software is built. For example, you
1602 might want to build GHC optimised (so that it runs fast) or
1603 unoptimised (so that you can compile it fast after you've
1604 modified it. Or, you might want to compile it with debugging on
1605 (so that extra consistency-checking code gets included) or off.
1608 <para>All of this stuff is called the
1609 <emphasis>configuration</emphasis> of your build. You set the
1610 configuration using a three-step process.</para>
1614 <term>Step 1: get ready for configuration.</term>
1616 <para>NOTE: if you're starting from a source distribution,
1617 rather than CVS sources, you can skip this step.</para>
1619 <para>Change directory to
1620 <constant>$(FPTOOLS_TOP)</constant> and
1621 issue the command</para>
1622 <programlisting>autoreconf</programlisting>
1623 <indexterm><primary>autoreconf</primary></indexterm>
1624 <para>(with no arguments). This GNU program (recursively) converts
1625 <filename>$(FPTOOLS_TOP)/configure.ac</filename> and
1626 <filename>$(FPTOOLS_TOP)/aclocal.m4</filename>
1627 to a shell script called
1628 <filename>$(FPTOOLS_TOP)/configure</filename>.
1629 If <command>autoreconf</command> bleats that it can't write the file <filename>configure</filename>,
1630 then delete the latter and try again. Note that you must use <command>autoreconf</command>,
1631 and not the old <command>autoconf</command>! If you erroneously use the latter, you'll get
1632 a message like "No rule to make target 'mk/config.h.in'".
1635 <para>Some projects, including GHC, have their own configure script.
1636 <command>autoreconf</command> takes care of that, too, so all you have
1637 to do is calling <command>autoreconf</command> in the top-level directory
1638 <filename>$(FPTOOLS_TOP)</filename>.</para>
1640 <para>These steps are completely platform-independent; they just mean
1641 that the human-written files (<filename>configure.ac</filename> and
1642 <filename>aclocal.m4</filename>) can be short, although the resulting
1643 files (the <command>configure</command> shell scripts and the C header
1644 template <filename>mk/config.h.in</filename>) are long.</para>
1649 <term>Step 2: system configuration.</term>
1651 <para>Runs the newly-created <command>configure</command>
1652 script, thus:</para>
1654 <programlisting>./configure <optional><parameter>args</parameter></optional></programlisting>
1656 <para><command>configure</command>'s mission is to scurry
1657 round your computer working out what architecture it has,
1658 what operating system, whether it has the
1659 <function>vfork</function> system call, where
1660 <command>tar</command> is kept, whether
1661 <command>gcc</command> is available, where various obscure
1662 <literal>#include</literal> files are, whether it's a
1663 leap year, and what the systems manager had for lunch. It
1664 communicates these snippets of information in two
1671 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1673 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1674 substituting for things between
1675 “<literal>@</literal>” brackets. So,
1676 “<literal>@HaveGcc@</literal>” will be
1677 replaced by “<literal>YES</literal>” or
1678 “<literal>NO</literal>” depending on what
1679 <command>configure</command> finds.
1680 <filename>mk/config.mk</filename> is included by every
1681 Makefile (directly or indirectly), so the
1682 configuration information is thereby communicated to
1683 all Makefiles.</para>
1687 <para> It translates
1688 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1690 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1691 The latter is <literal>#include</literal>d by
1692 various C programs, which can thereby make use of
1693 configuration information.</para>
1697 <para><command>configure</command> takes some optional
1698 arguments. Use <literal>./configure --help</literal> to
1699 get a list of the available arguments. Here are some of
1700 the ones you might need:</para>
1704 <term><literal>--with-ghc=<parameter>path</parameter></literal>
1705 <indexterm><primary><literal>--with-ghc</literal></primary></indexterm>
1708 <para>Specifies the path to an installed GHC which
1709 you would like to use. This compiler will be used
1710 for compiling GHC-specific code (eg. GHC itself).
1711 This option <emphasis>cannot</emphasis> be specified
1712 using <filename>build.mk</filename> (see later),
1713 because <command>configure</command> needs to
1714 auto-detect the version of GHC you're using. The
1715 default is to look for a compiler named
1716 <literal>ghc</literal> in your path.</para>
1721 <term><literal>--with-hc=<parameter>path</parameter></literal>
1722 <indexterm><primary><literal>--with-hc</literal></primary></indexterm>
1725 <para>Specifies the path to any installed Haskell
1726 compiler. This compiler will be used for compiling
1727 generic Haskell code. The default is to use
1728 <literal>ghc</literal>.</para>
1733 <term><literal>--with-gcc=<parameter>path</parameter></literal>
1734 <indexterm><primary><literal>--with-gcc</literal></primary></indexterm>
1737 <para>Specifies the path to the installed GCC. This
1738 compiler will be used to compile all C files,
1739 <emphasis>except</emphasis> any generated by the
1740 installed Haskell compiler, which will have its own
1741 idea of which C compiler (if any) to use. The
1742 default is to use <literal>gcc</literal>.</para>
1750 <term>Step 3: build configuration.</term>
1752 <para>Next, you say how this build of
1753 <literal>fptools</literal> is to differ from the standard
1754 defaults by creating a new file
1755 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1756 <emphasis>in the build tree</emphasis>. This file is the
1757 one and only file you edit in the build tree, precisely
1758 because it says how this build differs from the source.
1759 (Just in case your build tree does die, you might want to
1760 keep a private directory of <filename>build.mk</filename>
1761 files, and use a symbolic link in each build tree to point
1762 to the appropriate one.) So
1763 <filename>mk/build.mk</filename> never exists in the
1764 source tree—you create one in each build tree from
1765 the template. We'll discuss what to put in it
1771 <para>And that's it for configuration. Simple, eh?</para>
1773 <para>What do you put in your build-specific configuration file
1774 <filename>mk/build.mk</filename>? <emphasis>For almost all
1775 purposes all you will do is put make variable definitions that
1776 override those in</emphasis>
1777 <filename>mk/config.mk.in</filename>. The whole point of
1778 <filename>mk/config.mk.in</filename>—and its derived
1779 counterpart <filename>mk/config.mk</filename>—is to define
1780 the build configuration. It is heavily commented, as you will
1781 see if you look at it. So generally, what you do is look at
1782 <filename>mk/config.mk.in</filename>, and add definitions in
1783 <filename>mk/build.mk</filename> that override any of the
1784 <filename>config.mk</filename> definitions that you want to
1785 change. (The override occurs because the main boilerplate file,
1786 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1787 includes <filename>build.mk</filename> after
1788 <filename>config.mk</filename>.)</para>
1790 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1791 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1793 <para>For example, <filename>config.mk.in</filename> contains
1794 the definition:</para>
1796 <programlisting>GhcHcOpts=-O -Rghc-timing</programlisting>
1798 <para>The accompanying comment explains that this is the list of
1799 flags passed to GHC when building GHC itself. For doing
1800 development, it is wise to add <literal>-DDEBUG</literal>, to
1801 enable debugging code. So you would add the following to
1802 <filename>build.mk</filename>:</para>
1804 <para>or, if you prefer,</para>
1806 <programlisting>GhcHcOpts += -DDEBUG</programlisting>
1808 <para>GNU <command>make</command> allows existing definitions to
1809 have new text appended using the “<literal>+=</literal>”
1810 operator, which is quite a convenient feature.)</para>
1812 <para>If you want to remove the <literal>-O</literal> as well (a
1813 good idea when developing, because the turn-around cycle gets a
1814 lot quicker), you can just override
1815 <literal>GhcLibHcOpts</literal> altogether:</para>
1817 <programlisting>GhcHcOpts=-DDEBUG -Rghc-timing</programlisting>
1819 <para>When reading <filename>config.mk.in</filename>, remember
1820 that anything between “@...@” signs is going to be substituted
1821 by <command>configure</command> later. You
1822 <emphasis>can</emphasis> override the resulting definition if
1823 you want, but you need to be a bit surer what you are doing.
1824 For example, there's a line that says:</para>
1826 <programlisting>TAR = @TarCmd@</programlisting>
1828 <para>This defines the Make variables <constant>TAR</constant>
1829 to the pathname for a <command>tar</command> that
1830 <command>configure</command> finds somewhere. If you have your
1831 own pet <command>tar</command> you want to use instead, that's
1832 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1834 <programlisting>TAR = mytar</programlisting>
1836 <para>You do not <emphasis>have</emphasis> to have a
1837 <filename>mk/build.mk</filename> file at all; if you don't,
1838 you'll get all the default settings from
1839 <filename>mk/config.mk.in</filename>.</para>
1841 <para>You can also use <filename>build.mk</filename> to override
1842 anything that <command>configure</command> got wrong. One place
1843 where this happens often is with the definition of
1844 <constant>FPTOOLS_TOP_ABS</constant>: this
1845 variable is supposed to be the canonical path to the top of your
1846 source tree, but if your system uses an automounter then the
1847 correct directory is hard to find automatically. If you find
1848 that <command>configure</command> has got it wrong, just put the
1849 correct definition in <filename>build.mk</filename>.</para>
1853 <sect2 id="sec-storysofar">
1854 <title>The story so far</title>
1856 <para>Let's summarise the steps you need to carry to get
1857 yourself a fully-configured build tree from scratch.</para>
1861 <para> Get your source tree from somewhere (CVS repository
1862 or source distribution). Say you call the root directory
1863 <filename>myfptools</filename> (it does not have to be
1864 called <filename>fptools</filename>). Make sure that you
1865 have the essential files (see <xref
1866 linkend="sec-source-tree"/>).</para>
1871 <para>(Optional) Use <command>lndir</command> or
1872 <command>mkshadowdir</command> to create a build tree.</para>
1874 <programlisting>$ cd myfptools
1875 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4</programlisting>
1877 <para>(N.B. <command>mkshadowdir</command>'s first argument
1878 is taken relative to its second.) You probably want to give
1879 the build tree a name that suggests its main defining
1880 characteristic (in your mind at least), in case you later
1885 <para>Change directory to the build tree. Everything is
1886 going to happen there now.</para>
1888 <programlisting>$ cd /scratch/joe-bloggs/myfptools-sun4</programlisting>
1893 <para>Prepare for system configuration:</para>
1895 <programlisting>$ autoreconf</programlisting>
1897 <para>(You can skip this step if you are starting from a
1898 source distribution, and you already have
1899 <filename>configure</filename> and
1900 <filename>mk/config.h.in</filename>.)</para>
1904 <para>Do system configuration:</para>
1906 <programlisting>$ ./configure</programlisting>
1908 <para>Don't forget to check whether you need to add any
1909 arguments to <literal>configure</literal>; for example, a
1910 common requirement is to specify which GHC to use with
1911 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1915 <para>Create the file <filename>mk/build.mk</filename>,
1916 adding definitions for your desired configuration
1919 <programlisting>$ emacs mk/build.mk</programlisting>
1923 <para>You can make subsequent changes to
1924 <filename>mk/build.mk</filename> as often as you like. You do
1925 not have to run any further configuration programs to make these
1926 changes take effect. In theory you should, however, say
1927 <command>gmake clean</command>, <command>gmake all</command>,
1928 because configuration option changes could affect
1929 anything—but in practice you are likely to know what's
1934 <title>Making things</title>
1936 <para>At this point you have made yourself a fully-configured
1937 build tree, so you are ready to start building real
1940 <para>The first thing you need to know is that <emphasis>you
1941 must use GNU <command>make</command>, usually called
1942 <command>gmake</command>, not standard Unix
1943 <command>make</command></emphasis>. If you use standard Unix
1944 <command>make</command> you will get all sorts of error messages
1945 (but no damage) because the <literal>fptools</literal>
1946 <command>Makefiles</command> use GNU <command>make</command>'s
1947 facilities extensively.</para>
1949 <para>To just build the whole thing, <command>cd</command> to
1950 the top of your <literal>fptools</literal> tree and type
1951 <command>gmake</command>. This will prepare the tree and build
1952 the various projects in the correct order.</para>
1955 <sect2 id="sec-bootstrapping">
1956 <title>Bootstrapping GHC</title>
1958 <para>GHC requires a 2-stage bootstrap in order to provide
1959 full functionality, including GHCi. By a 2-stage bootstrap, we
1960 mean that the compiler is built once using the installed GHC,
1961 and then again using the compiler built in the first stage. You
1962 can also build a stage 3 compiler, but this normally isn't
1963 necessary except to verify that the stage 2 compiler is working
1966 <para>Note that when doing a bootstrap, the stage 1 compiler
1967 must be built, followed by the runtime system and libraries, and
1968 then the stage 2 compiler. The correct ordering is implemented
1969 by the top-level fptools <filename>Makefile</filename>, so if
1970 you want everything to work automatically it's best to start
1971 <command>make</command> from the top of the tree. When building
1972 GHC, the top-level fptools <filename>Makefile</filename> is set
1973 up to do a 2-stage bootstrap by default (when you say
1974 <command>make</command>). Some other targets it supports
1981 <para>Build everything as normal, including the stage 1
1989 <para>Build the stage 2 compiler only.</para>
1996 <para>Build the stage 3 compiler only.</para>
2001 <term>bootstrap</term> <term>bootstrap2</term>
2003 <para>Build stage 1 followed by stage 2.</para>
2008 <term>bootstrap3</term>
2010 <para>Build stages 1, 2 and 3.</para>
2015 <term>install</term>
2017 <para>Install everything, including the compiler built in
2018 stage 2. To override the stage, say <literal>make install
2019 stage=<replaceable>n</replaceable></literal> where
2020 <replaceable>n</replaceable> is the stage to install.</para>
2025 <para>The top-level <filename>Makefile</filename> also arranges
2026 to do the appropriate <literal>make boot</literal> steps (see
2027 below) before actually building anything.</para>
2029 <para>The <literal>stage1</literal>, <literal>stage2</literal>
2030 and <literal>stage3</literal> targets also work in the
2031 <literal>ghc/compiler</literal> directory, but don't forget that
2032 each stage requires its own <literal>make boot</literal> step:
2033 for example, you must do</para>
2035 <screen>$ make boot stage=2</screen>
2037 <para>before <literal>make stage2</literal> in
2038 <literal>ghc/compiler</literal>.</para>
2041 <sect2 id="sec-standard-targets">
2042 <title>Standard Targets</title>
2043 <indexterm><primary>targets, standard makefile</primary></indexterm>
2044 <indexterm><primary>makefile targets</primary></indexterm>
2046 <para>In any directory you should be able to make the following:</para>
2050 <term><literal>boot</literal></term>
2052 <para>does the one-off preparation required to get ready
2053 for the real work. Notably, it does <command>gmake
2054 depend</command> in all directories that contain programs.
2055 It also builds the necessary tools for compilation to
2058 <para>Invoking the <literal>boot</literal> target
2059 explicitly is not normally necessary. From the top-level
2060 <literal>fptools</literal> directory, invoking
2061 <literal>gmake</literal> causes <literal>gmake boot
2062 all</literal> to be invoked in each of the project
2063 subdirectories, in the order specified by
2064 <literal>$(AllTargets)</literal> in
2065 <literal>config.mk</literal>.</para>
2067 <para>If you're working in a subdirectory somewhere and
2068 need to update the dependencies, <literal>gmake
2069 boot</literal> is a good way to do it.</para>
2074 <term><literal>all</literal></term>
2076 <para>makes all the final target(s) for this Makefile.
2077 Depending on which directory you are in a “final
2078 target” may be an executable program, a library
2079 archive, a shell script, or a Postscript file. Typing
2080 <command>gmake</command> alone is generally the same as
2081 typing <command>gmake all</command>.</para>
2086 <term><literal>install</literal></term>
2088 <para>installs the things built by <literal>all</literal>
2089 (except for the documentation). Where does it install
2090 them? That is specified by
2091 <filename>mk/config.mk.in</filename>; you can override it
2092 in <filename>mk/build.mk</filename>, or by running
2093 <command>configure</command> with command-line arguments
2094 like <literal>--bindir=/home/simonpj/bin</literal>; see
2095 <literal>./configure --help</literal> for the full
2101 <term><literal>install-docs</literal></term>
2103 <para>installs the documentation. Otherwise behaves just
2104 like <literal>install</literal>.</para>
2109 <term><literal>uninstall</literal></term>
2111 <para>reverses the effect of
2112 <literal>install</literal>.</para>
2117 <term><literal>clean</literal></term>
2119 <para>Delete all files from the current directory that are
2120 normally created by building the program. Don't delete
2121 the files that record the configuration, or files
2122 generated by <command>gmake boot</command>. Also preserve
2123 files that could be made by building, but normally aren't
2124 because the distribution comes with them.</para>
2129 <term><literal>distclean</literal></term>
2131 <para>Delete all files from the current directory that are
2132 created by configuring or building the program. If you
2133 have unpacked the source and built the program without
2134 creating any other files, <literal>make
2135 distclean</literal> should leave only the files that were
2136 in the distribution.</para>
2141 <term><literal>mostlyclean</literal></term>
2143 <para>Like <literal>clean</literal>, but may refrain from
2144 deleting a few files that people normally don't want to
2150 <term><literal>maintainer-clean</literal></term>
2152 <para>Delete everything from the current directory that
2153 can be reconstructed with this Makefile. This typically
2154 includes everything deleted by
2155 <literal>distclean</literal>, plus more: C source files
2156 produced by Bison, tags tables, Info files, and so
2159 <para>One exception, however: <literal>make
2160 maintainer-clean</literal> should not delete
2161 <filename>configure</filename> even if
2162 <filename>configure</filename> can be remade using a rule
2163 in the <filename>Makefile</filename>. More generally,
2164 <literal>make maintainer-clean</literal> should not delete
2165 anything that needs to exist in order to run
2166 <filename>configure</filename> and then begin to build the
2172 <term><literal>check</literal></term>
2174 <para>run the test suite.</para>
2179 <para>All of these standard targets automatically recurse into
2180 sub-directories. Certain other standard targets do not:</para>
2184 <term><literal>configure</literal></term>
2186 <para>is only available in the root directory
2187 <constant>$(FPTOOLS_TOP)</constant>; it has
2188 been discussed in <xref
2189 linkend="sec-build-config"/>.</para>
2194 <term><literal>depend</literal></term>
2196 <para>make a <filename>.depend</filename> file in each
2197 directory that needs it. This <filename>.depend</filename>
2198 file contains mechanically-generated dependency
2199 information; for example, suppose a directory contains a
2200 Haskell source module <filename>Foo.lhs</filename> which
2201 imports another module <literal>Baz</literal>. Then the
2202 generated <filename>.depend</filename> file will contain
2203 the dependency:</para>
2205 <programlisting>Foo.o : Baz.hi</programlisting>
2207 <para>which says that the object file
2208 <filename>Foo.o</filename> depends on the interface file
2209 <filename>Baz.hi</filename> generated by compiling module
2210 <literal>Baz</literal>. The <filename>.depend</filename>
2211 file is automatically included by every Makefile.</para>
2216 <term><literal>binary-dist</literal></term>
2218 <para>make a binary distribution. This is the target we
2219 use to build the binary distributions of GHC and
2225 <term><literal>dist</literal></term>
2227 <para>make a source distribution. Note that this target
2228 does “make distclean” as part of its work;
2229 don't use it if you want to keep what you've built.</para>
2234 <para>Most <filename>Makefile</filename>s have targets other
2235 than these. You can discover them by looking in the
2236 <filename>Makefile</filename> itself.</para>
2240 <title>Using a project from the build tree</title>
2242 <para>If you want to build GHC (say) and just use it direct from
2243 the build tree without doing <literal>make install</literal>
2244 first, you can run the in-place driver script:
2245 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2247 <para> Do <emphasis>NOT</emphasis> use
2248 <filename>ghc/compiler/ghc</filename>, or
2249 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2250 scripts intended for installation, and contain hard-wired paths
2251 to the installed libraries, rather than the libraries in the
2254 <para>Happy can similarly be run from the build tree, using
2255 <filename>happy/src/happy-inplace</filename>, and similarly for
2256 Alex and Haddock.</para>
2260 <title>Fast Making</title>
2262 <indexterm><primary>fastmake</primary></indexterm>
2263 <indexterm><primary>dependencies, omitting</primary></indexterm>
2264 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2266 <para>Sometimes the dependencies get in the way: if you've made
2267 a small change to one file, and you're absolutely sure that it
2268 won't affect anything else, but you know that
2269 <command>make</command> is going to rebuild everything anyway,
2270 the following hack may be useful:</para>
2272 <programlisting>gmake FAST=YES</programlisting>
2274 <para>This tells the make system to ignore dependencies and just
2275 build what you tell it to. In other words, it's equivalent to
2276 temporarily removing the <filename>.depend</filename> file in
2277 the current directory (where <command>mkdependHS</command> and
2278 friends store their dependency information).</para>
2280 <para>A bit of history: GHC used to come with a
2281 <command>fastmake</command> script that did the above job, but
2282 GNU make provides the features we need to do it without
2283 resorting to a script. Also, we've found that fastmaking is
2284 less useful since the advent of GHC's recompilation checker (see
2285 the User's Guide section on "Separate Compilation").</para>
2289 <sect1 id="sec-makefile-arch">
2290 <title>The <filename>Makefile</filename> architecture</title>
2291 <indexterm><primary>makefile architecture</primary></indexterm>
2293 <para><command>make</command> is great if everything
2294 works—you type <command>gmake install</command> and lo! the
2295 right things get compiled and installed in the right places. Our
2296 goal is to make this happen often, but somehow it often doesn't;
2297 instead some weird error message eventually emerges from the
2298 bowels of a directory you didn't know existed.</para>
2300 <para>The purpose of this section is to give you a road-map to
2301 help you figure out what is going right and what is going
2305 <title>Debugging</title>
2307 <para>Debugging <filename>Makefile</filename>s is something of a
2308 black art, but here's a couple of tricks that we find
2309 particularly useful. The following command allows you to see
2310 the contents of any make variable in the context of the current
2311 <filename>Makefile</filename>:</para>
2313 <screen>$ make show VALUE=HS_SRCS</screen>
2315 <para>where you can replace <literal>HS_SRCS</literal> with the
2316 name of any variable you wish to see the value of.</para>
2318 <para>GNU make has a <option>-d</option> option which generates
2319 a dump of the decision procedure used to arrive at a conclusion
2320 about which files should be recompiled. Sometimes useful for
2321 tracking down problems with superfluous or missing
2322 recompilations.</para>
2326 <title>A small project</title>
2328 <para>To get started, let us look at the
2329 <filename>Makefile</filename> for an imaginary small
2330 <literal>fptools</literal> project, <literal>small</literal>.
2331 Each project in <literal>fptools</literal> has its own directory
2332 in <constant>FPTOOLS_TOP</constant>, so the
2333 <literal>small</literal> project will have its own directory
2334 <constant>FPOOLS_TOP/small/</constant>. Inside the
2335 <filename>small/</filename> directory there will be a
2336 <filename>Makefile</filename>, looking something like
2339 <indexterm><primary>Makefile, minimal</primary></indexterm>
2341 <programlisting># Makefile for fptools project "small"
2344 include $(TOP)/mk/boilerplate.mk
2346 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2349 include $(TOP)/target.mk</programlisting>
2351 <para>this <filename>Makefile</filename> has three
2356 <para>The first section includes
2359 One of the most important
2360 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2361 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2366 a file of “boilerplate” code from the level
2367 above (which in this case will be
2368 <filename>FPTOOLS_TOP/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2369 As its name suggests, <filename>boilerplate.mk</filename>
2370 consists of a large quantity of standard
2371 <filename>Makefile</filename> code. We discuss this
2372 boilerplate in more detail in <xref linkend="sec-boiler"/>.
2373 <indexterm><primary>include, directive in
2374 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2375 inclusion</primary></indexterm></para>
2377 <para>Before the <literal>include</literal> statement, you
2378 must define the <command>make</command> variable
2379 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2380 to be the directory containing the <filename>mk</filename>
2381 directory in which the <filename>boilerplate.mk</filename>
2382 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2384 <programlisting>include ../mk/boilerplate.mk # NO NO NO</programlisting>
2387 <para>Why? Because the <filename>boilerplate.mk</filename>
2388 file needs to know where it is, so that it can, in turn,
2389 <literal>include</literal> other files. (Unfortunately,
2390 when an <literal>include</literal>d file does an
2391 <literal>include</literal>, the filename is treated relative
2392 to the directory in which <command>gmake</command> is being
2393 run, not the directory in which the
2394 <literal>include</literal>d sits.) In general,
2395 <emphasis>every file <filename>foo.mk</filename> assumes
2397 <filename>$(TOP)/mk/foo.mk</filename>
2398 refers to itself.</emphasis> It is up to the
2399 <filename>Makefile</filename> doing the
2400 <literal>include</literal> to ensure this is the case.</para>
2402 <para>Files intended for inclusion in other
2403 <filename>Makefile</filename>s are written to have the
2404 following property: <emphasis>after
2405 <filename>foo.mk</filename> is <literal>include</literal>d,
2406 it leaves <constant>TOP</constant> containing the same value
2407 as it had just before the <literal>include</literal>
2408 statement</emphasis>. In our example, this invariant
2409 guarantees that the <literal>include</literal> for
2410 <filename>target.mk</filename> will look in the same
2411 directory as that for <filename>boilerplate.mk</filename>.</para>
2415 <para> The second section defines the following standard
2416 <command>make</command> variables:
2417 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2418 (the source files from which is to be built), and
2419 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2420 (the executable binary to be built). We will discuss in
2421 more detail what the “standard variables” are,
2422 and how they affect what happens, in <xref
2423 linkend="sec-targets"/>.</para>
2425 <para>The definition for <constant>SRCS</constant> uses the
2426 useful GNU <command>make</command> construct
2427 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2428 which expands to a list of all the files matching the
2429 pattern <literal>pat</literal> in the current directory. In
2430 this example, <constant>SRCS</constant> is set to the list
2431 of all the <filename>.lhs</filename> and
2432 <filename>.c</filename> files in the directory. (Let's
2433 suppose there is one of each, <filename>Foo.lhs</filename>
2434 and <filename>Baz.c</filename>.)</para>
2438 <para>The last section includes a second file of standard
2440 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2441 It contains the rules that tell <command>gmake</command> how
2442 to make the standard targets (<xref
2443 linkend="sec-standard-targets"/>). Why, you ask, can't this
2444 standard code be part of
2445 <filename>boilerplate.mk</filename>? Good question. We
2446 discuss the reason later, in <xref
2447 linkend="sec-boiler-arch"/>.</para>
2449 <para>You do not <emphasis>have</emphasis> to
2450 <literal>include</literal> the
2451 <filename>target.mk</filename> file. Instead, you can write
2452 rules of your own for all the standard targets. Usually,
2453 though, you will find quite a big payoff from using the
2454 canned rules in <filename>target.mk</filename>; the price
2455 tag is that you have to understand what canned rules get
2456 enabled, and what they do (<xref
2457 linkend="sec-targets"/>).</para>
2461 <para>In our example <filename>Makefile</filename>, most of the
2462 work is done by the two <literal>include</literal>d files. When
2463 you say <command>gmake all</command>, the following things
2468 <para><command>gmake</command> figures out that the object
2469 files are <filename>Foo.o</filename> and
2470 <filename>Baz.o</filename>.</para>
2474 <para>It uses a boilerplate pattern rule to compile
2475 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2476 using a Haskell compiler. (Which one? That is set in the
2477 build configuration.)</para>
2481 <para>It uses another standard pattern rule to compile
2482 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2483 using a C compiler. (Ditto.)</para>
2487 <para>It links the resulting <filename>.o</filename> files
2488 together to make <literal>small</literal>, using the Haskell
2489 compiler to do the link step. (Why not use
2490 <command>ld</command>? Because the Haskell compiler knows
2491 what standard libraries to link in. How did
2492 <command>gmake</command> know to use the Haskell compiler to
2493 do the link, rather than the C compiler? Because we set the
2494 variable <constant>HS_PROG</constant> rather than
2495 <constant>C_PROG</constant>.)</para>
2499 <para>All <filename>Makefile</filename>s should follow the above
2500 three-section format.</para>
2504 <title>A larger project</title>
2506 <para>Larger projects are usually structured into a number of
2507 sub-directories, each of which has its own
2508 <filename>Makefile</filename>. (In very large projects, this
2509 sub-structure might be iterated recursively, though that is
2510 rare.) To give you the idea, here's part of the directory
2511 structure for the (rather large) GHC project:</para>
2513 <screen>$(FPTOOLS_TOP)/ghc/
2520 ...source files for documentation...
2523 ...source files for driver...
2526 parser/...source files for parser...
2527 renamer/...source files for renamer...
2530 <para>The sub-directories <filename>docs</filename>,
2531 <filename>driver</filename>, <filename>compiler</filename>, and
2532 so on, each contains a sub-component of GHC, and each has its
2533 own <filename>Makefile</filename>. There must also be a
2534 <filename>Makefile</filename> in
2535 <filename>$(FPTOOLS_TOP)/ghc</filename>.
2536 It does most of its work by recursively invoking
2537 <command>gmake</command> on the <filename>Makefile</filename>s
2538 in the sub-directories. We say that
2539 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2540 <filename>Makefile</filename></emphasis>, because it does little
2541 except organise its children, while the
2542 <filename>Makefile</filename>s in the sub-directories are all
2543 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2544 principle the sub-directories might themselves contain a
2545 non-leaf <filename>Makefile</filename> and several
2546 sub-sub-directories, but that does not happen in GHC.)</para>
2548 <para>The <filename>Makefile</filename> in
2549 <filename>ghc/compiler</filename> is considered a leaf
2550 <filename>Makefile</filename> even though the
2551 <filename>ghc/compiler</filename> has sub-directories, because
2552 these sub-directories do not themselves have
2553 <filename>Makefile</filename>s in them. They are just used to
2554 structure the collection of modules that make up GHC, but all
2555 are managed by the single <filename>Makefile</filename> in
2556 <filename>ghc/compiler</filename>.</para>
2558 <para>You will notice that <filename>ghc/</filename> also
2559 contains a directory <filename>ghc/mk/</filename>. It contains
2560 GHC-specific <filename>Makefile</filename> boilerplate code.
2561 More precisely:</para>
2565 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2566 at the top of <filename>ghc/Makefile</filename>, and of all
2567 the leaf <filename>Makefile</filename>s in the
2568 sub-directories. It in turn <literal>include</literal>s the
2569 main boilerplate file
2570 <filename>mk/boilerplate.mk</filename>.</para>
2574 <para><filename>ghc/mk/target.mk</filename> is
2575 <literal>include</literal>d at the bottom of
2576 <filename>ghc/Makefile</filename>, and of all the leaf
2577 <filename>Makefile</filename>s in the sub-directories. It
2578 in turn <literal>include</literal>s the file
2579 <filename>mk/target.mk</filename>.</para>
2583 <para>So these two files are the place to look for GHC-wide
2584 customisation of the standard boilerplate.</para>
2587 <sect2 id="sec-boiler-arch">
2588 <title>Boilerplate architecture</title>
2589 <indexterm><primary>boilerplate architecture</primary></indexterm>
2591 <para>Every <filename>Makefile</filename> includes a
2592 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2593 file at the top, and
2594 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2595 file at the bottom. In this section we discuss what is in these
2596 files, and why there have to be two of them. In general:</para>
2600 <para><filename>boilerplate.mk</filename> consists of:</para>
2604 <para><emphasis>Definitions of millions of
2605 <command>make</command> variables</emphasis> that
2606 collectively specify the build configuration. Examples:
2607 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2608 the options to feed to the Haskell compiler;
2609 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2610 the sub-directories to enable within the
2611 <literal>nofib</literal> project;
2612 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2613 the name of the Haskell compiler to use when compiling
2614 GHC in the <literal>ghc</literal> project.</para>
2618 <para><emphasis>Standard pattern rules</emphasis> that
2619 tell <command>gmake</command> how to construct one file
2620 from another.</para>
2624 <para><filename>boilerplate.mk</filename> needs to be
2625 <literal>include</literal>d at the <emphasis>top</emphasis>
2626 of each <filename>Makefile</filename>, so that the user can
2627 replace the boilerplate definitions or pattern rules by
2628 simply giving a new definition or pattern rule in the
2629 <filename>Makefile</filename>. <command>gmake</command>
2630 simply takes the last definition as the definitive one.</para>
2632 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2633 definitions, it is also quite common to
2634 <emphasis>augment</emphasis> them. For example, a
2635 <filename>Makefile</filename> might say:</para>
2637 <programlisting>SRC_HC_OPTS += -O</programlisting>
2639 <para>thereby adding “<option>-O</option>” to
2641 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2645 <para><filename>target.mk</filename> contains
2646 <command>make</command> rules for the standard targets
2647 described in <xref linkend="sec-standard-targets"/>. These
2648 rules are selectively included, depending on the setting of
2649 certain <command>make</command> variables. These variables
2650 are usually set in the middle section of the
2651 <filename>Makefile</filename> between the two
2652 <literal>include</literal>s.</para>
2654 <para><filename>target.mk</filename> must be included at the
2655 end (rather than being part of
2656 <filename>boilerplate.mk</filename>) for several tiresome
2662 <para><command>gmake</command> commits target and
2663 dependency lists earlier than it should. For example,
2664 <filename>target.mk</filename> has a rule that looks
2667 <programlisting>$(HS_PROG) : $(OBJS)
2668 $(HC) $(LD_OPTS) $< -o $@</programlisting>
2670 <para>If this rule was in
2671 <filename>boilerplate.mk</filename> then
2672 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2674 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2675 would not have their final values at the moment
2676 <command>gmake</command> encountered the rule. Alas,
2677 <command>gmake</command> takes a snapshot of their
2678 current values, and wires that snapshot into the rule.
2679 (In contrast, the commands executed when the rule
2680 “fires” are only substituted at the moment
2681 of firing.) So, the rule must follow the definitions
2682 given in the <filename>Makefile</filename> itself.</para>
2686 <para>Unlike pattern rules, ordinary rules cannot be
2687 overriden or replaced by subsequent rules for the same
2688 target (at least, not without an error message).
2689 Including ordinary rules in
2690 <filename>boilerplate.mk</filename> would prevent the
2691 user from writing rules for specific targets in specific
2696 <para>There are a couple of other reasons I've
2697 forgotten, but it doesn't matter too much.</para>
2704 <sect2 id="sec-boiler">
2705 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2706 <indexterm><primary>boilerplate.mk</primary></indexterm>
2708 <para>If you look at
2709 <filename>$(FPTOOLS_TOP)/mk/boilerplate.mk</filename>
2710 you will find that it consists of the following sections, each
2711 held in a separate file:</para>
2715 <term><filename>config.mk</filename>
2716 <indexterm><primary>config.mk</primary></indexterm>
2719 <para>is the build configuration file we discussed at
2720 length in <xref linkend="sec-build-config"/>.</para>
2725 <term><filename>paths.mk</filename>
2726 <indexterm><primary>paths.mk</primary></indexterm>
2729 <para>defines <command>make</command> variables for
2730 pathnames and file lists. This file contains code for
2731 automatically compiling lists of source files and deriving
2732 lists of object files from those. The results can be
2733 overriden in the <filename>Makefile</filename>, but in
2734 most cases the automatic setup should do the right
2737 <para>The following variables may be set in the
2738 <filename>Makefile</filename> to affect how the automatic
2739 source file search is done:</para>
2743 <term><literal>ALL_DIRS</literal>
2744 <indexterm><primary><literal>ALL_DIRS</literal></primary></indexterm>
2747 <para>Set to a list of directories to search in
2748 addition to the current directory for source
2754 <term><literal>EXCLUDE_SRCS</literal>
2755 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary></indexterm>
2758 <para>Set to a list of source files (relative to the
2759 current directory) to omit from the automatic
2760 search. The source searching machinery is clever
2761 enough to know that if you exclude a source file
2762 from which other sources are derived, then the
2763 derived sources should also be excluded. For
2764 example, if you set <literal>EXCLUDED_SRCS</literal>
2765 to include <filename>Foo.y</filename>, then
2766 <filename>Foo.hs</filename> will also be
2772 <term><literal>EXTRA_SRCS</literal>
2773 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary></indexterm>
2776 <para>Set to a list of extra source files (perhaps
2777 in directories not listed in
2778 <literal>ALL_DIRS</literal>) that should be
2784 <para>The results of the automatic source file search are
2785 placed in the following make variables:</para>
2789 <term><literal>SRCS</literal>
2790 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2793 <para>All source files found, sorted and without
2794 duplicates, including those which might not exist
2795 yet but will be derived from other existing sources.
2796 <literal>SRCS</literal> <emphasis>can</emphasis> be
2797 overriden if necessary, in which case the variables
2798 below will follow suit.</para>
2803 <term><literal>HS_SRCS</literal>
2804 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2807 <para>all Haskell source files in the current
2808 directory, including those derived from other source
2809 files (eg. Happy sources also give rise to Haskell
2815 <term><literal>HS_OBJS</literal>
2816 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2819 <para>Object files derived from
2820 <literal>HS_SRCS</literal>.</para>
2825 <term><literal>HS_IFACES</literal>
2826 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2829 <para>Interface files (<literal>.hi</literal> files)
2830 derived from <literal>HS_SRCS</literal>.</para>
2835 <term><literal>C_SRCS</literal>
2836 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2839 <para>All C source files found.</para>
2844 <term><literal>C_OBJS</literal>
2845 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2848 <para>Object files derived from
2849 <literal>C_SRCS</literal>.</para>
2854 <term><literal>SCRIPT_SRCS</literal>
2855 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2858 <para>All script source files found
2859 (<literal>.lprl</literal> files).</para>
2864 <term><literal>SCRIPT_OBJS</literal>
2865 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2868 <para><quote>object</quote> files derived from
2869 <literal>SCRIPT_SRCS</literal>
2870 (<literal>.prl</literal> files).</para>
2875 <term><literal>HSC_SRCS</literal>
2876 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2879 <para>All <literal>hsc2hs</literal> source files
2880 (<literal>.hsc</literal> files).</para>
2885 <term><literal>HAPPY_SRCS</literal>
2886 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2889 <para>All <literal>happy</literal> source files
2890 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2895 <term><literal>OBJS</literal>
2896 <indexterm><primary>OBJS</primary></indexterm>
2899 <para>the concatenation of
2900 <literal>$(HS_OBJS)</literal>,
2901 <literal>$(C_OBJS)</literal>, and
2902 <literal>$(SCRIPT_OBJS)</literal>.</para>
2907 <para>Any or all of these definitions can easily be
2908 overriden by giving new definitions in your
2909 <filename>Makefile</filename>.</para>
2911 <para>What, exactly, does <filename>paths.mk</filename>
2912 consider a <quote>source file</quote> to be? It's based
2913 on the file's suffix (e.g. <filename>.hs</filename>,
2914 <filename>.lhs</filename>, <filename>.c</filename>,
2915 <filename>.hy</filename>, etc), but this is the kind of
2916 detail that changes, so rather than enumerate the source
2917 suffices here the best thing to do is to look in
2918 <filename>paths.mk</filename>.</para>
2923 <term><filename>opts.mk</filename>
2924 <indexterm><primary>opts.mk</primary></indexterm>
2927 <para>defines <command>make</command> variables for option
2928 strings to pass to each program. For example, it defines
2929 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2930 the option strings to pass to the Haskell compiler. See
2931 <xref linkend="sec-suffix"/>.</para>
2936 <term><filename>suffix.mk</filename>
2937 <indexterm><primary>suffix.mk</primary></indexterm>
2940 <para>defines standard pattern rules—see <xref
2941 linkend="sec-suffix"/>.</para>
2946 <para>Any of the variables and pattern rules defined by the
2947 boilerplate file can easily be overridden in any particular
2948 <filename>Makefile</filename>, because the boilerplate
2949 <literal>include</literal> comes first. Definitions after this
2950 <literal>include</literal> directive simply override the default
2951 ones in <filename>boilerplate.mk</filename>.</para>
2954 <sect2 id="sec-suffix">
2955 <title>Pattern rules and options</title>
2956 <indexterm><primary>Pattern rules</primary></indexterm>
2959 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2960 defines standard <emphasis>pattern rules</emphasis> that say how
2961 to build one kind of file from another, for example, how to
2962 build a <filename>.o</filename> file from a
2963 <filename>.c</filename> file. (GNU <command>make</command>'s
2964 <emphasis>pattern rules</emphasis> are more powerful and easier
2965 to use than Unix <command>make</command>'s <emphasis>suffix
2966 rules</emphasis>.)</para>
2968 <para>Almost all the rules look something like this:</para>
2970 <programlisting>%.o : %.c
2972 $(CC) $(CC_OPTS) -c $< -o $@</programlisting>
2974 <para>Here's how to understand the rule. It says that
2975 <emphasis>something</emphasis><filename>.o</filename> (say
2976 <filename>Foo.o</filename>) can be built from
2977 <emphasis>something</emphasis><filename>.c</filename>
2978 (<filename>Foo.c</filename>), by invoking the C compiler (path
2979 name held in <constant>$(CC)</constant>), passing to it
2980 the options <constant>$(CC_OPTS)</constant> and
2981 the rule's dependent file of the rule
2982 <literal>$<</literal> (<filename>Foo.c</filename> in
2983 this case), and putting the result in the rule's target
2984 <literal>$@</literal> (<filename>Foo.o</filename> in this
2987 <para>Every program is held in a <command>make</command>
2988 variable defined in <filename>mk/config.mk</filename>—look
2989 in <filename>mk/config.mk</filename> for the complete list. One
2990 important one is the Haskell compiler, which is called
2991 <constant>$(HC)</constant>.</para>
2993 <para>Every program's options are are held in a
2994 <command>make</command> variables called
2995 <constant><prog>_OPTS</constant>. the
2996 <constant><prog>_OPTS</constant> variables are
2997 defined in <filename>mk/opts.mk</filename>. Almost all of them
2998 are defined like this:</para>
3000 <programlisting>CC_OPTS = \
3001 $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)</programlisting>
3003 <para>The four variables from which
3004 <constant>CC_OPTS</constant> is built have the following
3009 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
3011 <para>options passed to all C compilations.</para>
3016 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
3018 <para>options passed to C compilations for way
3019 <literal><way></literal>. For example,
3020 <constant>WAY_mp_CC_OPTS</constant>
3021 gives options to pass to the C compiler when compiling way
3022 <literal>mp</literal>. The variable
3023 <constant>WAY_CC_OPTS</constant> holds
3024 options to pass to the C compiler when compiling the
3025 standard way. (<xref linkend="sec-ways"/> dicusses
3026 multi-way compilation.)</para>
3031 <term><constant><module>_CC_OPTS</constant>:</term>
3033 <para>options to pass to the C compiler that are specific
3034 to module <literal><module></literal>. For example,
3035 <constant>SMap_CC_OPTS</constant> gives the
3036 specific options to pass to the C compiler when compiling
3037 <filename>SMap.c</filename>.</para>
3042 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3044 <para>extra options to pass to all C compilations. This
3045 is intended for command line use, thus:</para>
3047 <programlisting>gmake libHS.a EXTRA_CC_OPTS="-v"</programlisting>
3053 <sect2 id="sec-targets">
3054 <title>The main <filename>mk/target.mk</filename> file</title>
3055 <indexterm><primary>target.mk</primary></indexterm>
3057 <para><filename>target.mk</filename> contains canned rules for
3058 all the standard targets described in <xref
3059 linkend="sec-standard-targets"/>. It is complicated by the fact
3060 that you don't want all of these rules to be active in every
3061 <filename>Makefile</filename>. Rather than have a plethora of
3062 tiny files which you can include selectively, there is a single
3063 file, <filename>target.mk</filename>, which selectively includes
3064 rules based on whether you have defined certain variables in
3065 your <filename>Makefile</filename>. This section explains what
3066 rules you get, what variables control them, and what the rules
3067 do. Hopefully, you will also get enough of an idea of what is
3068 supposed to happen that you can read and understand any weird
3069 special cases yourself.</para>
3073 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3075 <para>If <constant>HS_PROG</constant> is defined,
3076 you get rules with the following targets:</para>
3080 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3082 <para>itself. This rule links
3083 <constant>$(OBJS)</constant> with the Haskell
3084 runtime system to get an executable called
3085 <constant>$(HS_PROG)</constant>.</para>
3090 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3093 <constant>$(HS_PROG)</constant> in
3094 <constant>$(bindir)</constant>.</para>
3103 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3105 <para>is similar to <constant>HS_PROG</constant>,
3106 except that the link step links
3107 <constant>$(C_OBJS)</constant> with the C
3108 runtime system.</para>
3113 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3115 <para>is similar to <constant>HS_PROG</constant>,
3116 except that it links
3117 <constant>$(LIB_OBJS)</constant> to make the
3118 library archive <constant>$(LIBRARY)</constant>,
3119 and <literal>install</literal> installs it in
3120 <constant>$(libdir)</constant>.</para>
3125 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3127 <para>…</para>
3132 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3134 <para>…</para>
3139 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3141 <para>If <constant>HS_SRCS</constant> is defined
3142 and non-empty, a rule for the target
3143 <literal>depend</literal> is included, which generates
3144 dependency information for Haskell programs. Similarly
3145 for <constant>C_SRCS</constant>.</para>
3150 <para>All of these rules are “double-colon” rules,
3153 <programlisting>install :: $(HS_PROG)
3154 ...how to install it...</programlisting>
3156 <para>GNU <command>make</command> treats double-colon rules as
3157 separate entities. If there are several double-colon rules for
3158 the same target it takes each in turn and fires it if its
3159 dependencies say to do so. This means that you can, for
3160 example, define both <constant>HS_PROG</constant> and
3161 <constant>LIBRARY</constant>, which will generate two rules for
3162 <literal>install</literal>. When you type <command>gmake
3163 install</command> both rules will be fired, and both the program
3164 and the library will be installed, just as you wanted.</para>
3167 <sect2 id="sec-subdirs">
3168 <title>Recursion</title>
3169 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3170 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3172 <para>In leaf <filename>Makefile</filename>s the variable
3173 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3174 is undefined. In non-leaf <filename>Makefile</filename>s,
3175 <constant>SUBDIRS</constant> is set to the list of
3176 sub-directories that contain subordinate
3177 <filename>Makefile</filename>s. <emphasis>It is up to you to
3178 set <constant>SUBDIRS</constant> in the
3179 <filename>Makefile</filename>.</emphasis> There is no automation
3180 here—<constant>SUBDIRS</constant> is too important to
3183 <para>When <constant>SUBDIRS</constant> is defined,
3184 <filename>target.mk</filename> includes a rather neat rule for
3185 the standard targets (<xref linkend="sec-standard-targets"/> that
3186 simply invokes <command>make</command> recursively in each of
3187 the sub-directories.</para>
3189 <para><emphasis>These recursive invocations are guaranteed to
3190 occur in the order in which the list of directories is specified
3191 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3192 be important. For example, when you say <command>gmake
3193 boot</command> it can be important that the recursive invocation
3194 of <command>make boot</command> is done in one sub-directory
3195 (the include files, say) before another (the source files).
3196 Generally, put the most independent sub-directory first, and the
3197 most dependent last.</para>
3200 <sect2 id="sec-ways">
3201 <title>Way management</title>
3202 <indexterm><primary>way management</primary></indexterm>
3204 <para>We sometimes want to build essentially the same system in
3205 several different “ways”. For example, we want to build GHC's
3206 <literal>Prelude</literal> libraries with and without profiling,
3207 so that there is an appropriately-built library archive to link
3208 with when the user compiles his program. It would be possible
3209 to have a completely separate build tree for each such “way”,
3210 but it would be horribly bureaucratic, especially since often
3211 only parts of the build tree need to be constructed in multiple
3215 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3216 contains some clever magic to allow you to build several
3217 versions of a system; and to control locally how many versions
3218 are built and how they differ. This section explains the
3221 <para>The files for a particular way are distinguished by
3222 munging the suffix. The <quote>normal way</quote> is always
3223 built, and its files have the standard suffices
3224 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3225 In addition, you can build one or more extra ways, each
3226 distinguished by a <emphasis>way tag</emphasis>. The object
3227 files and interface files for one of these extra ways are
3228 distinguished by their suffix. For example, way
3229 <literal>mp</literal> has files
3230 <filename>.mp_o</filename> and
3231 <filename>.mp_hi</filename>. Library archives have their
3232 way tag the other side of the dot, for boring reasons; thus,
3233 <filename>libHS_mp.a</filename>.</para>
3235 <para>A <command>make</command> variable called
3236 <constant>way</constant> holds the current way tag.
3237 <emphasis><constant>way</constant> is only ever set on the
3238 command line of <command>gmake</command></emphasis> (usually in
3239 a recursive invocation of <command>gmake</command> by the
3240 system). It is never set inside a
3241 <filename>Makefile</filename>. So it is a global constant for
3242 any one invocation of <command>gmake</command>. Two other
3243 <command>make</command> variables,
3244 <constant>way_</constant> and
3245 <constant>_way</constant> are immediately derived from
3246 <constant>$(way)</constant> and never altered. If
3247 <constant>way</constant> is not set, then neither are
3248 <constant>way_</constant> and
3249 <constant>_way</constant>, and the invocation of
3250 <command>make</command> will build the <quote>normal
3251 way</quote>. If <constant>way</constant> is set, then the other
3252 two variables are set in sympathy. For example, if
3253 <constant>$(way)</constant> is “<literal>mp</literal>”,
3254 then <constant>way_</constant> is set to
3255 “<literal>mp_</literal>” and
3256 <constant>_way</constant> is set to
3257 “<literal>_mp</literal>”. These three variables are
3258 then used when constructing file names.</para>
3260 <para>So how does <command>make</command> ever get recursively
3261 invoked with <constant>way</constant> set? There are two ways
3262 in which this happens:</para>
3266 <para>For some (but not all) of the standard targets, when
3267 in a leaf sub-directory, <command>make</command> is
3268 recursively invoked for each way tag in
3269 <constant>$(WAYS)</constant>. You set
3270 <constant>WAYS</constant> in the
3271 <filename>Makefile</filename> to the list of way tags you
3272 want these targets built for. The mechanism here is very
3273 much like the recursive invocation of
3274 <command>make</command> in sub-directories (<xref
3275 linkend="sec-subdirs"/>). It is up to you to set
3276 <constant>WAYS</constant> in your
3277 <filename>Makefile</filename>; this is how you control what
3278 ways will get built.</para>
3282 <para>For a useful collection of targets (such as
3283 <filename>libHS_mp.a</filename>,
3284 <filename>Foo.mp_o</filename>) there is a rule which
3285 recursively invokes <command>make</command> to make the
3286 specified target, setting the <constant>way</constant>
3287 variable. So if you say <command>gmake
3288 Foo.mp_o</command> you should see a recursive
3289 invocation <command>gmake Foo.mp_o way=mp</command>,
3290 and <emphasis>in this recursive invocation the pattern rule
3291 for compiling a Haskell file into a <filename>.o</filename>
3292 file will match</emphasis>. The key pattern rules (in
3293 <filename>suffix.mk</filename>) look like this:
3295 <programlisting>%.$(way_)o : %.lhs
3296 $(HC) $(HC_OPTS) $< -o $@</programlisting>
3302 <para>You can invoke <command>make</command> with a
3303 particular <literal>way</literal> setting yourself, in order
3304 to build files related to a particular
3305 <literal>way</literal> in the current directory. eg.
3307 <screen>$ make way=p</screen>
3309 will build files for the profiling way only in the current
3316 <title>When the canned rule isn't right</title>
3318 <para>Sometimes the canned rule just doesn't do the right thing.
3319 For example, in the <literal>nofib</literal> suite we want the
3320 link step to print out timing information. The thing to do here
3321 is <emphasis>not</emphasis> to define
3322 <constant>HS_PROG</constant> or
3323 <constant>C_PROG</constant>, and instead define a special
3324 purpose rule in your own <filename>Makefile</filename>. By
3325 using different variable names you will avoid the canned rules
3326 being included, and conflicting with yours.</para>
3330 <sect1 id="building-docs">
3331 <title>Building the documentation</title>
3333 <sect2 id="pre-supposed-doc-tools">
3334 <title>Tools for building the Documentation</title>
3336 <para>The following additional tools are required if you want to
3337 format the documentation that comes with the
3338 <literal>fptools</literal> projects:</para>
3343 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3344 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3347 <para>Much of our documentation is written in SGML, using
3348 the DocBook DTD. Instructions on installing and
3349 configuring the DocBook tools are below.</para>
3355 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3356 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3359 <para>A decent TeX distribution is required if you want to
3360 produce printable documentation. We recomment teTeX,
3361 which includes just about everything you need.</para>
3367 <indexterm><primary>Haddock</primary></indexterm>
3370 <para>Haddock is a Haskell documentation tool that we use
3371 for automatically generating documentation from the
3372 library source code. It is an <literal>fptools</literal>
3373 project in itself. To build documentation for the
3374 libraries (<literal>fptools/libraries</literal>) you
3375 should check out and build Haddock in
3376 <literal>fptools/haddock</literal>. Haddock requires GHC
3384 <title>Installing the DocBook tools</title>
3387 <title>Installing the DocBook tools on Linux</title>
3389 <para>If you're on a recent RedHat system (7.0+), you probably
3390 have working DocBook tools already installed. The configure
3391 script should detect your setup and you're away.</para>
3393 <para>If you don't have DocBook tools installed, and you are
3394 using a system that can handle RedHat RPM packages, you can
3395 probably use the <ulink
3396 url="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3397 DocBook tools</ulink>, which is the most shrink-wrapped SGML
3398 suite that we could find. You need all the RPMs except for
3399 psgml (i.e. <filename>docbook</filename>,
3400 <filename>jade</filename>, <filename>jadetex</filename>,
3401 <filename>sgmlcommon</filename> and
3402 <filename>stylesheets</filename>). Note that most of these
3403 RPMs are architecture neutral, so are likely to be found in a
3404 <filename>noarch</filename> directory. The SuSE RPMs also
3405 work; the RedHat ones <emphasis>don't</emphasis> in RedHat 6.2
3406 (7.0 and later should be OK), but they are easy to fix: just
3408 <filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</filename>
3409 to <filename>/usr/lib/sgml/lib/dblib.dsl</filename>. </para>
3413 <title>Installing DocBook on FreeBSD</title>
3415 <para>On FreeBSD systems, the easiest way to get DocBook up
3416 and running is to install it from the ports tree or a
3417 pre-compiled package (packages are available from your local
3418 FreeBSD mirror site).</para>
3420 <para>To use the ports tree, do this:
3421 <screen>$ cd /usr/ports/textproc/docproj
3422 $ make install</screen>
3423 This installs the FreeBSD documentation project tools, which
3424 includes everything needed to format the GHC
3425 documentation.</para>
3429 <title>Installing from binaries on Windows</title>
3431 <para>It's a good idea to use Norman Walsh's <ulink
3432 url="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3433 notes</ulink> as a guide. You should get version 3.1 of
3434 DocBook, and note that his file <filename>test.sgm</filename>
3435 won't work, as it needs version 3.0. You should unpack Jade
3436 into <filename>\Jade</filename>, along with the entities,
3437 DocBook into <filename>\docbook</filename>, and the DocBook
3438 stylesheets into <filename>\docbook\stylesheets</filename> (so
3439 they actually end up in
3440 <filename>\docbook\stylesheets\docbook</filename>).</para>
3445 <title>Installing the DocBook tools from source</title>
3450 <para>Install <ulink
3451 url="http://openjade.sourceforge.net/">OpenJade</ulink>
3452 (Windows binaries are available as well as sources). If you
3453 want DVI, PS, or PDF then install JadeTeX from the
3454 <filename>dsssl</filename> subdirectory. (If you get the
3457 <screen>! LaTeX Error: Unknown option implicit=false' for package hyperref'.</screen>
3459 your version of <command>hyperref</command> is out of date;
3460 download it from CTAN
3461 (<filename>macros/latex/contrib/supported/hyperref</filename>),
3462 and make it, ensuring that you have first removed or renamed
3463 your old copy. If you start getting file not found errors
3464 when making the test for <command>hyperref</command>, you
3465 can abort at that point and proceed straight to
3466 <command>make install</command>, or enter them as
3467 <filename>../</filename><emphasis>filename</emphasis>.)</para>
3469 <para>Make links from <filename>virtex</filename> to
3470 <filename>jadetex</filename> and
3471 <filename>pdfvirtex</filename> to
3472 <filename>pdfjadetex</filename> (otherwise DVI, PostScript
3473 and PDF output will not work). Copy
3474 <filename>dsssl/*.{dtd,dsl}</filename> and
3475 <filename>catalog</filename> to
3476 <filename>/usr/[local/]lib/sgml</filename>.</para>
3480 <title>DocBook and the DocBook stylesheets</title>
3482 <para>Get a Zip of <ulink
3483 url="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ulink>
3484 and install the contents in
3485 <filename>/usr/[local/]/lib/sgml</filename>.</para>
3487 <para>Get the <ulink
3488 url="http://nwalsh.com/docbook/dsssl/">DocBook
3489 stylesheets</ulink> and install in
3490 <filename>/usr/[local/]lib/sgml/stylesheets</filename>
3491 (thereby creating a subdirectory docbook). For indexing,
3492 copy or link <filename>collateindex.pl</filename> from the
3493 DocBook stylesheets archive in <filename>bin</filename> into
3494 a directory on your <constant>PATH</constant>.</para>
3496 <para>Download the <ulink
3497 url="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3498 entities</ulink> into
3499 <filename>/usr/[local/]lib/sgml</filename>.</para>
3505 <title>Configuring the DocBook tools</title>
3507 <para>Once the DocBook tools are installed, the configure script
3508 will detect them and set up the build system accordingly. If you
3509 have a system that isn't supported, let us know, and we'll try
3514 <title>Remaining problems</title>
3516 <para>If you install from source, you'll get a pile of warnings
3519 <screen>DTDDECL catalog entries are not supported</screen>
3521 every time you build anything. These can safely be ignored, but
3522 if you find them tedious you can get rid of them by removing all
3523 the <constant>DTDDECL</constant> entries from
3524 <filename>docbook.cat</filename>.</para>
3528 <title>Building the documentation</title>
3530 <para>To build documentation in a certain format, you can
3531 say, for example,</para>
3533 <screen>$ make html</screen>
3535 <para>to build HTML documentation below the current directory.
3536 The available formats are: <literal>dvi</literal>,
3537 <literal>ps</literal>, <literal>pdf</literal>,
3538 <literal>html</literal>, and <literal>rtf</literal>. Note that
3539 not all documentation can be built in all of these formats: HTML
3540 documentation is generally supported everywhere, and DocBook
3541 documentation might support the other formats (depending on what
3542 other tools you have installed).</para>
3544 <para>All of these targets are recursive; that is, saying
3545 <literal>make html</literal> will make HTML docs for all the
3546 documents recursively below the current directory.</para>
3548 <para>Because there are many different formats that the DocBook
3549 documentation can be generated in, you have to select which ones
3550 you want by setting the <literal>SGMLDocWays</literal> variable
3551 to a list of them. For example, in
3552 <filename>build.mk</filename> you might have a line:</para>
3554 <screen>SGMLDocWays = html ps</screen>
3556 <para>This will cause the documentation to be built in the requested
3557 formats as part of the main build (the default is not to build
3558 any documentation at all).</para>
3562 <title>Installing the documentation</title>
3564 <para>To install the documentation, use:</para>
3566 <screen>$ make install-docs</screen>
3568 <para>This will install the documentation into
3569 <literal>$(datadir)</literal> (which defaults to
3570 <literal>$(prefix)/share</literal>). The exception is HTML
3571 documentation, which goes into
3572 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3574 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3575 to a list of formats, the <literal>install-docs</literal> target
3576 won't do anything for SGML documentation.</para>
3582 <sect1 id="sec-porting-ghc">
3583 <title>Porting GHC</title>
3585 <para>This section describes how to port GHC to a currenly
3586 unsupported platform. There are two distinct
3587 possibilities:</para>
3591 <para>The hardware architecture for your system is already
3592 supported by GHC, but you're running an OS that isn't
3593 supported (or perhaps has been supported in the past, but
3594 currently isn't). This is the easiest type of porting job,
3595 but it still requires some careful bootstrapping. Proceed to
3596 <xref linkend="sec-booting-from-hc"/>.</para>
3600 <para>Your system's hardware architecture isn't supported by
3601 GHC. This will be a more difficult port (though by comparison
3602 perhaps not as difficult as porting gcc). Proceed to <xref
3603 linkend="unregisterised-porting"/>.</para>
3607 <sect2 id="sec-booting-from-hc">
3608 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3610 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3611 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3612 <indexterm><primary>porting GHC</primary></indexterm>
3614 <para>Bootstrapping GHC on a system without GHC already
3615 installed is achieved by taking the intermediate C files (known
3616 as HC files) from a GHC compilation on a supported system to the
3617 target machine, and compiling them using gcc to get a working
3620 <para><emphasis>NOTE: GHC versions 5.xx were hard to bootstrap
3621 from C. We recommend using GHC 6.0.1 or
3622 later.</emphasis></para>
3624 <para>HC files are platform-dependent, so you have to get a set
3625 that were generated on similar hardware. There may be some
3626 supplied on the GHC download page, otherwise you'll have to
3627 compile some up yourself, or start from
3628 <emphasis>unregisterised</emphasis> HC files - see <xref
3629 linkend="unregisterised-porting"/>.</para>
3631 <para>The following steps should result in a working GHC build
3632 with full libraries:</para>
3636 <para>Unpack the HC files on top of a fresh source tree
3637 (make sure the source tree version matches the version of
3638 the HC files <emphasis>exactly</emphasis>!). This will
3639 place matching <filename>.hc</filename> files next to the
3640 corresponding Haskell source (<filename>.hs</filename> or
3641 <filename>.lhs</filename>) in the compiler subdirectory
3642 <filename>ghc/compiler</filename> and in the libraries
3643 (subdirectories of <filename>hslibs</filename> and
3644 <literal>libraries</literal>).</para>
3648 <para>The actual build process is fully automated by the
3649 <filename>hc-build</filename> script located in the
3650 <filename>distrib</filename> directory. If you eventually
3651 want to install GHC into the directory
3652 <replaceable>dir</replaceable>, the following
3653 command will execute the whole build process (it won't
3654 install yet):</para>
3656 <screen>foo% distrib/hc-build --prefix=<replaceable>dir</replaceable></screen>
3657 <indexterm><primary>--hc-build</primary></indexterm>
3659 <para>By default, the installation directory is
3660 <filename>/usr/local</filename>. If that is what you want,
3661 you may omit the argument to <filename>hc-build</filename>.
3662 Generally, any option given to <filename>hc-build</filename>
3663 is passed through to the configuration script
3664 <filename>configure</filename>. If
3665 <filename>hc-build</filename> successfully completes the
3666 build process, you can install the resulting system, as
3669 <screen>foo% make install</screen>
3674 <sect2 id="unregisterised-porting">
3675 <title>Porting GHC to a new architecture</title>
3677 <para>The first step in porting to a new architecture is to get
3678 an <firstterm>unregisterised</firstterm> build working. An
3679 unregisterised build is one that compiles via vanilla C only.
3680 By contrast, a registerised build uses the following
3681 architecture-specific hacks for speed:</para>
3685 <para>Global register variables: certain abstract machine
3686 <quote>registers</quote> are mapped to real machine
3687 registers, depending on how many machine registers are
3689 <filename>ghc/includes/MachRegs.h</filename>).</para>
3693 <para>Assembly-mangling: when compiling via C, we feed the
3694 assembly generated by gcc though a Perl script known as the
3695 <firstterm>mangler</firstterm> (see
3696 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3697 mangler rearranges the assembly to support tail-calls and
3698 various other optimisations.</para>
3702 <para>In an unregisterised build, neither of these hacks are
3703 used — the idea is that the C code generated by the
3704 compiler should compile using gcc only. The lack of these
3705 optimisations costs about a factor of two in performance, but
3706 since unregisterised compilation is usually just a step on the
3707 way to a full registerised port, we don't mind too much.</para>
3709 <para>Notes on GHC portability in general: we've tried to stick
3710 to writing portable code in most parts of the system, so it
3711 should compile on any POSIXish system with gcc, but in our
3712 experience most systems differ from the standards in one way or
3713 another. Deal with any problems as they arise - if you get
3714 stuck, ask the experts on
3715 <email>glasgow-haskell-users@haskell.org</email>.</para>
3717 <para>Lots of useful information about the innards of GHC is
3718 available in the <ulink
3719 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3720 Commentary</ulink>, which might be helpful if you run into some
3721 code which needs tweaking for your system.</para>
3724 <title>Cross-compiling to produce an unregisterised GHC</title>
3726 <para>In this section, we explain how to bootstrap GHC on a
3727 new platform, using unregisterised intermediate C files. We
3728 haven't put a great deal of effort into automating this
3729 process, for two reasons: it is done very rarely, and the
3730 process usually requires human intervention to cope with minor
3731 porting issues anyway.</para>
3733 <para>The following step-by-step instructions should result in
3734 a fully working, albeit unregisterised, GHC. Firstly, you
3735 need a machine that already has a working GHC (we'll call this
3736 the <firstterm>host</firstterm> machine), in order to
3737 cross-compile the intermediate C files that we will use to
3738 bootstrap the compiler on the <firstterm>target</firstterm>
3743 <para>On the target machine:</para>
3747 <para>Unpack a source tree (preferably a released
3748 version). We will call the path to the root of this
3749 tree <replaceable>T</replaceable>.</para>
3753 <screen>$ cd <replaceable>T</replaceable>
3754 $ ./configure --enable-hc-boot --enable-hc-boot-unregisterised</screen>
3756 <para>You might need to update
3757 <filename>configure.in</filename> to recognise the new
3758 architecture, and re-generate
3759 <filename>configure</filename> with
3760 <literal>autoreconf</literal>.</para>
3764 <screen>$ cd <replaceable>T</replaceable>/ghc/includes
3765 $ make config.h</screen>
3771 <para>On the host machine:</para>
3775 <para>Unpack a source tree (same released version). Call
3776 this directory <replaceable>H</replaceable>.</para>
3780 <screen>$ cd <replaceable>H</replaceable>
3781 $ ./configure</screen>
3786 <filename><replaceable>H</replaceable>/mk/build.mk</filename>,
3787 with the following contents:</para>
3789 <programlisting>GhcUnregisterised = YES
3790 GhcLibHcOpts = -O -H32m -keep-hc-files
3793 GhcWithNativeCodeGen = NO
3794 GhcWithInterpreter = NO
3795 GhcStage1HcOpts = -O -H32m -fasm
3796 GhcStage2HcOpts = -O -fvia-C -keep-hc-files</programlisting>
3801 <filename><replaceable>H</replaceable>/mk/config.mk</filename>:</para>
3804 <para>change <literal>TARGETPLATFORM</literal>
3805 appropriately, and set the variables involving
3806 <literal>TARGET</literal> to the correct values for
3807 the target platform. This step is necessary because
3808 currently <literal>configure</literal> doesn't cope
3809 with specifying different values for the
3810 <literal>--host</literal> and
3811 <literal>--target</literal> flags.</para>
3814 <para>copy <literal>LeadingUnderscore</literal>
3815 setting from target.</para>
3822 <filename><replaceable>T</replaceable>/ghc/includes/config.h</filename>
3824 <filename><replaceable>H</replaceable>/ghc/includes</filename>.
3825 Note that we are building on the host machine, using the
3826 target machine's <literal>config.h</literal> file. This
3827 is so that the intermediate C files generated here will
3828 be suitable for compiling on the target system.</para>
3833 <para>Touch <literal>config.h</literal>, just to make
3834 sure it doesn't get replaced during the build:</para>
3835 <screen>$ touch <replaceable>H</replaceable>/ghc/includes/config.h</screen>
3839 <para>Now build the compiler:</para>
3840 <screen>$ cd <replaceable>H</replaceable>/glafp-utils && make boot && make
3841 $ cd <replaceable>H</replaceable>/ghc && make boot && make</screen>
3842 <para>Don't worry if the build falls over in the RTS, we
3843 don't need the RTS yet.</para>
3847 <screen>$ cd <replaceable>H</replaceable>/libraries
3848 $ make boot && make</screen>
3852 <screen>$ cd <replaceable>H</replaceable>/ghc
3853 $ make boot stage=2 && make stage=2</screen>
3857 <screen>$ cd <replaceable>H</replaceable>/ghc/utils
3859 $ make -k HC=<replaceable>H</replaceable>/ghc/compiler/stage1/ghc-inplace \
3860 EXTRA_HC_OPTS='-O -fvia-C -keep-hc-files'</screen>
3864 <screen>$ cd <replaceable>H</replaceable>
3865 $ make hc-file-bundle Project=Ghc</screen>
3870 <filename><replaceable>H</replaceable>/*-hc.tar.gz</filename>
3871 to <filename><replaceable>T</replaceable>/..</filename>.</para>
3877 <para>On the target machine:</para>
3879 <para>At this stage we simply need to bootstrap a compiler
3880 from the intermediate C files we generated above. The
3881 process of bootstrapping from C files is automated by the
3882 script in <literal>distrib/hc-build</literal>, and is
3883 described in <xref linkend="sec-booting-from-hc"/>.</para>
3885 <screen>$ ./distrib/hc-build --enable-hc-boot-unregisterised</screen>
3887 <para>However, since this is a bootstrap on a new machine,
3888 the automated process might not run to completion the
3889 first time. For that reason, you might want to treat the
3890 <literal>hc-build</literal> script as a list of
3891 instructions to follow, rather than as a fully automated
3892 script. This way you'll be able to restart the process
3893 part-way through if you need to fix anything on the
3896 <para>Don't bother with running
3897 <literal>make install</literal> in the newly
3898 bootstrapped tree; just use the compiler in that tree to
3899 build a fresh compiler from scratch, this time without
3900 booting from C files. Before doing this, you might want
3901 to check that the bootstrapped compiler is generating
3902 working binaries:</para>
3904 <screen>$ cat >hello.hs
3905 main = putStrLn "Hello World!\n"
3907 $ <replaceable>T</replaceable>/ghc/compiler/ghc-inplace hello.hs -o hello
3909 Hello World!</screen>
3911 <para>Once you have the unregisterised compiler up and
3912 running, you can use it to start a registerised port. The
3913 following sections describe the various parts of the
3914 system that will need architecture-specific tweaks in
3915 order to get a registerised build going.</para>
3922 <title>Porting the RTS</title>
3924 <para>The following files need architecture-specific code for a
3925 registerised build:</para>
3929 <term><filename>ghc/includes/MachRegs.h</filename>
3930 <indexterm><primary><filename>MachRegs.h</filename></primary></indexterm>
3933 <para>Defines the STG-register to machine-register
3934 mapping. You need to know your platform's C calling
3935 convention, and which registers are generally available
3936 for mapping to global register variables. There are
3937 plenty of useful comments in this file.</para>
3941 <term><filename>ghc/includes/TailCalls.h</filename>
3942 <indexterm><primary><filename>TailCalls.h</filename></primary></indexterm>
3945 <para>Macros that cooperate with the mangler (see <xref
3946 linkend="sec-mangler"/>) to make proper tail-calls
3951 <term><filename>ghc/rts/Adjustor.c</filename>
3952 <indexterm><primary><filename>Adjustor.c</filename></primary></indexterm>
3956 <literal>foreign import "wrapper"</literal>
3958 <literal>foreign export dynamic</literal>).
3959 Not essential for getting GHC bootstrapped, so this file
3960 can be deferred until later if necessary.</para>
3964 <term><filename>ghc/rts/StgCRun.c</filename>
3965 <indexterm><primary><filename>StgCRun.c</filename></primary></indexterm>
3968 <para>The little assembly layer between the C world and
3969 the Haskell world. See the comments and code for the
3970 other architectures in this file for pointers.</para>
3974 <term><filename>ghc/rts/MBlock.h</filename>
3975 <indexterm><primary><filename>MBlock.h</filename></primary></indexterm>
3977 <term><filename>ghc/rts/MBlock.c</filename>
3978 <indexterm><primary><filename>MBlock.c</filename></primary></indexterm>
3981 <para>These files are really OS-specific rather than
3982 architecture-specific. In <filename>MBlock.h</filename>
3983 is specified the absolute location at which the RTS
3984 should try to allocate memory on your platform (try to
3985 find an area which doesn't conflict with code or dynamic
3986 libraries). In <filename>Mblock.c</filename> you might
3987 need to tweak the call to <literal>mmap()</literal> for
3994 <sect3 id="sec-mangler">
3995 <title>The mangler</title>
3997 <para>The mangler is an evil Perl-script that rearranges the
3998 assembly code output from gcc to do two main things:</para>
4002 <para>Remove function prologues and epilogues, and all
4003 movement of the C stack pointer. This is to support
4004 tail-calls: every code block in Haskell code ends in an
4005 explicit jump, so we don't want the C-stack overflowing
4006 while we're jumping around between code blocks.</para>
4009 <para>Move the <firstterm>info table</firstterm> for a
4010 closure next to the entry code for that closure. In
4011 unregisterised code, info tables contain a pointer to the
4012 entry code, but in registerised compilation we arrange
4013 that the info table is shoved right up against the entry
4014 code, and addressed backwards from the entry code pointer
4015 (this saves a word in the info table and an extra
4016 indirection when jumping to the closure entry
4021 <para>The mangler is abstracted to a certain extent over some
4022 architecture-specific things such as the particular assembler
4023 directives used to herald symbols. Take a look at the
4024 definitions for other architectures and use these as a
4025 starting point.</para>
4029 <title>The native code generator</title>
4031 <para>The native code generator isn't essential to getting a
4032 registerised build going, but it's a desirable thing to have
4033 because it can cut compilation times in half. The native code
4034 generator is described in some detail in the <ulink
4035 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
4036 commentary</ulink>.</para>
4042 <para>To support GHCi, you need to port the dynamic linker
4043 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
4044 currently supports the ELF and PEi386 object file formats - if
4045 your platform uses one of these then things will be
4046 significantly easier. The majority of Unix platforms use the
4047 ELF format these days. Even so, there are some
4048 machine-specific parts of the ELF linker: for example, the
4049 code for resolving particular relocation types is
4050 machine-specific, so some porting of this code to your
4051 architecture will probaly be necessary.</para>
4053 <para>If your system uses a different object file format, then
4054 you have to write a linker — good luck!</para>
4060 <sect1 id="sec-build-pitfalls">
4061 <title>Known pitfalls in building Glasgow Haskell
4063 <indexterm><primary>problems, building</primary></indexterm>
4064 <indexterm><primary>pitfalls, in building</primary></indexterm>
4065 <indexterm><primary>building pitfalls</primary></indexterm></title>
4068 WARNINGS about pitfalls and known “problems”:
4077 One difficulty that comes up from time to time is running out of space
4078 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
4079 compensate for the vagaries of different sysadmin approaches to temp
4081 <indexterm><primary>tmp, running out of space in</primary></indexterm>
4083 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
4084 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
4087 The best way around it is to say
4089 <programlisting>export TMPDIR=<dir></programlisting>
4091 in your <filename>build.mk</filename> file.
4092 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
4101 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
4102 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
4110 When compiling via C, you'll sometimes get “warning: assignment from
4111 incompatible pointer type” out of GCC. Harmless.
4118 Similarly, <command>ar</command>chiving warning messages like the following are not
4121 <screen>ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
4122 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
4131 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
4132 get an “Out of heap space” error message. These can vary with the
4133 vagaries of different systems, it seems. The solution is simple:
4140 If you're compiling with GHC 4.00 or later, then the
4141 <emphasis>maximum</emphasis> heap size must have been reached. This
4142 is somewhat unlikely, since the maximum is set to 64M by default.
4143 Anyway, you can raise it with the
4144 <option>-optCrts-M<size></option> flag (add this flag to
4145 <constant><module>_HC_OPTS</constant>
4146 <command>make</command> variable in the appropriate
4147 <filename>Makefile</filename>).
4154 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
4163 and try again: <command>gmake</command>. (see <xref linkend="sec-suffix"/> for information about
4164 <constant><module>_HC_OPTS</constant>.)
4166 Alternatively, just cut to the chase:
4168 <screen>% cd ghc/compiler
4169 % make EXTRA_HC_OPTS=-optCrts-M128M</screen>
4177 If you try to compile some Haskell, and you get errors from GCC about
4178 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4179 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4181 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4182 this bug also suggests that you have an old GCC.
4190 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4193 <screen>% cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4194 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4196 ? # or, on some machines: ar s $i
4200 We'd be interested to know if this is still necessary.
4208 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4209 a bit from one Unix to another. One particular gotcha is macro calls
4213 <programlisting>SLIT("Hello, world")</programlisting>
4216 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4217 arguments, so you get
4220 <screen>:731: macro `SLIT' used with too many (2) args</screen>
4223 Alas, <command>cpp</command> doesn't tell you the offending file!
4225 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4236 <sect1 id="platforms"><title>Platforms, scripts, and file names</title>
4238 GHC is designed both to be built, and to run, on both Unix and Windows. This flexibility
4239 gives rise to a good deal of brain-bending detail, which we have tried to collect in this chapter.
4242 <sect2 id="cygwin-and-mingw"><title>Windows platforms: Cygwin, MSYS, and MinGW</title>
4244 <para> The build system is built around Unix-y makefiles. Because it's not native,
4245 the Windows situation for building GHC is particularly confusing. This section
4246 tries to clarify, and to establish terminology.</para>
4248 <sect3 id="ghc-mingw"><title>MinGW</title>
4250 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4251 is a collection of header
4252 files and import libraries that allow one to use <command>gcc</command> and produce
4253 native Win32 programs that do not rely on any third-party DLLs. The
4254 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4255 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4259 <para> The down-side of MinGW is that the MinGW libraries do not support anything like the full
4264 <sect3 id="ghc-cygwin"><title>Cygwin and MSYS</title>
4266 <para>You can't use the MinGW to <emphasis>build</emphasis> GHC, because MinGW doesn't have a shell,
4267 or the standard Unix commands such as <command>mv</command>, <command>rm</command>,
4268 <command>ls</command>, nor build-system stuff such as <command>make</command> and <command>cvs</command>.
4269 For that, there are two choices: <ulink url="http://www.cygwin.com">Cygwin</ulink>
4270 and <ulink url="http://www.mingw.org/msys.shtml">MSYS</ulink>:
4274 Cygwin comes with compilation tools (<command>gcc</command>, <command>ld</command> and so on), which
4275 compile code that has access to all of Posix. The price is that the executables must be
4276 dynamically linked with the Cygwin DLL, so that <emphasis>you cannot run a Cywin-compiled program on a machine
4277 that doesn't have Cygwin</emphasis>. Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4278 does not change, but the implementation certainly does. Even the interfaces to functions
4279 it exports seem to change occasionally. </para>
4283 MSYS is a fork of the Cygwin tree, so they
4284 are fundamentally similar. However, MSYS is by design much smaller and simpler. Access to the file system goes
4285 through fewer layers, so MSYS is quite a bit faster too.
4288 <para>Furthermore, MSYS provides no compilation tools; it relies instead on the MinGW tools. These
4289 compile binaries that run with no DLL support, on any Win32 system.
4290 However, MSYS does come with all the make-system tools, such as <command>make</command>, <command>autoconf</command>,
4291 <command>cvs</command>, <command>ssh</command> etc. To get these, you have to download the
4292 MsysDTK (Developer Tool Kit) package, as well as the base MSYS package.
4294 <para>MSYS does have a DLL, but it's only used by MSYS commands (<command>sh</command>, <command>rm</command>,
4295 <command>ssh</command> and so on),
4296 not by programs compiled under MSYS.
4304 <sect3><title>Targeting MinGW</title>
4306 <para>We want GHC to compile programs that work on any Win32 system. Hence:
4309 GHC does invoke a C compiler, assembler, linker and so on, but we ensure that it only
4310 invokes the MinGW tools, not the Cygwin ones. That means that the programs GHC compiles
4311 will work on any system, but it also means that the programs GHC compiles do not have access
4312 to all of Posix. In particular, they cannot import the (Haskell) Posix
4313 library; they have to do
4314 their input output using standard Haskell I/O libraries, or native Win32 bindings.</para>
4315 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4319 To make the GHC distribution self-contained, the GHC distribution includes the MinGW <command>gcc</command>,
4320 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4323 So <emphasis>GHC targets MinGW</emphasis>, not Cygwin.
4324 It is in principle possible to build a version of GHC, <emphasis>GHC-cygwin</emphasis>,
4325 that targets Cygwin instead. The up-side of GHC-cygwin is
4326 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4327 <emphasis>We do not support GHC-cygwin, however; it is beyond our resources.</emphasis>
4330 <para>While GHC <emphasis>targets</emphasis> MinGW, that says nothing about
4331 how GHC is <emphasis>built</emphasis>. We use both MSYS and Cygwin as build environments for
4332 GHC; both work fine, though MSYS is rather lighter weight.</para>
4334 <para>In your build tree, you build a compiler called <command>ghc-inplace</command>. It
4335 uses the <command>gcc</command> that you specify using the
4336 <option>--with-gcc</option> flag when you run
4337 <command>configure</command> (see below).
4338 The makefiles are careful to use <command>ghc-inplace</command> (not <command>gcc</command>)
4339 to compile any C files, so that it will in turn invoke the correct <command>gcc</command> rather that
4340 whatever one happens to be in your path. However, the makefiles do use whatever <command>ld</command>
4341 and <command>ar</command> happen to be in your path. This is a bit naughty, but (a) they are only
4342 used to glom together .o files into a bigger .o file, or a .a file,
4343 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4344 Cygwin and MinGW use the same .o file format. So its ok.
4348 <sect3><title> File names </title>
4350 <para>Cygwin, MSYS, and the underlying Windows file system all understand file paths of form <literal>c:/tmp/foo</literal>.
4354 MSYS programs understand <filename>/bin</filename>, <filename>/usr/bin</filename>, and map Windows's lettered drives as
4355 <filename>/c/tmp/foo</filename> etc. The exact mount table is given in the doc subdirectory of the MSYS distribution.
4357 <para> When it invokes a command, the MSYS shell sees whether the invoked binary lives in the MSYS <filename>/bin</filename>
4358 directory. If so, it just invokes it. If not, it assumes the program is no an MSYS program, and walks over the command-line
4359 arguments changing MSYS paths into native-compatible paths. It does this inside sub-arguments and inside quotes. For example,
4361 <programlisting>foogle -B/c/tmp/baz</programlisting>
4362 the MSYS shell will actually call <literal>foogle</literal> with argument <literal>-Bc:/tmp/baz</literal>.
4366 Cygwin programs have a more complicated mount table, and map the lettered drives as <filename>/cygdrive/c/tmp/foo</filename>.
4368 <para>The Cygwin shell does no argument processing when invoking non-Cygwin programs.
4374 <sect3><title>Host System vs Target System</title>
4377 In the source code you'll find various ifdefs looking like:
4378 <programlisting>#ifdef mingw32_HOST_OS
4380 #endif</programlisting>
4382 <programlisting>#ifdef mingw32_TARGET_OS
4384 #endif</programlisting>
4385 These macros are set by the configure script (via the file config.h).
4386 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4389 <para>The "host" system is the one on which GHC itself will be run.</para>
4392 <para>The "target" system is the one for which the program compiled by GHC will be run.</para>
4395 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4396 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4403 <sect2><title>Wrapper scripts</title>
4406 Many programs, including GHC itself and hsc2hs, need to find associated binaries and libraries.
4407 For <emphasis>installed</emphasis> programs, the strategy depends on the platform. We'll use
4408 GHC itself as an example:
4411 On Unix, the command <command>ghc</command> is a shell script, generated by adding installation
4412 paths to the front of the source file <filename>ghc.sh</filename>,
4413 that invokes the real binary, passing "-B<emphasis>path</emphasis>" as an argument to tell <command>ghc</command>
4414 where to find its supporting files.
4418 On vanilla Windows, it turns out to be much harder to make reliable script to be run by the
4419 native Windows shell <command>cmd</command> (e.g. limits on the length
4420 of the command line). So instead we invoke the GHC binary directly, with no -B flag.
4421 GHC uses the Windows <literal>getExecDir</literal> function to find where the executable is,
4422 and from that figures out where the supporting files are.
4425 (You can find the layout of GHC's supporting files in the
4426 section "Layout of installed files" of Section 2 of the GHC user guide.)
4429 Things work differently for <emphasis>in-place</emphasis> execution, where you want to
4430 execute a program that has just been built in a build tree. The difference is that the
4431 layout of the supporting files is different.
4432 In this case, whether on Windows or Unix, we always use a shell script. This works OK
4433 on Windows because the script is executed by MSYS or Cygwin, which don't have the
4434 shortcomings of the native Windows <command>cmd</command> shell.
4441 <sect1 id="winbuild"><title>Instructions for building under Windows</title>
4444 This section gives detailed instructions for how to build
4445 GHC from source on your Windows machine. Similar instructions for
4446 installing and running GHC may be found in the user guide. In general,
4447 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4450 Make sure you read the preceding section on platforms (<xref linkend="platforms"/>)
4451 before reading section.
4452 You don't need Cygwin or MSYS to <emphasis>use</emphasis> GHC,
4453 but you do need one or the other to <emphasis>build</emphasis> GHC.</para>
4456 <sect2 id="msys-install"><title>Installing and configuring MSYS</title>
4459 MSYS is a lightweight alternative to Cygwin.
4460 You don't need MSYS to <emphasis>use</emphasis> GHC,
4461 but you do need it or Cygwin to <emphasis>build</emphasis> GHC.
4462 Here's how to install MSYS.
4465 Go to <ulink url="http://www.mingw.org/download.shtml">http://www.mingw.org/download.shtml</ulink> and
4466 download the following (of course, the version numbers will differ):
4468 <listitem><para>The main MSYS package (binary is sufficient): <literal>MSYS-1.0.9.exe</literal>
4470 <listitem><para>The MSYS developer's toolkit (binary is sufficient): <literal>msysDTK-1.0.1.exe</literal>.
4471 This provides <command>make</command>, <command>autoconf</command>,
4472 <command>ssh</command>, <command>cvs</command> and probably more besides.
4475 Run both executables (in the order given above) to install them. I put them in <literal>c:/msys</literal>
4479 Set the following environment variables
4481 <listitem><para><literal>PATH</literal>: add <literal>c:/msys/1.0/bin</literal> to your path. (Of course, the version number may differ.)
4484 <listitem><para><literal>HOME</literal>: set to your home directory (e.g. <literal>c:/userid</literal>).
4485 This is where, among other things, <command>ssh</command> will look for your <literal>.ssh</literal> directory.
4488 <listitem><para><literal>SHELL</literal>: set to <literal>c:/msys/1.0/bin/sh.exe</literal>
4491 <listitem><para><literal>CVS_RSH</literal>: set to <literal>c:/msys/1.0/bin/ssh.exe</literal>. Only necessary if
4495 <listitem><para><literal>MAKE_MODE</literal>: set to <literal>UNIX</literal>. (I'm not certain this is necessary for MSYS.)
4502 Check that the <literal>CYGWIN</literal> environment variable is <emphasis>not</emphasis> set. It's a bad bug
4503 that MSYS is affected by this, but if you have CYGWIN set to "ntsec ntea", which is right for Cygwin, it
4504 causes the MSYS <command>ssh</command> to bogusly fail complaining that your <filename>.ssh/identity</filename>
4505 file has too-liberal permissinos.
4510 <para>Here are some points to bear in mind when using MSYS:
4512 <listitem> <para> MSYS does some kind of special magic to binaries stored in
4513 <filename>/bin</filename> and <filename>/usr/bin</filename>, which are by default both mapped
4514 to <filename>c:/msys/1.0/bin</filename> (assuming you installed MSYS in <filename>c:/msys</filename>).
4515 Do not put any other binaries (such as GHC or Alex) in this directory or its sub-directories:
4516 they fail in mysterious ways. However, it's fine to put other binaries in <filename>/usr/local/bin</filename>,
4517 which maps to <filename>c:/msys/1.0/local/bin</filename>.</para></listitem>
4519 <listitem> <para> MSYS seems to implement symbolic links by copying, so sharing is lost.
4523 Win32 has a <command>find</command> command which is not the same as MSYS's find.
4524 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4525 before the MSYS one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4526 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4527 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4531 MSYS comes with <command>bzip</command>, and MSYS's <command>tar</command>'s <literal>-j</literal>
4532 will bunzip an archive (e.g. <literal>tar xvjf foo.tar.bz2</literal>). Useful when you get a
4533 bzip'd dump.</para></listitem>
4539 <sect2><title>Installing and configuring Cygwin</title>
4541 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4542 The installation process is straightforward; we install it in <filename>c:/cygwin</filename>.
4543 During the installation dialogue, make sure that you select all of the following:
4544 <command>cvs</command>,
4545 <command>openssh</command>,
4546 <command>autoconf</command>,
4547 <command>binutils</command> (includes ld and (I think) ar),
4548 <command>gcc</command>,
4549 <command>flex</command>,
4550 <command>make</command>.
4551 If you miss out any of these, strange things will happen to you. To see thse packages,
4552 click on the "View" button in the "Select Packages"
4553 stage of Cygwin's installation dialogue, until the view says "Full". The default view, which is
4554 "Category" isn't very helpful, and the "View" button is rather unobtrousive.
4556 <para> Now set the following user environment variables:
4559 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4560 <constant>PATH</constant></para></listitem>
4564 Set <constant>MAKE_MODE</constant> to <literal>UNIX</literal>. If you
4565 don't do this you get very weird messages when you type
4566 <command>make</command>, such as:
4567 <screen>/c: /c: No such file or directory</screen>
4571 <listitem><para> Set <constant>SHELL</constant> to
4572 <filename>c:/cygwin/bin/bash</filename>. When you invoke a shell in Emacs, this
4573 <constant>SHELL</constant> is what you get.
4576 <listitem><para> Set <constant>HOME</constant> to point to your
4577 home directory. This is where, for example,
4578 <command>bash</command> will look for your <filename>.bashrc</filename>
4579 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4585 There are a few other things to do:
4589 By default, cygwin provides the command shell <filename>ash</filename>
4590 as <filename>sh.exe</filename>. We have often seen build-system problems that
4591 turn out to be due to bugs in <filename>ash</filename>
4593 and length of command lines). On the other hand <filename>bash</filename> seems
4595 So, in <filename>cygwin/bin</filename>
4596 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4597 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4598 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4599 you can't rename a running program!
4605 Some script files used in the make system start with "<command>#!/bin/perl</command>",
4606 (and similarly for <command>sh</command>). Notice the hardwired path!
4607 So you need to ensure that your <filename>/bin</filename> directory has the following
4610 <listitem> <para><command>sh</command></para></listitem>
4611 <listitem> <para><command>perl</command></para></listitem>
4612 <listitem> <para><command>cat</command></para></listitem>
4614 All these come in Cygwin's <filename>bin</filename> directory, which you probably have
4615 installed as <filename>c:/cygwin/bin</filename>. By default Cygwin mounts "<filename>/</filename>" as
4616 <filename>c:/cygwin</filename>, so if you just take the defaults it'll all work ok.
4617 (You can discover where your Cygwin
4618 root directory <filename>/</filename> is by typing <command>mount</command>.)
4619 Provided <filename>/bin</filename> points to the Cygwin <filename>bin</filename>
4620 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4621 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4627 <para>Finally, here are some things to be aware of when using Cygwin:
4629 <listitem> <para>Cygwin doesn't deal well with filenames that include
4630 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4634 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4635 magical text in it. So other programs that don't use Cygwin's
4636 I/O libraries won't recognise such files as symlinks.
4637 In particular, programs compiled by GHC are meant to be runnable
4638 without having Cygwin, so they don't use the Cygwin library, so
4639 they don't recognise symlinks.
4643 See the notes in <xref linkend="msys-install"/> about <command>find</command> and <command>bzip</command>,
4644 which apply to Cygwin too.
4652 <sect2 id="configure-ssh"><title>Configuring SSH</title>
4654 <para><command>ssh</command> comes with Cygwin, provided you remember to ask for it when
4655 you install Cygwin. (If not, the installer lets you update easily.) Look for <command>openssh</command>
4656 (not ssh) in the Cygwin list of applications!</para>
4658 <para>There are several strange things about <command>ssh</command> on Windows that you need to know.
4662 The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
4663 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
4664 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
4665 and run it as follows:
4666 <screen>c:\tmp> set CYGWIN32=tty
4667 c:\tmp> c:/user/local/bin/ssh-keygen1</screen> </para>
4670 <listitem><para> (Cygwin-only problem, I think.)
4671 <command>ssh</command> needs to access your directory <filename>.ssh</filename>, in your home directory.
4672 To determine your home directory <command>ssh</command> first looks in
4673 <filename>c:/cygwin/etc/passwd</filename> (or wherever you have Cygwin installed). If there's an entry
4674 there with your userid, it'll use that entry to determine your home directory, <emphasis>ignoring
4675 the setting of the environment variable $HOME</emphasis>. If the home directory is
4676 bogus, <command>ssh</command> fails horribly. The best way to see what is going on is to say
4677 <programlisting>ssh -v cvs.haskell.org</programlisting>
4678 which makes <command>ssh</command> print out information about its activity.
4680 <para> You can fix this problem, either by correcting the home-directory field in
4681 <filename>c:/cygwin/etc/passwd</filename>, or by simply deleting the entire entry for your userid. If
4682 you do that, <command>ssh</command> uses the $HOME environment variable instead.
4688 <para>To protect your
4689 <literal>.ssh</literal> from access by anyone else,
4690 right-click your <literal>.ssh</literal> directory, and
4691 select <literal>Properties</literal>. If you are not on
4692 the access control list, add yourself, and give yourself
4693 full permissions (the second panel). Remove everyone else
4694 from the access control list. Don't leave them there but
4695 deny them access, because 'they' may be a list that
4696 includes you!</para>
4700 <para>In fact <command>ssh</command> 3.6.1 now seems to <emphasis>require</emphasis>
4701 you to have Unix permissions 600 (read/write for owner only)
4702 on the <literal>.ssh/identity</literal> file, else it
4703 bombs out. For your local C drive, it seems that <literal>chmod 600 identity</literal> works,
4704 but on Windows NT/XP, it doesn't work on a network drive (exact dteails obscure).
4705 The solution seems to be to set the $CYGWIN environment
4706 variable to "<literal>ntsec neta</literal>". The $CYGWIN environment variable is discussed
4707 in <ulink url="http://cygwin.com/cygwin-ug-net/using-cygwinenv.html">the Cygwin User's Guide</ulink>,
4708 and there are more details in <ulink url="http://cygwin.com/faq/faq_4.html#SEC44">the Cygwin FAQ</ulink>.
4715 <sect2><title>Other things you need to install</title>
4717 <para>You have to install the following other things to build GHC, listed below.</para>
4719 <para>On Windows you often install executables in directories with spaces, such as
4720 "<filename>Program Files</filename>". However, the <literal>make</literal> system for fptools doesn't
4721 deal with this situation (it'd have to do more quoting of binaries), so you are strongly advised
4722 to put binaries for all tools in places with no spaces in their path.
4723 On both MSYS and Cygwin, it's perfectly OK to install such programs in the standard Unixy places,
4724 <filename>/usr/local/bin</filename> and <filename>/usr/local/lib</filename>. But it doesn't matter,
4725 provided they are in your path.
4729 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4730 This is what you will use to compile GHC. Add it in your
4731 <constant>PATH</constant>: the installer tells you the path element
4732 you need to add upon completion.
4738 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4739 Happy is a parser generator used to compile the Haskell grammar. Under MSYS or Cygwin you can easily
4740 build it from the source distribution using
4741 <programlisting>./configure
4743 make install</programlisting>
4744 This should install it in <filename>/usr/local/bin</filename> (which maps to <filename>c:/msys/1.0/local/bin</filename>
4746 Make sure the installation directory is in your
4747 <constant>PATH</constant>.
4752 <para>Install Alex. This can be done by building from the
4753 source distribution in the same way as Happy. Sources are
4754 available from <ulink
4755 url="http://www.haskell.org/alex">http://www.haskell.org/alex</ulink>.</para>
4759 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4760 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw"/>).
4761 Just pick up a mingw bundle at
4762 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4763 We install it in <filename>c:/mingw</filename>.
4765 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4766 They are only going to get used by explicit access (via the --with-gcc flag you
4767 give to <command>configure</command> later). If you do add them to your path
4768 you are likely to get into a mess because their names overlap with Cygwin binaries.
4774 <para>We use <command>emacs</command> a lot, so we install that too.
4775 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4776 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4777 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4778 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4779 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4780 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4786 <para> Finally, check out a copy of GHC sources from
4787 the CVS repository, following the instructions above (<xref linkend="cvs-access"/>).
4794 <sect2><title>Building GHC</title>
4797 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source"/>);
4798 the bullets below only tell
4799 you about Windows-specific wrinkles.</para>
4803 If you used <command>autoconf</command> instead of <command>autoreconf</command>,
4804 you'll get an error when you run <filename>./configure</filename>:
4807 creating mk/config.h
4808 mk/config.h is unchanged
4810 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4811 ./configure: ./configure: No such file or directory
4812 configure: error: ./configure failed for ghc</screen>
4816 <listitem> <para><command>autoreconf</command> seems to create the file <filename>configure</filename>
4817 read-only. So if you need to run autoreconf again (which I sometimes do for safety's sake),
4819 <screen>/usr/bin/autoconf: cannot create configure: permission denied</screen>
4820 Solution: delete <filename>configure</filename> first.
4825 After <command>autoreconf</command> run <command>./configure</command> in
4826 <filename>fptools/</filename> thus:
4828 <screen>./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc</screen>
4829 This is the point at which you specify that you are building GHC-mingw
4830 (see <xref linkend="ghc-mingw"/>). </para>
4832 <para> Both these options are important! It's possible to get into
4833 trouble using the wrong C compiler!</para>
4835 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4836 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4837 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4838 understand a Cygwin path. For example, if you
4839 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4840 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4841 time it tries to invoke it. Worse, the failure comes with
4842 no error message whatsoever. GHC simply fails silently when first invoked,
4843 typically leaving you with this:
4844 <screen>make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4845 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4846 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4847 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4848 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4849 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4850 -optc-fomit-frame-pointer -O2 -static
4851 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4852 make[2]: *** [Adjustor.o] Error 1
4853 make[1]: *** [all] Error 1
4854 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4855 make: *** [all] Error 1</screen>
4860 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin"/>)
4861 you'll have to do something more like:
4862 <screen>./configure --with-gcc=...the Cygwin gcc...</screen>
4867 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4868 This file occasionally remembers out-of-date configuration information, which
4869 can be really confusing.
4873 <listitem><para> You almost certainly want to set
4874 <programlisting>SplitObjs = NO</programlisting>
4875 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config"/>).
4876 This tells the build system not to split each library into a myriad of little object files, one
4877 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4878 it dramatically increases the time taken to build the libraries in the first place.
4882 <listitem><para> Do not attempt to build the documentation.
4883 It needs all kinds of wierd Jade stuff that we haven't worked out for
4884 Win32.</para></listitem>