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.</term>
68 <indexterm><primary>CVS repository</primary>
71 <para>We make releases infrequently. If you want more
72 up-to-the minute (but less tested) source code then you need
73 to get access to our CVS repository.</para>
75 <para>All the <literal>fptools</literal> source code is held
76 in a CVS repository. CVS is a pretty good source-code
77 control system, and best of all it works over the
80 <para>The repository holds source code only. It holds no
81 mechanically generated files at all. So if you check out a
82 source tree from CVS you will need to install every utility
83 so that you can build all the derived files from
86 <para>More information about our CVS repository can be found
87 in <xref linkend="sec-cvs">.</para>
92 <para>If you are going to do any building from sources (either
93 from a source distribution or the CVS repository) then you need to
94 read all of this manual in detail.</para>
98 <title>Using the CVS repository</title>
100 <para>We use <ulink url="http://www.cvshome.org/">CVS</ulink> (Concurrent Version System) to keep track of our
101 sources for various software projects. CVS lets several people
102 work on the same software at the same time, allowing changes to be
103 checked in incrementally. </para>
105 <para>This section is a set of guidelines for how to use our CVS
106 repository, and will probably evolve in time. The main thing to
107 remember is that most mistakes can be undone, but if there's
108 anything you're not sure about feel free to bug the local CVS
109 meister (namely Jeff Lewis
110 <email>jlewis@galois.com</email>). </para>
112 <sect2 id="cvs-access">
113 <title>Getting access to the CVS Repository</title>
115 <para>You can access the repository in one of two ways:
116 read-only (<xref linkend="cvs-read-only">), or read-write (<xref
117 linkend="cvs-read-write">).</para>
119 <sect3 id="cvs-read-only">
120 <title>Remote Read-only CVS Access</title>
122 <para>Read-only access is available to anyone - there's no
123 need to ask us first. With read-only CVS access you can do
124 anything except commit changes to the repository. You can
125 make changes to your local tree, and still use CVS's merge
126 facility to keep your tree up to date, and you can generate
127 patches using 'cvs diff' in order to send to us for
130 <para>To get read-only access to the repository:</para>
134 <para>Make sure that <application>cvs</application> is
135 installed on your machine.</para>
138 <para>Set your <literal>$CVSROOT</literal> environment variable to
139 <literal>:pserver:anoncvs@glass.cse.ogi.edu:/cvs</literal></para>
140 <para>If you set <literal>$CVSROOT</literal> in a shell script, be sure not to
141 have any trailing spaces on that line, otherwise CVS will respond with
142 a perplexing message like
144 /cvs : no such repository
145 </programlisting></para>
148 <para>Run the command</para>
152 <para>The password is simply <literal>cvs</literal>. This
153 sets up a file in your home directory called
154 <literal>.cvspass</literal>, which squirrels away the
155 dummy password, so you only need to do this step once.</para>
159 <para>Now go to <xref linkend="cvs-first">.</para>
164 <sect3 id="cvs-read-write">
165 <title>Remote Read-Write CVS Access</title>
167 <para>We generally supply read-write access to folk doing
168 serious development on some part of the source tree, when
169 going through us would be a pain. If you're developing some
170 feature, or think you have the time and inclination to fix
171 bugs in our sources, feel free to ask for read-write
172 access. There is a certain amount of responsibility that goes
173 with commit privileges; we are more likely to grant you access
174 if you've demonstrated your competence by sending us patches
175 via mail in the past.</para>
177 <para>To get remote read-write CVS access, you need to do the
178 following steps.</para>
182 <para>Make sure that <literal>cvs</literal> and
183 <literal>ssh</literal> are both installed on your
188 <para>Generate a DSA private-key/public-key pair, thus:</para>
192 <para>(<literal>ssh-keygen</literal> comes with
193 <literal>ssh</literal>.) Running <literal>ssh-keygen
194 -d</literal> creates the private and public keys in
195 <literal>$HOME/.ssh/id_dsa</literal> and
196 <literal>$HOME/.ssh/id_dsa.pub</literal> respectively
197 (assuming you accept the standard defaults).</para>
199 <para><literal>ssh-keygen -d</literal> will only work if
200 you have Version 2 <literal>ssh</literal> installed; it
201 will fail harmlessly otherwise. If you only have Version
202 1 you can instead generate an RSA key pair using plain</para>
207 <para>Doing so creates the private and public RSA keys in
208 <literal>$HOME/.ssh/identity</literal> and
209 <literal>$HOME/.ssh/identity.pub</literal>
212 <para>[Deprecated.] Incidentally, you can force a Version
213 2 <literal>ssh</literal> to use the Version 1 protocol by
214 creating <literal>$HOME/config</literal> with the
215 following in it:</para>
223 <para>In both cases, <literal>ssh-keygen</literal> will
224 ask for a <firstterm>passphrase</firstterm>. The
225 passphrase is a password that protects your private key.
226 In response to the 'Enter passphrase' question, you can
230 <para>[Recommended.] Enter a passphrase, which you
231 will quote each time you use CVS.
232 <literal>ssh-agent</literal> makes this entirely
236 <para>[Deprecated.] Just hit return (i.e. use an empty
237 passphrase); then you won't need to quote the
238 passphrase when using CVS. The downside is that
239 anyone who can see into your <literal>.ssh</literal>
240 directory, and thereby get your private key, can mess
241 up the repository. So you must keep the
242 <literal>.ssh</literal> directory with draconian
243 no-access permissions.</para>
249 <emphasis>Windows users: see the notes in <xref linkend="configure-ssh"> about <command>ssh</command> wrinkles!</emphasis>
256 <para>Send a message to to the CVS repository
257 administrator (currently Jeff Lewis
258 <email>jeff@galois.com</email>), containing:</para>
261 <para>Your desired user-name.</para>
264 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
265 <literal>.ssh/identity.pub</literal>).</para>
268 <para>He will set up your account.</para>
272 <para>Set the following environment variables:</para>
276 <constant>$HOME</constant>: points to your home directory. This is where CVS
277 will look for its <filename>.cvsrc</filename> file.
283 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
285 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
286 <literal>ssh</literal> assumes that your CVS client
287 understands how to execute shell script
288 ("#!"s,really), which is what
289 <literal>ssh</literal> is. This may not be the case on
290 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
291 <literal>ssh1</literal>.</para>
295 <para><literal>$CVSROOT</literal> to
296 <literal>:ext:</literal><replaceable>your-username</replaceable>
297 <literal>@cvs.haskell.org:/home/cvs/root</literal>
298 where <replaceable>your-username</replaceable> is your user name on
299 <literal>cvs.haskell.org</literal>.
301 <para>The <literal>CVSROOT</literal> environment variable will
302 be recorded in the checked-out tree, so you don't need to set
303 this every time. </para>
309 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
310 if you want to use an Emacs buffer for typing in those long commit messages.
316 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
317 set this to point to <filename>bash.exe</filename>.
328 Put the following in <filename>$HOME/.cvsrc</filename>:
339 These are the default options for the specified CVS commands,
340 and represent better defaults than the usual ones. (Feel
341 free to change them.)
345 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
346 the 8.3 DOS filesystem, but that restriction should have
347 been lifted by now (i.e., you're using VFAT or later filesystems.) If
348 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
356 <para>[Experts.] Once your account is set up, you can get
357 access from other machines without bothering Jeff, thus:</para>
360 <para>Generate a public/private key pair on the new
364 <para>Use ssh to log in to
365 <literal>cvs.haskell.org</literal>, from your old
369 <para>Add the public key for the new machine to the file
370 <literal>$HOME/ssh/authorized_keys</literal> on
371 <literal>cvs.haskell.org</literal>.
372 (<literal>authorized_keys2</literal>, I think, for Version
376 <para>Make sure that the new version of
377 <literal>authorized_keys</literal> still has 600 file
386 <sect2 id="cvs-first">
387 <title>Checking Out a Source Tree</title>
391 <para>Make sure you set your <literal>CVSROOT</literal>
392 environment variable according to either of the remote
393 methods above. The Approved Way to check out a source tree
394 is as follows:</para>
397 $ cvs checkout fpconfig
400 <para>At this point you have a new directory called
401 <literal>fptools</literal> which contains the basic stuff
402 for the fptools suite, including the configuration files and
403 some other junk. </para>
405 <para>[Windows users.] The following messages appear to be harmless:
407 setsockopt IPTOS_LOWDELAY: Invalid argument
408 setsockopt IPTOS_THROUGHPUT: Invalid argument
413 <para>You can call the fptools directory whatever you like,
414 CVS won't mind: </para>
417 $ mv fptools <replaceable>directory</replaceable>
420 <para> NB: after you've read the CVS manual you might be
421 tempted to try</para>
423 $ cvs checkout -d <replaceable>directory</replaceable> fpconfig
426 <para>instead of checking out <literal>fpconfig</literal>
427 and then renaming it. But this doesn't work, and will
428 result in checking out the entire repository instead of just
429 the <literal>fpconfig</literal> bit.</para>
431 $ cd <replaceable>directory</replaceable>
432 $ cvs checkout ghc hslibs libraries
435 <para>The second command here checks out the relevant
436 modules you want to work on. For a GHC build, for instance,
437 you need at least the <literal>ghc</literal>,
438 <literal>hslibs</literal> and <literal>libraries</literal>
439 modules (for a full list of the projects available, see
440 <xref linkend="projects">).</para>
442 <para>Remember that if you do not have
443 <literal>happy</literal> and/or <literal>Alex</literal>
444 installed, you need to check them out as well.</para>
449 <sect2 id="cvs-committing">
450 <title>Committing Changes</title>
452 <para>This is only if you have read-write access to the
453 repository. For anoncvs users, CVS will issue a "read-only
454 repository" error if you try to commit changes.</para>
458 <para>Build the software, if necessary. Unless you're just
459 working on documentation, you'll probably want to build the
460 software in order to test any changes you make.</para>
464 <para>Make changes. Preferably small ones first.</para>
468 <para>Test them. You can see exactly what changes you've
469 made by using the <literal>cvs diff</literal> command:</para>
473 <para>lists all the changes (using the
474 <literal>diff</literal> command) in and below the current
475 directory. In emacs, <literal>C-c C-v =</literal> runs
476 <literal>cvs diff</literal> on the current buffer and shows
477 you the results.</para>
481 <para>If you changed something in the
482 <literal>fptools/libraries</literal> subdirectories, also run
483 <literal>make html</literal> to check if the documentation can
484 be generated successfully, too.</para>
488 <para>Before checking in a change, you need to update your
495 <para>This pulls in any changes that other people have made,
496 and merges them with yours. If there are any conflicts, CVS
497 will tell you, and you'll have to resolve them before you
498 can check your changes in. The documentation describes what
499 to do in the event of a conflict.</para>
501 <para>It's not always necessary to do a full cvs update
502 before checking in a change, since CVS will always tell you
503 if you try to check in a file that someone else has changed.
504 However, you should still update at regular intervals to
505 avoid making changes that don't work in conjuction with
506 changes that someone else made. Keeping an eye on what goes
507 by on the mailing list can help here.</para>
511 <para>When you're happy that your change isn't going to
512 break anything, check it in. For a one-file change:</para>
515 $ cvs commit <replaceable>filename</replaceable>
518 <para>CVS will then pop up an editor for you to enter a
519 "commit message", this is just a short description
520 of what your change does, and will be kept in the history of
523 <para>If you're using emacs, simply load up the file into a
524 buffer and type <literal>C-x C-q</literal>, and emacs will
525 prompt for a commit message and then check in the file for
528 <para>For a multiple-file change, things are a bit
529 trickier. There are several ways to do this, but this is the
530 way I find easiest. First type the commit message into a
531 temporary file. Then either</para>
534 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
537 <para>or, if nothing else has changed in this part of the
541 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
544 <para>where <replaceable>directory</replaceable> is a common
545 parent directory for all your changes, and
546 <replaceable>commit-message</replaceable> is the name of the
547 file containing the commit message.</para>
549 <para>Shortly afterwards, you'll get some mail from the
550 relevant mailing list saying which files changed, and giving
551 the commit message. For a multiple-file change, you should
552 still get only <emphasis>one</emphasis> message.</para>
557 <sect2 id="cvs-update">
558 <title>Updating Your Source Tree</title>
560 <para>It can be tempting to cvs update just part of a source
561 tree to bring in some changes that someone else has made, or
562 before committing your own changes. This is NOT RECOMMENDED!
563 Quite often changes in one part of the tree are dependent on
564 changes in another part of the tree (the
565 <literal>mk/*.mk</literal> files are a good example where
566 problems crop up quite often). Having an inconsistent tree is a
567 major cause of headaches. </para>
569 <para>So, to avoid a lot of hassle, follow this recipe for
570 updating your tree:</para>
574 $ cvs update -P 2>&1 | tee log</screen>
576 <para>Look at the log file, and fix any conflicts (denoted by a
577 <quote>C</quote> in the first column). New directories may have
578 appeared in the repository; CVS doesn't check these out by
579 default, so to get new directories you have to explicitly do
581 $ cvs update -d</screen>
582 in each project subdirectory. Don't do this at the top level,
583 because then <emphasis>all</emphasis> the projects will be
586 <para>If you're using multiple build trees, then for every build
587 tree you have pointing at this source tree, you need to update
588 the links in case any new files have appeared: </para>
591 $ cd <replaceable>build-tree</replaceable>
592 $ lndir <replaceable>source-tree</replaceable>
595 <para>Some files might have been removed, so you need to remove
596 the links pointing to these non-existent files:</para>
599 $ find . -xtype l -exec rm '{}' \;
602 <para>To be <emphasis>really</emphasis> safe, you should do
605 <screen>$ gmake all</screen>
607 <para>from the top-level, to update the dependencies and build
608 any changed files. </para>
611 <sect2 id="cvs-tags">
612 <title>GHC Tag Policy</title>
614 <para>If you want to check out a particular version of GHC,
615 you'll need to know how we tag versions in the repository. The
616 policy (as of 4.04) is:</para>
620 <para>The tree is branched before every major release. The
621 branch tag is <literal>ghc-x-xx-branch</literal>, where
622 <literal>x-xx</literal> is the version number of the release
623 with the <literal>'.'</literal> replaced by a
624 <literal>'-'</literal>. For example, the 4.04 release lives
625 on <literal>ghc-4-04-branch</literal>.</para>
629 <para>The release itself is tagged with
630 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
631 called <literal>ghc-4-06</literal>.</para>
635 <para>We didn't always follow these guidelines, so to see
636 what tags there are for previous versions, do <literal>cvs
637 log</literal> on a file that's been around for a while (like
638 <literal>fptools/ghc/README</literal>).</para>
642 <para>So, to check out a fresh GHC 4.06 tree you would
646 $ cvs co -r ghc-4-06 fpconfig
648 $ cvs co -r ghc-4-06 ghc hslibs
652 <sect2 id="cvs-hints">
653 <title>General Hints</title>
657 <para>As a general rule: commit changes in small units,
658 preferably addressing one issue or implementing a single
659 feature. Provide a descriptive log message so that the
660 repository records exactly which changes were required to
661 implement a given feature/fix a bug. I've found this
662 <emphasis>very</emphasis> useful in the past for finding out
663 when a particular bug was introduced: you can just wind back
664 the CVS tree until the bug disappears.</para>
668 <para>Keep the sources at least *buildable* at any given
669 time. No doubt bugs will creep in, but it's quite easy to
670 ensure that any change made at least leaves the tree in a
671 buildable state. We do nightly builds of GHC to keep an eye
672 on what things work/don't work each day and how we're doing
673 in relation to previous verions. This idea is truely wrecked
674 if the compiler won't build in the first place!</para>
678 <para>To check out extra bits into an already-checked-out
679 tree, use the following procedure. Suppose you have a
680 checked-out fptools tree containing just ghc, and you want
681 to add nofib to it:</para>
692 $ cvs update -d nofib
695 <para>(the -d flag tells update to create a new
696 directory). If you just want part of the nofib suite, you
701 $ cvs checkout nofib/spectral
704 <para>This works because <literal>nofib</literal> is a
705 module in its own right, and spectral is a subdirectory of
706 the nofib module. The path argument to checkout must always
707 start with a module name. There's no equivalent form of this
708 command using <literal>update</literal>.</para>
714 <sect1 id="projects">
715 <title>What projects are there?</title>
717 <para>The <literal>fptools</literal> suite consists of several
718 <firstterm>projects</firstterm>, most of which can be downloaded,
719 built and installed individually. Each project corresponds to a
720 subdirectory in the source tree, and if checking out from CVS then
721 each project can be checked out individually by sitting in the top
722 level of your source tree and typing <command>cvs checkout
723 <replaceable>project</replaceable></command>.</para>
725 <para>Here is a list of the projects currently available:</para>
729 <term><literal>alex</literal></term>
730 <indexterm><primary><literal>alex</literal></primary>
731 <secondary>project</secondary></indexterm>
734 url="http://www.haskell.org/alex/">Alex</ulink> lexical
735 analyser generator for Haskell.</para>
740 <term><literal>ghc</literal></term>
741 <indexterm><primary><literal>ghc</literal></primary>
742 <secondary>project</secondary></indexterm>
744 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
745 Haskell Compiler</ulink> (minus libraries). Absolutely
746 required for building GHC.</para>
751 <term><literal>glafp-utils</literal></term>
752 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
754 <para>Utility programs, some of which are used by the
755 build/installation system. Required for pretty much
761 <term><literal>greencard</literal></term>
762 <indexterm><primary><literal>greencard</literal></primary><secondary>project</secondary></indexterm>
765 url="http://www.haskell.org/greencard/">GreenCard</ulink>
766 system for generating Haskell foreign function
772 <term><literal>haggis</literal></term>
773 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
776 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
777 Haskell GUI framework.</para>
782 <term><literal>haddock</literal></term>
783 <indexterm><primary><literal>haddock</literal></primary><secondary>project</secondary></indexterm>
786 url="http://www.haskell.org/haddock/">Haddock</ulink>
787 documentation tool.</para>
792 <term><literal>happy</literal></term>
793 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
796 url="http://www.haskell.org/happy/">Happy</ulink> Parser
802 <term><literal>hdirect</literal></term>
803 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
806 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
807 Haskell interoperability tool.</para>
812 <term><literal>hood</literal></term>
813 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
815 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
816 Object Observation Debugger</ulink>.</para>
821 <term><literal>hslibs</literal></term>
822 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
824 <para>Supplemental libraries for GHC
825 (<emphasis>required</emphasis> for building GHC).</para>
830 <term><literal>libraries</literal></term>
831 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
833 <para>Hierarchical Haskell library suite
834 (<emphasis>required</emphasis> for building GHC).</para>
839 <term><literal>mhms</literal></term>
840 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
842 <para>The Modular Haskell Metric System.</para>
847 <term><literal>nofib</literal></term>
848 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
850 <para>The NoFib suite: A collection of Haskell programs used
851 primarily for benchmarking.</para>
856 <term><literal>testsuite</literal></term>
857 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
859 <para>A testing framework, including GHC's regression test
865 <para>So, to build GHC you need at least the
866 <literal>ghc</literal>, <literal>libraries</literal> and
867 <literal>hslibs</literal> projects (a GHC source distribution will
868 already include the bits you need).</para>
871 <sect1 id="sec-build-checks">
872 <title>Things to check before you start</title>
874 <para>Here's a list of things to check before you get
880 <indexterm><primary>Disk space needed</primary></indexterm>
881 <para>Disk space needed: from about 100Mb for a basic GHC
882 build, up to probably 500Mb for a GHC build with everything
883 included (libraries built several different ways,
888 <para>Use an appropriate machine / operating system. <xref
889 linkend="sec-port-info"> lists the supported platforms; if
890 yours isn't amongst these then you can try porting GHC (see
891 <xref linkend="sec-porting-ghc">).</para>
895 <para>Be sure that the “pre-supposed” utilities are
896 installed. <Xref LinkEnd="sec-pre-supposed">
901 <para>If you have any problem when building or installing the
902 Glasgow tools, please check the “known pitfalls” (<Xref
903 LinkEnd="sec-build-pitfalls">). Also check the FAQ for the
904 version you're building, which is part of the User's Guide and
905 available on the <ulink URL="http://www.haskell.org/ghc/" >GHC web
908 <indexterm><primary>bugs</primary><secondary>known</secondary></indexterm>
910 <para>If you feel there is still some shortcoming in our
911 procedure or instructions, please report it.</para>
913 <para>For GHC, please see the <ulink
914 url="http://www.haskell.org/ghc/docs/latest/set/bug-reporting.html">bug-reporting
915 section of the GHC Users' Guide</ulink>, to maximise the
916 usefulness of your report.</para>
918 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
919 <para>If in doubt, please send a message to
920 <email>glasgow-haskell-bugs@haskell.org</email>.
921 <indexterm><primary>bugs</primary><secondary>mailing
922 list</secondary></indexterm></para>
927 <sect1 id="sec-port-info">
928 <title>What machines the Glasgow tools run on</title>
930 <indexterm><primary>ports</primary><secondary>GHC</secondary></indexterm>
931 <indexterm><primary>GHC</primary><secondary>ports</secondary></indexterm>
932 <indexterm><primary>platforms</primary><secondary>supported</secondary></indexterm>
934 <para>The main question is whether or not the Haskell compiler
935 (GHC) runs on your platform.</para>
937 <para>A “platform” is a
938 architecture/manufacturer/operating-system combination, such as
939 <literal>sparc-sun-solaris2</literal>. Other common ones are
940 <literal>alpha-dec-osf2</literal>,
941 <literal>hppa1.1-hp-hpux9</literal>,
942 <literal>i386-unknown-linux</literal>,
943 <literal>i386-unknown-solaris2</literal>,
944 <literal>i386-unknown-freebsd</literal>,
945 <literal>i386-unknown-cygwin32</literal>,
946 <literal>m68k-sun-sunos4</literal>,
947 <literal>mips-sgi-irix5</literal>,
948 <literal>sparc-sun-sunos4</literal>,
949 <literal>sparc-sun-solaris2</literal>,
950 <literal>powerpc-ibm-aix</literal>.</para>
952 <para>Some libraries may only work on a limited number of
953 platforms; for example, a sockets library is of no use unless the
954 operating system supports the underlying BSDisms.</para>
957 <title>What platforms the Haskell compiler (GHC) runs on</title>
959 <indexterm><primary>fully-supported platforms</primary></indexterm>
960 <indexterm><primary>native-code generator</primary></indexterm>
961 <indexterm><primary>registerised ports</primary></indexterm>
962 <indexterm><primary>unregisterised ports</primary></indexterm>
964 <para>The GHC hierarchy of Porting Goodness: (a) Best is a
965 native-code generator; (b) next best is a
966 “registerised” port; (c) the bare minimum is an
967 “unregisterised” port.
968 (“Unregisterised” is so terrible that we won't say
969 more about it).</para>
971 <para>We use Sparcs running Solaris 2.7 and x86 boxes running
972 FreeBSD and Linux, so those are the best supported platforms,
973 unsurprisingly.</para>
975 <para>Here's everything that's known about GHC ports. We
976 identify platforms by their “canonical”
977 CPU/Manufacturer/OS triple.</para>
981 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:</term>
982 <indexterm><primary>alpha-dec-osf</primary></indexterm>
983 <indexterm><primary>alpha-dec-linux</primary></indexterm>
984 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
985 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
986 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
989 <para>The OSF port is currently working (as of GHC version
990 5.02.1) and well supported. The native code generator is
991 currently non-working. Other operating systems will
992 require some minor porting.</para>
997 <term>sparc-sun-sunos4</term>
998 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
1000 <para>Probably works with minor tweaks, hasn't been tested
1006 <term>sparc-sun-solaris2</term>
1007 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
1009 <para>Fully supported (at least for Solaris 2.7 and 2.6),
1010 including native-code generator.</para>
1015 <term>sparc-unknown-openbsd</term>
1016 <indexterm><primary>sparc-unknown-openbsd</primary></indexterm>
1018 <para>Supported, including native-code generator. The
1019 same should also be true of NetBSD</para>
1024 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
1025 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1027 <para>A registerised port is available for version 4.08,
1028 but GHC hasn't been built on that platform since (as far
1029 as we know). No native-code generator.</para>
1034 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1035 <indexterm><primary>i386-*-linux</primary></indexterm>
1037 <para>GHC works registerised and has a native code
1038 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1039 or later. NOTE about <literal>glibc</literal> versions:
1040 GHC binaries built on a system running <literal>glibc
1041 2.0</literal> won't work on a system running
1042 <literal>glibc 2.1</literal>, and vice versa. In general,
1043 don't expect compatibility between
1044 <literal>glibc</literal> versions, even if the shared
1045 library version hasn't changed.</para>
1050 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or
1052 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1054 <para>GHC works registerised. Pre-built packages are
1055 available in the native package format, so if you just
1056 need binaries you're better off just installing the
1057 package (it might even be on your installation
1063 <term>i386-unknown-openbsd (PCs running OpenBSD)</term>
1064 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1066 <para>Supported, with native code generator. Packages are
1067 available through the ports system in the native package
1073 <term>i386-unknown-netbsd (PCs running NetBSD)</term>
1074 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1076 <para>Will require some minor porting effort, but should
1077 work registerised.</para>
1082 <term>i386-unknown-mingw32 (PCs running Windows)</term>
1083 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1085 <para>Fully supported under Win9x, WinNT, Win2k, and
1086 WinXP. Includes a native code generator. Building from
1087 source requires a recent <ulink
1088 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1089 to be installed.</para>
1094 <term>ia64-unknown-linux</term>
1095 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1097 <para>Supported, except there is no native code
1103 <term>x86_64-unknown-linux</term>
1104 <indexterm><primary>x86_64-unknown-linux</primary></indexterm>
1106 <para>GHC currently works unregisterised. A registerised
1107 port is in progress.</para>
1112 <term>amd64-unknown-openbsd</term>
1113 <indexterm><primary>amd64-unknown-linux</primary></indexterm>
1115 <para>(This is the same as x86_64-unknown-openbsd). GHC
1116 currently works unregisterised. A registerised port is in
1122 <term>mips-sgi-irix5</term>
1123 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1125 <para>Port has worked in the past, but hasn't been tested
1126 for some time (and will certainly have rotted in various
1127 ways). As usual, we don't have access to machines and
1128 there hasn't been an overwhelming demand for this port,
1129 but feel free to get in touch.</para>
1134 <term>mips64-sgi-irix6</term>
1135 <indexterm><primary>mips-sgi-irix6</primary></indexterm>
1137 <para>GHC currently works unregisterised.</para>
1142 <term>powerpc-ibm-aix</term>
1143 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1145 <para>Port currently doesn't work, needs some minimal
1146 porting effort. As usual, we don't have access to
1147 machines and there hasn't been an overwhelming demand for
1148 this port, but feel free to get in touch.</para>
1153 <term>powerpc-apple-darwin</term>
1154 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1156 <para>Supported registerised. Native code generator is
1157 almost working.</para>
1162 <term>powerpc-apple-linux</term>
1163 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1165 <para>Not supported (yet).</para>
1170 <para>Various other systems have had GHC ported to them in the
1171 distant past, including various Motorola 68k boxes. The 68k
1172 support still remains, but porting to one of these systems will
1173 certainly be a non-trivial task.</para>
1177 <title>What machines the other tools run on</title>
1179 <para>Unless you hear otherwise, the other tools work if GHC
1185 <sect1 id="sec-pre-supposed">
1186 <title>Installing pre-supposed utilities</title>
1188 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1189 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1191 <para>Here are the gory details about some utility programs you
1192 may need; <command>perl</command>, <command>gcc</command> and
1193 <command>happy</command> are the only important
1194 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1195 important if you're going for Parallel Haskell.) The
1196 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1197 script will tell you if you are missing something.</para>
1203 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1204 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1206 <para>GHC is required to build many of the tools, including
1207 GHC itself. If you need to port GHC to your platform
1208 because there isn't a binary distribution of GHC available,
1209 then see <xref linkend="sec-porting-ghc">.</para>
1211 <para>Which version of GHC you need will depend on the
1212 packages you intend to build. GHC itself will normally
1213 build using one of several older versions of itself - check
1214 the announcement or release notes for details.</para>
1220 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1221 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1223 <para><emphasis>You have to have Perl to proceed!</emphasis>
1224 Perl version 5 at least is required. GHC has been known to
1225 tickle bugs in Perl, so if you find that Perl crashes when
1226 running GHC try updating (or downgrading) your Perl
1227 installation. Versions of Perl that we use and are known to
1228 be fairly stable are 5.005 and 5.6.1.</para>
1230 <para>For Win32 platforms, you should use the binary
1231 supplied in the InstallShield (copy it to
1232 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1235 <para>Perl should be put somewhere so that it can be invoked
1236 by the <literal>#!</literal> script-invoking
1237 mechanism. The full pathname may need to be less than 32
1238 characters long on some systems.</para>
1243 <term>GNU C (<command>gcc</command>)</term>
1244 <indexterm><primary>pre-supposed: GCC (GNU C
1245 compiler)</primary></indexterm> <indexterm><primary>GCC (GNU C
1246 compiler), pre-supposed</primary></indexterm>
1248 <para>We recommend using GCC version 2.95.2 on all
1249 platforms. Failing that, version 2.7.2 is stable on most
1250 platforms. Earlier versions of GCC can be assumed not to
1251 work, and versions in between 2.7.2 and 2.95.2 (including
1252 <command>egcs</command>) have varying degrees of stability
1253 depending on the platform.</para>
1255 <para>GCC 3.2 is currently known to have problems building
1256 GHC on Sparc, but is stable on x86.</para>
1258 <para>If your GCC dies with “internal error” on
1259 some GHC source file, please let us know, so we can report
1260 it and get things improved. (Exception: on x86
1261 boxes—you may need to fiddle with GHC's
1262 <option>-monly-N-regs</option> option; see the User's
1268 <term>GNU Make</term>
1269 <indexterm><primary>make</primary><secondary>GNU</secondary>
1272 <para>The fptools build system makes heavy use of features
1273 specific to GNU <command>make</command>, so you must have
1274 this installed in order to build any of the fptools
1281 <indexterm><primary>Happy</primary></indexterm>
1283 <para>Happy is a parser generator tool for Haskell, and is
1284 used to generate GHC's parsers. Happy is written in
1285 Haskell, and is a project in the CVS repository
1286 (<literal>fptools/happy</literal>). It can be built from
1287 source, but bear in mind that you'll need GHC installed in
1288 order to build it. To avoid the chicken/egg problem,
1289 install a binary distribution of either Happy or GHC to get
1290 started. Happy distributions are available from <ulink
1291 url="http://www.haskell.org/happy/">Happy's Web
1292 Page</ulink>.</para>
1298 <indexterm><primary>Alex</primary></indexterm>
1300 <para>Alex is a lexical-analyser generator for Haskell,
1301 which GHC uses to generate its lexer. Like Happy, Alex is
1302 written in Haskell and is a project in the CVS repository.
1303 Alex distributions are available from <ulink
1304 url="http://www.haskell.org/alex/">Alex's Web
1305 Page</ulink>.</para>
1310 <term>autoconf</term>
1311 <indexterm><primary>pre-supposed: autoconf</primary></indexterm>
1312 <indexterm><primary>autoconf, pre-supposed</primary></indexterm>
1314 <para>GNU autoconf is needed if you intend to build from the
1315 CVS sources, it is <emphasis>not</emphasis> needed if you
1316 just intend to build a standard source distribution.</para>
1318 <para>Version 2.52 or later of the autoconf package is required.
1319 NB. version 2.13 will no longer work, as of GHC version
1322 <para><command>autoreconf</command> (from the autoconf package)
1323 recursively builds <command>configure</command> scripts from
1324 the corresponding <filename>configure.ac</filename> and
1325 <filename>aclocal.m4</filename> files. If you modify one of
1326 the latter files, you'll need <command>autoreconf</command> to
1327 rebuild the corresponding <filename>configure</filename>.</para>
1332 <term><command>sed</command></term>
1333 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1334 <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:</term>
1359 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1360 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1362 <para>PVM is the Parallel Virtual Machine on which
1363 Parallel Haskell programs run. (You only need this if you
1364 plan to run Parallel Haskell. Concurrent Haskell, which
1365 runs concurrent threads on a uniprocessor doesn't need
1366 it.) Underneath PVM, you can have (for example) a network
1367 of workstations (slow) or a multiprocessor box
1370 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1371 It is readily available on the net; I think I got it from
1372 <literal>research.att.com</literal>, in
1373 <filename>netlib</filename>.</para>
1375 <para>A PVM installation is slightly quirky, but easy to
1376 do. Just follow the <filename>Readme</filename>
1377 instructions.</para>
1382 <term><command>bash</command>:</term>
1383 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1385 <para>Sadly, the <command>gr2ps</command> script, used to
1386 convert “parallelism profiles” to PostScript,
1387 is written in Bash (GNU's Bourne Again shell). This bug
1388 will be fixed (someday).</para>
1394 <sect2 id="pre-supposed-other-tools">
1395 <title>Other useful tools</title>
1400 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1401 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1403 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1404 to build a couple of utilities in
1405 <literal>glafp-utils</literal>. Depending on your
1406 operating system, the supplied <command>lex</command> may
1407 or may not work; you should get the GNU version.</para>
1412 <para>More tools are required if you want to format the documentation
1413 that comes with GHC and other fptools projects. See <xref
1414 linkend="building-docs">.</para>
1418 <sect1 id="sec-building-from-source">
1419 <title>Building from source</title>
1421 <indexterm><primary>Building from source</primary></indexterm>
1422 <indexterm><primary>Source, building from</primary></indexterm>
1424 <para>You've been rash enough to want to build some of the Glasgow
1425 Functional Programming tools (GHC, Happy, nofib, etc.) from
1426 source. You've slurped the source, from the CVS repository or
1427 from a source distribution, and now you're sitting looking at a
1428 huge mound of bits, wondering what to do next.</para>
1430 <para>Gingerly, you type <command>make</command>. Wrong
1433 <para>This rest of this guide is intended for duffers like me, who
1434 aren't really interested in Makefiles and systems configurations,
1435 but who need a mental model of the interlocking pieces so that
1436 they can make them work, extend them consistently when adding new
1437 software, and lay hands on them gently when they don't
1440 <sect2 id="quick-start">
1441 <title>Quick Start</title>
1443 <para>If you are starting from a source distribution, and just
1444 want a completely standard build, then the following should
1447 <screen>$ autoreconf
1453 <para>For GHC, this will do a 2-stage bootstrap build of the
1454 compiler, with profiling libraries, and install the
1457 <para>If you want to do anything at all non-standard, or you
1458 want to do some development, read on...</para>
1461 <sect2 id="sec-source-tree">
1462 <title>Your source tree</title>
1464 <para>The source code is held in your <emphasis>source
1465 tree</emphasis>. The root directory of your source tree
1466 <emphasis>must</emphasis> contain the following directories and
1471 <para><filename>Makefile</filename>: the root
1476 <para><filename>mk/</filename>: the directory that contains
1477 the main Makefile code, shared by all the
1478 <literal>fptools</literal> software.</para>
1482 <para><filename>configure.ac</filename>,
1483 <filename>config.sub</filename>,
1484 <filename>config.guess</filename>: these files support the
1485 configuration process.</para>
1489 <para><filename>install-sh</filename>.</para>
1493 <para>All the other directories are individual
1494 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1495 system—for example, the Glasgow Haskell Compiler
1496 (<literal>ghc</literal>), the Happy parser generator
1497 (<literal>happy</literal>), the <literal>nofib</literal>
1498 benchmark suite, and so on. You can have zero or more of these.
1499 Needless to say, some of them are needed to build others.</para>
1501 <para>The important thing to remember is that even if you want
1502 only one project (<literal>happy</literal>, say), you must have
1503 a source tree whose root directory contains
1504 <filename>Makefile</filename>, <filename>mk/</filename>,
1505 <filename>configure.ac</filename>, and the project(s) you want
1506 (<filename>happy/</filename> in this case). You cannot get by
1507 with just the <filename>happy/</filename> directory.</para>
1511 <title>Build trees</title>
1512 <indexterm><primary>build trees</primary></indexterm>
1513 <indexterm><primary>link trees, for building</primary></indexterm>
1515 <para>If you just want to build the software once on a single
1516 platform, then your source tree can also be your build tree, and
1517 you can skip the rest of this section.</para>
1519 <para>We often want to build multiple versions of our software
1520 for different architectures, or with different options
1521 (e.g. profiling). It's very desirable to share a single copy of
1522 the source code among all these builds.</para>
1524 <para>So for every source tree we have zero or more
1525 <emphasis>build trees</emphasis>. Each build tree is initially
1526 an exact copy of the source tree, except that each file is a
1527 symbolic link to the source file, rather than being a copy of
1528 the source file. There are “standard” Unix
1529 utilities that make such copies, so standard that they go by
1531 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1532 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1533 are two (If you don't have either, the source distribution
1534 includes sources for the X11
1535 <command>lndir</command>—check out
1536 <filename>fptools/glafp-utils/lndir</filename>). See <Xref
1537 LinkEnd="sec-storysofar"> for a typical invocation.</para>
1539 <para>The build tree does not need to be anywhere near the
1540 source tree in the file system. Indeed, one advantage of
1541 separating the build tree from the source is that the build tree
1542 can be placed in a non-backed-up partition, saving your systems
1543 support people from backing up untold megabytes of
1544 easily-regenerated, and rapidly-changing, gubbins. The golden
1545 rule is that (with a single exception—<XRef
1546 LinkEnd="sec-build-config">) <emphasis>absolutely everything in
1547 the build tree is either a symbolic link to the source tree, or
1548 else is mechanically generated</emphasis>. It should be
1549 perfectly OK for your build tree to vanish overnight; an hour or
1550 two compiling and you're on the road again.</para>
1552 <para>You need to be a bit careful, though, that any new files
1553 you create (if you do any development work) are in the source
1554 tree, not a build tree!</para>
1556 <para>Remember, that the source files in the build tree are
1557 <emphasis>symbolic links</emphasis> to the files in the source
1558 tree. (The build tree soon accumulates lots of built files like
1559 <filename>Foo.o</filename>, as well.) You can
1560 <emphasis>delete</emphasis> a source file from the build tree
1561 without affecting the source tree (though it's an odd thing to
1562 do). On the other hand, if you <emphasis>edit</emphasis> a
1563 source file from the build tree, you'll edit the source-tree
1564 file directly. (You can set up Emacs so that if you edit a
1565 source file from the build tree, Emacs will silently create an
1566 edited copy of the source file in the build tree, leaving the
1567 source file unchanged; but the danger is that you think you've
1568 edited the source file whereas actually all you've done is edit
1569 the build-tree copy. More commonly you do want to edit the
1570 source file.)</para>
1572 <para>Like the source tree, the top level of your build tree
1573 must be (a linked copy of) the root directory of the
1574 <literal>fptools</literal> suite. Inside Makefiles, the root of
1575 your build tree is called
1576 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1577 In the rest of this document path names are relative to
1578 <constant>$(FPTOOLS_TOP)</constant> unless
1579 otherwise stated. For example, the file
1580 <filename>ghc/mk/target.mk</filename> is actually
1581 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</filename>.</para>
1584 <sect2 id="sec-build-config">
1585 <title>Getting the build you want</title>
1587 <para>When you build <literal>fptools</literal> you will be
1588 compiling code on a particular <emphasis>host
1589 platform</emphasis>, to run on a particular <emphasis>target
1590 platform</emphasis> (usually the same as the host
1591 platform)<indexterm><primary>platform</primary></indexterm>.
1592 The difficulty is that there are minor differences between
1593 different platforms; minor, but enough that the code needs to be
1594 a bit different for each. There are some big differences too:
1595 for a different architecture we need to build GHC with a
1596 different native-code generator.</para>
1598 <para>There are also knobs you can turn to control how the
1599 <literal>fptools</literal> software is built. For example, you
1600 might want to build GHC optimised (so that it runs fast) or
1601 unoptimised (so that you can compile it fast after you've
1602 modified it. Or, you might want to compile it with debugging on
1603 (so that extra consistency-checking code gets included) or off.
1606 <para>All of this stuff is called the
1607 <emphasis>configuration</emphasis> of your build. You set the
1608 configuration using a three-step process.</para>
1612 <term>Step 1: get ready for configuration.</term>
1614 <para>NOTE: if you're starting from a source distribution,
1615 rather than CVS sources, you can skip this step.</para>
1617 <para>Change directory to
1618 <constant>$(FPTOOLS_TOP)</constant> and
1619 issue the command</para>
1623 <indexterm><primary>autoreconf</primary></indexterm>
1624 <para>(with no arguments). This GNU program (recursively) converts
1625 <filename><constant>$(FPTOOLS_TOP)</constant>/configure.ac</filename> and
1626 <filename><constant>$(FPTOOLS_TOP)</constant>/aclocal.m4</filename>
1627 to a shell script called
1628 <filename><constant>$(FPTOOLS_TOP)</constant>/configure</filename>.
1629 If <command>autoreconf</command> bleats that it can't write the file <filename>configure</filename>,
1630 then delete the latter and try again. Note that you must use <command>autoreconf</command>,
1631 and not the old <command>autoconf</command>! If you erroneously use the latter, you'll get
1632 a message like "No rule to make target 'mk/config.h.in'".
1635 <para>Some projects, including GHC, have their own configure script.
1636 <command>autoreconf</command> takes care of that, too, so all you have
1637 to do is calling <command>autoreconf</command> in the top-level directory
1638 <constant>$(FPTOOLS_TOP)</constant>.</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>
1655 ./configure <optional><parameter>args</parameter></optional>
1658 <para><command>configure</command>'s mission is to scurry
1659 round your computer working out what architecture it has,
1660 what operating system, whether it has the
1661 <Function>vfork</Function> system call, where
1662 <command>tar</command> is kept, whether
1663 <command>gcc</command> is available, where various obscure
1664 <literal>#include</literal> files are, whether it's a
1665 leap year, and what the systems manager had for lunch. It
1666 communicates these snippets of information in two
1673 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1675 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1676 substituting for things between
1677 “<literal>@</literal>” brackets. So,
1678 “<literal>@HaveGcc@</literal>” will be
1679 replaced by “<literal>YES</literal>” or
1680 “<literal>NO</literal>” depending on what
1681 <command>configure</command> finds.
1682 <filename>mk/config.mk</filename> is included by every
1683 Makefile (directly or indirectly), so the
1684 configuration information is thereby communicated to
1685 all Makefiles.</para>
1689 <para> It translates
1690 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1692 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1693 The latter is <literal>#include</literal>d by
1694 various C programs, which can thereby make use of
1695 configuration information.</para>
1699 <para><command>configure</command> takes some optional
1700 arguments. Use <literal>./configure --help</literal> to
1701 get a list of the available arguments. Here are some of
1702 the ones you might need:</para>
1706 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1707 <indexterm><primary><literal>--with-ghc</literal></primary>
1710 <para>Specifies the path to an installed GHC which
1711 you would like to use. This compiler will be used
1712 for compiling GHC-specific code (eg. GHC itself).
1713 This option <emphasis>cannot</emphasis> be specified
1714 using <filename>build.mk</filename> (see later),
1715 because <command>configure</command> needs to
1716 auto-detect the version of GHC you're using. The
1717 default is to look for a compiler named
1718 <literal>ghc</literal> in your path.</para>
1723 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1724 <indexterm><primary><literal>--with-hc</literal></primary>
1727 <para>Specifies the path to any installed Haskell
1728 compiler. This compiler will be used for compiling
1729 generic Haskell code. The default is to use
1730 <literal>ghc</literal>.</para>
1735 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1736 <indexterm><primary><literal>--with-gcc</literal></primary>
1739 <para>Specifies the path to the installed GCC. This
1740 compiler will be used to compile all C files,
1741 <emphasis>except</emphasis> any generated by the
1742 installed Haskell compiler, which will have its own
1743 idea of which C compiler (if any) to use. The
1744 default is to use <literal>gcc</literal>.</para>
1752 <term>Step 3: build configuration.</term>
1754 <para>Next, you say how this build of
1755 <literal>fptools</literal> is to differ from the standard
1756 defaults by creating a new file
1757 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1758 <emphasis>in the build tree</emphasis>. This file is the
1759 one and only file you edit in the build tree, precisely
1760 because it says how this build differs from the source.
1761 (Just in case your build tree does die, you might want to
1762 keep a private directory of <filename>build.mk</filename>
1763 files, and use a symbolic link in each build tree to point
1764 to the appropriate one.) So
1765 <filename>mk/build.mk</filename> never exists in the
1766 source tree—you create one in each build tree from
1767 the template. We'll discuss what to put in it
1773 <para>And that's it for configuration. Simple, eh?</para>
1775 <para>What do you put in your build-specific configuration file
1776 <filename>mk/build.mk</filename>? <emphasis>For almost all
1777 purposes all you will do is put make variable definitions that
1778 override those in</emphasis>
1779 <filename>mk/config.mk.in</filename>. The whole point of
1780 <filename>mk/config.mk.in</filename>—and its derived
1781 counterpart <filename>mk/config.mk</filename>—is to define
1782 the build configuration. It is heavily commented, as you will
1783 see if you look at it. So generally, what you do is look at
1784 <filename>mk/config.mk.in</filename>, and add definitions in
1785 <filename>mk/build.mk</filename> that override any of the
1786 <filename>config.mk</filename> definitions that you want to
1787 change. (The override occurs because the main boilerplate file,
1788 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1789 includes <filename>build.mk</filename> after
1790 <filename>config.mk</filename>.)</para>
1792 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1793 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1795 <para>For example, <filename>config.mk.in</filename> contains
1796 the definition:</para>
1799 GhcHcOpts=-O -Rghc-timing
1802 <para>The accompanying comment explains that this is the list of
1803 flags passed to GHC when building GHC itself. For doing
1804 development, it is wise to add <literal>-DDEBUG</literal>, to
1805 enable debugging code. So you would add the following to
1806 <filename>build.mk</filename>:</para>
1808 <para>or, if you prefer,</para>
1811 GhcHcOpts += -DDEBUG
1814 <para>GNU <command>make</command> allows existing definitions to
1815 have new text appended using the “<literal>+=</literal>”
1816 operator, which is quite a convenient feature.)</para>
1818 <para>If you want to remove the <literal>-O</literal> as well (a
1819 good idea when developing, because the turn-around cycle gets a
1820 lot quicker), you can just override
1821 <literal>GhcLibHcOpts</literal> altogether:</para>
1824 GhcHcOpts=-DDEBUG -Rghc-timing
1827 <para>When reading <filename>config.mk.in</filename>, remember
1828 that anything between “@...@” signs is going to be substituted
1829 by <command>configure</command> later. You
1830 <emphasis>can</emphasis> override the resulting definition if
1831 you want, but you need to be a bit surer what you are doing.
1832 For example, there's a line that says:</para>
1838 <para>This defines the Make variables <constant>TAR</constant>
1839 to the pathname for a <command>tar</command> that
1840 <command>configure</command> finds somewhere. If you have your
1841 own pet <command>tar</command> you want to use instead, that's
1842 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1848 <para>You do not <emphasis>have</emphasis> to have a
1849 <filename>mk/build.mk</filename> file at all; if you don't,
1850 you'll get all the default settings from
1851 <filename>mk/config.mk.in</filename>.</para>
1853 <para>You can also use <filename>build.mk</filename> to override
1854 anything that <command>configure</command> got wrong. One place
1855 where this happens often is with the definition of
1856 <constant>FPTOOLS_TOP_ABS</constant>: this
1857 variable is supposed to be the canonical path to the top of your
1858 source tree, but if your system uses an automounter then the
1859 correct directory is hard to find automatically. If you find
1860 that <command>configure</command> has got it wrong, just put the
1861 correct definition in <filename>build.mk</filename>.</para>
1865 <sect2 id="sec-storysofar">
1866 <title>The story so far</title>
1868 <para>Let's summarise the steps you need to carry to get
1869 yourself a fully-configured build tree from scratch.</para>
1873 <para> Get your source tree from somewhere (CVS repository
1874 or source distribution). Say you call the root directory
1875 <filename>myfptools</filename> (it does not have to be
1876 called <filename>fptools</filename>). Make sure that you
1877 have the essential files (see <XRef
1878 LinkEnd="sec-source-tree">).</para>
1883 <para>(Optional) Use <command>lndir</command> or
1884 <command>mkshadowdir</command> to create a build tree.</para>
1888 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1891 <para>(N.B. <command>mkshadowdir</command>'s first argument
1892 is taken relative to its second.) You probably want to give
1893 the build tree a name that suggests its main defining
1894 characteristic (in your mind at least), in case you later
1899 <para>Change directory to the build tree. Everything is
1900 going to happen there now.</para>
1903 $ cd /scratch/joe-bloggs/myfptools-sun4
1909 <para>Prepare for system configuration:</para>
1915 <para>(You can skip this step if you are starting from a
1916 source distribution, and you already have
1917 <filename>configure</filename> and
1918 <filename>mk/config.h.in</filename>.)</para>
1922 <para>Do system configuration:</para>
1928 <para>Don't forget to check whether you need to add any
1929 arguments to <literal>configure</literal>; for example, a
1930 common requirement is to specify which GHC to use with
1931 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1935 <para>Create the file <filename>mk/build.mk</filename>,
1936 adding definitions for your desired configuration
1945 <para>You can make subsequent changes to
1946 <filename>mk/build.mk</filename> as often as you like. You do
1947 not have to run any further configuration programs to make these
1948 changes take effect. In theory you should, however, say
1949 <command>gmake clean</command>, <command>gmake all</command>,
1950 because configuration option changes could affect
1951 anything—but in practice you are likely to know what's
1956 <title>Making things</title>
1958 <para>At this point you have made yourself a fully-configured
1959 build tree, so you are ready to start building real
1962 <para>The first thing you need to know is that <emphasis>you
1963 must use GNU <command>make</command>, usually called
1964 <command>gmake</command>, not standard Unix
1965 <command>make</command></emphasis>. If you use standard Unix
1966 <command>make</command> you will get all sorts of error messages
1967 (but no damage) because the <literal>fptools</literal>
1968 <command>Makefiles</command> use GNU <command>make</command>'s
1969 facilities extensively.</para>
1971 <para>To just build the whole thing, <command>cd</command> to
1972 the top of your <literal>fptools</literal> tree and type
1973 <command>gmake</command>. This will prepare the tree and build
1974 the various projects in the correct order.</para>
1977 <sect2 id="sec-bootstrapping">
1978 <title>Bootstrapping GHC</title>
1980 <para>GHC requires a 2-stage bootstrap in order to provide
1981 full functionality, including GHCi. By a 2-stage bootstrap, we
1982 mean that the compiler is built once using the installed GHC,
1983 and then again using the compiler built in the first stage. You
1984 can also build a stage 3 compiler, but this normally isn't
1985 necessary except to verify that the stage 2 compiler is working
1988 <para>Note that when doing a bootstrap, the stage 1 compiler
1989 must be built, followed by the runtime system and libraries, and
1990 then the stage 2 compiler. The correct ordering is implemented
1991 by the top-level fptools <filename>Makefile</filename>, so if
1992 you want everything to work automatically it's best to start
1993 <command>make</command> from the top of the tree. When building
1994 GHC, the top-level fptools <filename>Makefile</filename> is set
1995 up to do a 2-stage bootstrap by default (when you say
1996 <command>make</command>). Some other targets it supports
2003 <para>Build everything as normal, including the stage 1
2011 <para>Build the stage 2 compiler only.</para>
2018 <para>Build the stage 3 compiler only.</para>
2023 <term>bootstrap</term> <term>bootstrap2</term>
2025 <para>Build stage 1 followed by stage 2.</para>
2030 <term>bootstrap3</term>
2032 <para>Build stages 1, 2 and 3.</para>
2037 <term>install</term>
2039 <para>Install everything, including the compiler built in
2040 stage 2. To override the stage, say <literal>make install
2041 stage=<replaceable>n</replaceable></literal> where
2042 <replaceable>n</replaceable> is the stage to install.</para>
2047 <para>The top-level <filename>Makefile</filename> also arranges
2048 to do the appropriate <literal>make boot</literal> steps (see
2049 below) before actually building anything.</para>
2051 <para>The <literal>stage1</literal>, <literal>stage2</literal>
2052 and <literal>stage3</literal> targets also work in the
2053 <literal>ghc/compiler</literal> directory, but don't forget that
2054 each stage requires its own <literal>make boot</literal> step:
2055 for example, you must do</para>
2057 <screen>$ make boot stage=2</screen>
2059 <para>before <literal>make stage2</literal> in
2060 <literal>ghc/compiler</literal>.</para>
2063 <sect2 id="sec-standard-targets">
2064 <title>Standard Targets</title>
2065 <indexterm><primary>targets, standard makefile</primary></indexterm>
2066 <indexterm><primary>makefile targets</primary></indexterm>
2068 <para>In any directory you should be able to make the following:</para>
2072 <term><literal>boot</literal></term>
2074 <para>does the one-off preparation required to get ready
2075 for the real work. Notably, it does <command>gmake
2076 depend</command> in all directories that contain programs.
2077 It also builds the necessary tools for compilation to
2080 <para>Invoking the <literal>boot</literal> target
2081 explicitly is not normally necessary. From the top-level
2082 <literal>fptools</literal> directory, invoking
2083 <literal>gmake</literal> causes <literal>gmake boot
2084 all</literal> to be invoked in each of the project
2085 subdirectories, in the order specified by
2086 <literal>$(AllTargets)</literal> in
2087 <literal>config.mk</literal>.</para>
2089 <para>If you're working in a subdirectory somewhere and
2090 need to update the dependencies, <literal>gmake
2091 boot</literal> is a good way to do it.</para>
2096 <term><literal>all</literal></term>
2098 <para>makes all the final target(s) for this Makefile.
2099 Depending on which directory you are in a “final
2100 target” may be an executable program, a library
2101 archive, a shell script, or a Postscript file. Typing
2102 <command>gmake</command> alone is generally the same as
2103 typing <command>gmake all</command>.</para>
2108 <term><literal>install</literal></term>
2110 <para>installs the things built by <literal>all</literal>
2111 (except for the documentation). Where does it install
2112 them? That is specified by
2113 <filename>mk/config.mk.in</filename>; you can override it
2114 in <filename>mk/build.mk</filename>, or by running
2115 <command>configure</command> with command-line arguments
2116 like <literal>--bindir=/home/simonpj/bin</literal>; see
2117 <literal>./configure --help</literal> for the full
2123 <term><literal>install-docs</literal></term>
2125 <para>installs the documentation. Otherwise behaves just
2126 like <literal>install</literal>.</para>
2131 <term><literal>uninstall</literal></term>
2133 <para>reverses the effect of
2134 <literal>install</literal>.</para>
2139 <term><literal>clean</literal></term>
2141 <para>Delete all files from the current directory that are
2142 normally created by building the program. Don't delete
2143 the files that record the configuration, or files
2144 generated by <command>gmake boot</command>. Also preserve
2145 files that could be made by building, but normally aren't
2146 because the distribution comes with them.</para>
2151 <term><literal>distclean</literal></term>
2153 <para>Delete all files from the current directory that are
2154 created by configuring or building the program. If you
2155 have unpacked the source and built the program without
2156 creating any other files, <literal>make
2157 distclean</literal> should leave only the files that were
2158 in the distribution.</para>
2163 <term><literal>mostlyclean</literal></term>
2165 <para>Like <literal>clean</literal>, but may refrain from
2166 deleting a few files that people normally don't want to
2172 <term><literal>maintainer-clean</literal></term>
2174 <para>Delete everything from the current directory that
2175 can be reconstructed with this Makefile. This typically
2176 includes everything deleted by
2177 <literal>distclean</literal>, plus more: C source files
2178 produced by Bison, tags tables, Info files, and so
2181 <para>One exception, however: <literal>make
2182 maintainer-clean</literal> should not delete
2183 <filename>configure</filename> even if
2184 <filename>configure</filename> can be remade using a rule
2185 in the <filename>Makefile</filename>. More generally,
2186 <literal>make maintainer-clean</literal> should not delete
2187 anything that needs to exist in order to run
2188 <filename>configure</filename> and then begin to build the
2194 <term><literal>check</literal></term>
2196 <para>run the test suite.</para>
2201 <para>All of these standard targets automatically recurse into
2202 sub-directories. Certain other standard targets do not:</para>
2206 <term><literal>configure</literal></term>
2208 <para>is only available in the root directory
2209 <constant>$(FPTOOLS_TOP)</constant>; it has
2210 been discussed in <XRef
2211 LinkEnd="sec-build-config">.</para>
2216 <term><literal>depend</literal></term>
2218 <para>make a <filename>.depend</filename> file in each
2219 directory that needs it. This <filename>.depend</filename>
2220 file contains mechanically-generated dependency
2221 information; for example, suppose a directory contains a
2222 Haskell source module <filename>Foo.lhs</filename> which
2223 imports another module <literal>Baz</literal>. Then the
2224 generated <filename>.depend</filename> file will contain
2225 the dependency:</para>
2231 <para>which says that the object file
2232 <filename>Foo.o</filename> depends on the interface file
2233 <filename>Baz.hi</filename> generated by compiling module
2234 <literal>Baz</literal>. The <filename>.depend</filename>
2235 file is automatically included by every Makefile.</para>
2240 <term><literal>binary-dist</literal></term>
2242 <para>make a binary distribution. This is the target we
2243 use to build the binary distributions of GHC and
2249 <term><literal>dist</literal></term>
2251 <para>make a source distribution. Note that this target
2252 does “make distclean” as part of its work;
2253 don't use it if you want to keep what you've built.</para>
2258 <para>Most <filename>Makefile</filename>s have targets other
2259 than these. You can discover them by looking in the
2260 <filename>Makefile</filename> itself.</para>
2264 <title>Using a project from the build tree</title>
2266 <para>If you want to build GHC (say) and just use it direct from
2267 the build tree without doing <literal>make install</literal>
2268 first, you can run the in-place driver script:
2269 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2271 <para> Do <emphasis>NOT</emphasis> use
2272 <filename>ghc/compiler/ghc</filename>, or
2273 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2274 scripts intended for installation, and contain hard-wired paths
2275 to the installed libraries, rather than the libraries in the
2278 <para>Happy can similarly be run from the build tree, using
2279 <filename>happy/src/happy-inplace</filename>, and similarly for
2280 Alex and Haddock.</para>
2284 <title>Fast Making</title>
2286 <indexterm><primary>fastmake</primary></indexterm>
2287 <indexterm><primary>dependencies, omitting</primary></indexterm>
2288 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2290 <para>Sometimes the dependencies get in the way: if you've made
2291 a small change to one file, and you're absolutely sure that it
2292 won't affect anything else, but you know that
2293 <command>make</command> is going to rebuild everything anyway,
2294 the following hack may be useful:</para>
2300 <para>This tells the make system to ignore dependencies and just
2301 build what you tell it to. In other words, it's equivalent to
2302 temporarily removing the <filename>.depend</filename> file in
2303 the current directory (where <command>mkdependHS</command> and
2304 friends store their dependency information).</para>
2306 <para>A bit of history: GHC used to come with a
2307 <command>fastmake</command> script that did the above job, but
2308 GNU make provides the features we need to do it without
2309 resorting to a script. Also, we've found that fastmaking is
2310 less useful since the advent of GHC's recompilation checker (see
2311 the User's Guide section on "Separate Compilation").</para>
2315 <sect1 id="sec-makefile-arch">
2316 <title>The <filename>Makefile</filename> architecture</title>
2317 <indexterm><primary>makefile architecture</primary></indexterm>
2319 <para><command>make</command> is great if everything
2320 works—you type <command>gmake install</command> and lo! the
2321 right things get compiled and installed in the right places. Our
2322 goal is to make this happen often, but somehow it often doesn't;
2323 instead some weird error message eventually emerges from the
2324 bowels of a directory you didn't know existed.</para>
2326 <para>The purpose of this section is to give you a road-map to
2327 help you figure out what is going right and what is going
2331 <title>Debugging</title>
2333 <para>Debugging <filename>Makefile</filename>s is something of a
2334 black art, but here's a couple of tricks that we find
2335 particularly useful. The following command allows you to see
2336 the contents of any make variable in the context of the current
2337 <filename>Makefile</filename>:</para>
2339 <screen>$ make show VALUE=HS_SRCS</screen>
2341 <para>where you can replace <literal>HS_SRCS</literal> with the
2342 name of any variable you wish to see the value of.</para>
2344 <para>GNU make has a <option>-d</option> option which generates
2345 a dump of the decision procedure used to arrive at a conclusion
2346 about which files should be recompiled. Sometimes useful for
2347 tracking down problems with superfluous or missing
2348 recompilations.</para>
2352 <title>A small project</title>
2354 <para>To get started, let us look at the
2355 <filename>Makefile</filename> for an imaginary small
2356 <literal>fptools</literal> project, <literal>small</literal>.
2357 Each project in <literal>fptools</literal> has its own directory
2358 in <constant>FPTOOLS_TOP</constant>, so the
2359 <literal>small</literal> project will have its own directory
2360 <constant>FPOOLS_TOP/small/</constant>. Inside the
2361 <filename>small/</filename> directory there will be a
2362 <filename>Makefile</filename>, looking something like
2365 <indexterm><primary>Makefile, minimal</primary></indexterm>
2368 # Makefile for fptools project "small"
2371 include $(TOP)/mk/boilerplate.mk
2373 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2376 include $(TOP)/target.mk
2379 <para>this <filename>Makefile</filename> has three
2384 <para>The first section includes
2387 One of the most important
2388 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2389 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2394 a file of “boilerplate” code from the level
2395 above (which in this case will be
2396 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2397 As its name suggests, <filename>boilerplate.mk</filename>
2398 consists of a large quantity of standard
2399 <filename>Makefile</filename> code. We discuss this
2400 boilerplate in more detail in <XRef LinkEnd="sec-boiler">.
2401 <indexterm><primary>include, directive in
2402 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2403 inclusion</primary></indexterm></para>
2405 <para>Before the <literal>include</literal> statement, you
2406 must define the <command>make</command> variable
2407 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2408 to be the directory containing the <filename>mk</filename>
2409 directory in which the <filename>boilerplate.mk</filename>
2410 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2413 include ../mk/boilerplate.mk # NO NO NO
2417 <para>Why? Because the <filename>boilerplate.mk</filename>
2418 file needs to know where it is, so that it can, in turn,
2419 <literal>include</literal> other files. (Unfortunately,
2420 when an <literal>include</literal>d file does an
2421 <literal>include</literal>, the filename is treated relative
2422 to the directory in which <command>gmake</command> is being
2423 run, not the directory in which the
2424 <literal>include</literal>d sits.) In general,
2425 <emphasis>every file <filename>foo.mk</filename> assumes
2427 <filename><constant>$(TOP)</constant>/mk/foo.mk</filename>
2428 refers to itself.</emphasis> It is up to the
2429 <filename>Makefile</filename> doing the
2430 <literal>include</literal> to ensure this is the case.</para>
2432 <para>Files intended for inclusion in other
2433 <filename>Makefile</filename>s are written to have the
2434 following property: <emphasis>after
2435 <filename>foo.mk</filename> is <literal>include</literal>d,
2436 it leaves <constant>TOP</constant> containing the same value
2437 as it had just before the <literal>include</literal>
2438 statement</emphasis>. In our example, this invariant
2439 guarantees that the <literal>include</literal> for
2440 <filename>target.mk</filename> will look in the same
2441 directory as that for <filename>boilerplate.mk</filename>.</para>
2445 <para> The second section defines the following standard
2446 <command>make</command> variables:
2447 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2448 (the source files from which is to be built), and
2449 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2450 (the executable binary to be built). We will discuss in
2451 more detail what the “standard variables” are,
2452 and how they affect what happens, in <XRef
2453 LinkEnd="sec-targets">.</para>
2455 <para>The definition for <constant>SRCS</constant> uses the
2456 useful GNU <command>make</command> construct
2457 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2458 which expands to a list of all the files matching the
2459 pattern <literal>pat</literal> in the current directory. In
2460 this example, <constant>SRCS</constant> is set to the list
2461 of all the <filename>.lhs</filename> and
2462 <filename>.c</filename> files in the directory. (Let's
2463 suppose there is one of each, <filename>Foo.lhs</filename>
2464 and <filename>Baz.c</filename>.)</para>
2468 <para>The last section includes a second file of standard
2470 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2471 It contains the rules that tell <command>gmake</command> how
2472 to make the standard targets (<Xref
2473 LinkEnd="sec-standard-targets">). Why, you ask, can't this
2474 standard code be part of
2475 <filename>boilerplate.mk</filename>? Good question. We
2476 discuss the reason later, in <Xref
2477 LinkEnd="sec-boiler-arch">.</para>
2479 <para>You do not <emphasis>have</emphasis> to
2480 <literal>include</literal> the
2481 <filename>target.mk</filename> file. Instead, you can write
2482 rules of your own for all the standard targets. Usually,
2483 though, you will find quite a big payoff from using the
2484 canned rules in <filename>target.mk</filename>; the price
2485 tag is that you have to understand what canned rules get
2486 enabled, and what they do (<Xref
2487 LinkEnd="sec-targets">).</para>
2491 <para>In our example <filename>Makefile</filename>, most of the
2492 work is done by the two <literal>include</literal>d files. When
2493 you say <command>gmake all</command>, the following things
2498 <para><command>gmake</command> figures out that the object
2499 files are <filename>Foo.o</filename> and
2500 <filename>Baz.o</filename>.</para>
2504 <para>It uses a boilerplate pattern rule to compile
2505 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2506 using a Haskell compiler. (Which one? That is set in the
2507 build configuration.)</para>
2511 <para>It uses another standard pattern rule to compile
2512 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2513 using a C compiler. (Ditto.)</para>
2517 <para>It links the resulting <filename>.o</filename> files
2518 together to make <literal>small</literal>, using the Haskell
2519 compiler to do the link step. (Why not use
2520 <command>ld</command>? Because the Haskell compiler knows
2521 what standard libraries to link in. How did
2522 <command>gmake</command> know to use the Haskell compiler to
2523 do the link, rather than the C compiler? Because we set the
2524 variable <constant>HS_PROG</constant> rather than
2525 <constant>C_PROG</constant>.)</para>
2529 <para>All <filename>Makefile</filename>s should follow the above
2530 three-section format.</para>
2534 <title>A larger project</title>
2536 <para>Larger projects are usually structured into a number of
2537 sub-directories, each of which has its own
2538 <filename>Makefile</filename>. (In very large projects, this
2539 sub-structure might be iterated recursively, though that is
2540 rare.) To give you the idea, here's part of the directory
2541 structure for the (rather large) GHC project:</para>
2551 ...source files for documentation...
2554 ...source files for driver...
2557 parser/...source files for parser...
2558 renamer/...source files for renamer...
2562 <para>The sub-directories <filename>docs</filename>,
2563 <filename>driver</filename>, <filename>compiler</filename>, and
2564 so on, each contains a sub-component of GHC, and each has its
2565 own <filename>Makefile</filename>. There must also be a
2566 <filename>Makefile</filename> in
2567 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>.
2568 It does most of its work by recursively invoking
2569 <command>gmake</command> on the <filename>Makefile</filename>s
2570 in the sub-directories. We say that
2571 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2572 <filename>Makefile</filename></emphasis>, because it does little
2573 except organise its children, while the
2574 <filename>Makefile</filename>s in the sub-directories are all
2575 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2576 principle the sub-directories might themselves contain a
2577 non-leaf <filename>Makefile</filename> and several
2578 sub-sub-directories, but that does not happen in GHC.)</para>
2580 <para>The <filename>Makefile</filename> in
2581 <filename>ghc/compiler</filename> is considered a leaf
2582 <filename>Makefile</filename> even though the
2583 <filename>ghc/compiler</filename> has sub-directories, because
2584 these sub-directories do not themselves have
2585 <filename>Makefile</filename>s in them. They are just used to
2586 structure the collection of modules that make up GHC, but all
2587 are managed by the single <filename>Makefile</filename> in
2588 <filename>ghc/compiler</filename>.</para>
2590 <para>You will notice that <filename>ghc/</filename> also
2591 contains a directory <filename>ghc/mk/</filename>. It contains
2592 GHC-specific <filename>Makefile</filename> boilerplate code.
2593 More precisely:</para>
2597 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2598 at the top of <filename>ghc/Makefile</filename>, and of all
2599 the leaf <filename>Makefile</filename>s in the
2600 sub-directories. It in turn <literal>include</literal>s the
2601 main boilerplate file
2602 <filename>mk/boilerplate.mk</filename>.</para>
2606 <para><filename>ghc/mk/target.mk</filename> is
2607 <literal>include</literal>d at the bottom of
2608 <filename>ghc/Makefile</filename>, and of all the leaf
2609 <filename>Makefile</filename>s in the sub-directories. It
2610 in turn <literal>include</literal>s the file
2611 <filename>mk/target.mk</filename>.</para>
2615 <para>So these two files are the place to look for GHC-wide
2616 customisation of the standard boilerplate.</para>
2619 <sect2 id="sec-boiler-arch">
2620 <title>Boilerplate architecture</title>
2621 <indexterm><primary>boilerplate architecture</primary></indexterm>
2623 <para>Every <filename>Makefile</filename> includes a
2624 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2625 file at the top, and
2626 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2627 file at the bottom. In this section we discuss what is in these
2628 files, and why there have to be two of them. In general:</para>
2632 <para><filename>boilerplate.mk</filename> consists of:</para>
2636 <para><emphasis>Definitions of millions of
2637 <command>make</command> variables</emphasis> that
2638 collectively specify the build configuration. Examples:
2639 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2640 the options to feed to the Haskell compiler;
2641 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2642 the sub-directories to enable within the
2643 <literal>nofib</literal> project;
2644 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2645 the name of the Haskell compiler to use when compiling
2646 GHC in the <literal>ghc</literal> project.</para>
2650 <para><emphasis>Standard pattern rules</emphasis> that
2651 tell <command>gmake</command> how to construct one file
2652 from another.</para>
2656 <para><filename>boilerplate.mk</filename> needs to be
2657 <literal>include</literal>d at the <emphasis>top</emphasis>
2658 of each <filename>Makefile</filename>, so that the user can
2659 replace the boilerplate definitions or pattern rules by
2660 simply giving a new definition or pattern rule in the
2661 <filename>Makefile</filename>. <command>gmake</command>
2662 simply takes the last definition as the definitive one.</para>
2664 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2665 definitions, it is also quite common to
2666 <emphasis>augment</emphasis> them. For example, a
2667 <filename>Makefile</filename> might say:</para>
2673 <para>thereby adding “<option>-O</option>” to
2675 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2679 <para><filename>target.mk</filename> contains
2680 <command>make</command> rules for the standard targets
2681 described in <Xref LinkEnd="sec-standard-targets">. These
2682 rules are selectively included, depending on the setting of
2683 certain <command>make</command> variables. These variables
2684 are usually set in the middle section of the
2685 <filename>Makefile</filename> between the two
2686 <literal>include</literal>s.</para>
2688 <para><filename>target.mk</filename> must be included at the
2689 end (rather than being part of
2690 <filename>boilerplate.mk</filename>) for several tiresome
2696 <para><command>gmake</command> commits target and
2697 dependency lists earlier than it should. For example,
2698 <FIlename>target.mk</FIlename> has a rule that looks
2702 $(HS_PROG) : $(OBJS)
2703 $(HC) $(LD_OPTS) $< -o $@
2706 <para>If this rule was in
2707 <filename>boilerplate.mk</filename> then
2708 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2710 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2711 would not have their final values at the moment
2712 <command>gmake</command> encountered the rule. Alas,
2713 <command>gmake</command> takes a snapshot of their
2714 current values, and wires that snapshot into the rule.
2715 (In contrast, the commands executed when the rule
2716 “fires” are only substituted at the moment
2717 of firing.) So, the rule must follow the definitions
2718 given in the <filename>Makefile</filename> itself.</para>
2722 <para>Unlike pattern rules, ordinary rules cannot be
2723 overriden or replaced by subsequent rules for the same
2724 target (at least, not without an error message).
2725 Including ordinary rules in
2726 <filename>boilerplate.mk</filename> would prevent the
2727 user from writing rules for specific targets in specific
2732 <para>There are a couple of other reasons I've
2733 forgotten, but it doesn't matter too much.</para>
2740 <sect2 id="sec-boiler">
2741 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2742 <indexterm><primary>boilerplate.mk</primary></indexterm>
2744 <para>If you look at
2745 <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename>
2746 you will find that it consists of the following sections, each
2747 held in a separate file:</para>
2751 <term><filename>config.mk</filename></term>
2752 <indexterm><primary>config.mk</primary></indexterm>
2754 <para>is the build configuration file we discussed at
2755 length in <Xref LinkEnd="sec-build-config">.</para>
2760 <term><filename>paths.mk</filename></term>
2761 <indexterm><primary>paths.mk</primary></indexterm>
2763 <para>defines <command>make</command> variables for
2764 pathnames and file lists. This file contains code for
2765 automatically compiling lists of source files and deriving
2766 lists of object files from those. The results can be
2767 overriden in the <filename>Makefile</filename>, but in
2768 most cases the automatic setup should do the right
2771 <para>The following variables may be set in the
2772 <filename>Makefile</filename> to affect how the automatic
2773 source file search is done:</para>
2777 <term><literal>ALL_DIRS</literal></term>
2778 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2781 <para>Set to a list of directories to search in
2782 addition to the current directory for source
2788 <term><literal>EXCLUDE_SRCS</literal></term>
2789 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2792 <para>Set to a list of source files (relative to the
2793 current directory) to omit from the automatic
2794 search. The source searching machinery is clever
2795 enough to know that if you exclude a source file
2796 from which other sources are derived, then the
2797 derived sources should also be excluded. For
2798 example, if you set <literal>EXCLUDED_SRCS</literal>
2799 to include <filename>Foo.y</filename>, then
2800 <filename>Foo.hs</filename> will also be
2806 <term><literal>EXTRA_SRCS</literal></term>
2807 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2810 <para>Set to a list of extra source files (perhaps
2811 in directories not listed in
2812 <literal>ALL_DIRS</literal>) that should be
2818 <para>The results of the automatic source file search are
2819 placed in the following make variables:</para>
2823 <term><literal>SRCS</literal></term>
2824 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2826 <para>All source files found, sorted and without
2827 duplicates, including those which might not exist
2828 yet but will be derived from other existing sources.
2829 <literal>SRCS</literal> <emphasis>can</emphasis> be
2830 overriden if necessary, in which case the variables
2831 below will follow suit.</para>
2836 <term><literal>HS_SRCS</literal></term>
2837 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2839 <para>all Haskell source files in the current
2840 directory, including those derived from other source
2841 files (eg. Happy sources also give rise to Haskell
2847 <term><literal>HS_OBJS</literal></term>
2848 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2850 <para>Object files derived from
2851 <literal>HS_SRCS</literal>.</para>
2856 <term><literal>HS_IFACES</literal></term>
2857 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2859 <para>Interface files (<literal>.hi</literal> files)
2860 derived from <literal>HS_SRCS</literal>.</para>
2865 <term><literal>C_SRCS</literal></term>
2866 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2868 <para>All C source files found.</para>
2873 <term><literal>C_OBJS</literal></term>
2874 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2876 <para>Object files derived from
2877 <literal>C_SRCS</literal>.</para>
2882 <term><literal>SCRIPT_SRCS</literal></term>
2883 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2885 <para>All script source files found
2886 (<literal>.lprl</literal> files).</para>
2891 <term><literal>SCRIPT_OBJS</literal></term>
2892 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2894 <para><quote>object</quote> files derived from
2895 <literal>SCRIPT_SRCS</literal>
2896 (<literal>.prl</literal> files).</para>
2901 <term><literal>HSC_SRCS</literal></term>
2902 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2904 <para>All <literal>hsc2hs</literal> source files
2905 (<literal>.hsc</literal> files).</para>
2910 <term><literal>HAPPY_SRCS</literal></term>
2911 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2913 <para>All <literal>happy</literal> source files
2914 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2919 <term><literal>OBJS</literal></term>
2920 <indexterm><primary>OBJS</primary></indexterm>
2922 <para>the concatenation of
2923 <literal>$(HS_OBJS)</literal>,
2924 <literal>$(C_OBJS)</literal>, and
2925 <literal>$(SCRIPT_OBJS)</literal>.</para>
2930 <para>Any or all of these definitions can easily be
2931 overriden by giving new definitions in your
2932 <filename>Makefile</filename>.</para>
2934 <para>What, exactly, does <filename>paths.mk</filename>
2935 consider a <quote>source file</quote> to be? It's based
2936 on the file's suffix (e.g. <filename>.hs</filename>,
2937 <filename>.lhs</filename>, <filename>.c</filename>,
2938 <filename>.hy</filename>, etc), but this is the kind of
2939 detail that changes, so rather than enumerate the source
2940 suffices here the best thing to do is to look in
2941 <filename>paths.mk</filename>.</para>
2946 <term><filename>opts.mk</filename></term>
2947 <indexterm><primary>opts.mk</primary></indexterm>
2949 <para>defines <command>make</command> variables for option
2950 strings to pass to each program. For example, it defines
2951 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2952 the option strings to pass to the Haskell compiler. See
2953 <Xref LinkEnd="sec-suffix">.</para>
2958 <term><filename>suffix.mk</filename></term>
2959 <indexterm><primary>suffix.mk</primary></indexterm>
2961 <para>defines standard pattern rules—see <Xref
2962 LinkEnd="sec-suffix">.</para>
2967 <para>Any of the variables and pattern rules defined by the
2968 boilerplate file can easily be overridden in any particular
2969 <filename>Makefile</filename>, because the boilerplate
2970 <literal>include</literal> comes first. Definitions after this
2971 <literal>include</literal> directive simply override the default
2972 ones in <filename>boilerplate.mk</filename>.</para>
2975 <sect2 id="sec-suffix">
2976 <title>Pattern rules and options</title>
2977 <indexterm><primary>Pattern rules</primary></indexterm>
2980 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2981 defines standard <emphasis>pattern rules</emphasis> that say how
2982 to build one kind of file from another, for example, how to
2983 build a <filename>.o</filename> file from a
2984 <filename>.c</filename> file. (GNU <command>make</command>'s
2985 <emphasis>pattern rules</emphasis> are more powerful and easier
2986 to use than Unix <command>make</command>'s <emphasis>suffix
2987 rules</emphasis>.)</para>
2989 <para>Almost all the rules look something like this:</para>
2994 $(CC) $(CC_OPTS) -c $< -o $@
2997 <para>Here's how to understand the rule. It says that
2998 <emphasis>something</emphasis><filename>.o</filename> (say
2999 <filename>Foo.o</filename>) can be built from
3000 <emphasis>something</emphasis><filename>.c</filename>
3001 (<filename>Foo.c</filename>), by invoking the C compiler (path
3002 name held in <constant>$(CC)</constant>), passing to it
3003 the options <constant>$(CC_OPTS)</constant> and
3004 the rule's dependent file of the rule
3005 <literal>$<</literal> (<filename>Foo.c</filename> in
3006 this case), and putting the result in the rule's target
3007 <literal>$@</literal> (<filename>Foo.o</filename> in this
3010 <para>Every program is held in a <command>make</command>
3011 variable defined in <filename>mk/config.mk</filename>—look
3012 in <filename>mk/config.mk</filename> for the complete list. One
3013 important one is the Haskell compiler, which is called
3014 <constant>$(HC)</constant>.</para>
3016 <para>Every program's options are are held in a
3017 <command>make</command> variables called
3018 <constant><prog>_OPTS</constant>. the
3019 <constant><prog>_OPTS</constant> variables are
3020 defined in <filename>mk/opts.mk</filename>. Almost all of them
3021 are defined like this:</para>
3024 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
3027 <para>The four variables from which
3028 <constant>CC_OPTS</constant> is built have the following
3033 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
3035 <para>options passed to all C compilations.</para>
3040 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
3042 <para>options passed to C compilations for way
3043 <literal><way></literal>. For example,
3044 <constant>WAY_mp_CC_OPTS</constant>
3045 gives options to pass to the C compiler when compiling way
3046 <literal>mp</literal>. The variable
3047 <constant>WAY_CC_OPTS</constant> holds
3048 options to pass to the C compiler when compiling the
3049 standard way. (<Xref LinkEnd="sec-ways"> dicusses
3050 multi-way compilation.)</para>
3055 <term><constant><module>_CC_OPTS</constant>:</term>
3057 <para>options to pass to the C compiler that are specific
3058 to module <literal><module></literal>. For example,
3059 <constant>SMap_CC_OPTS</constant> gives the
3060 specific options to pass to the C compiler when compiling
3061 <filename>SMap.c</filename>.</para>
3066 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3068 <para>extra options to pass to all C compilations. This
3069 is intended for command line use, thus:</para>
3072 gmake libHS.a EXTRA_CC_OPTS="-v"
3079 <sect2 id="sec-targets">
3080 <title>The main <filename>mk/target.mk</filename> file</title>
3081 <indexterm><primary>target.mk</primary></indexterm>
3083 <para><filename>target.mk</filename> contains canned rules for
3084 all the standard targets described in <Xref
3085 LinkEnd="sec-standard-targets">. It is complicated by the fact
3086 that you don't want all of these rules to be active in every
3087 <filename>Makefile</filename>. Rather than have a plethora of
3088 tiny files which you can include selectively, there is a single
3089 file, <filename>target.mk</filename>, which selectively includes
3090 rules based on whether you have defined certain variables in
3091 your <filename>Makefile</filename>. This section explains what
3092 rules you get, what variables control them, and what the rules
3093 do. Hopefully, you will also get enough of an idea of what is
3094 supposed to happen that you can read and understand any weird
3095 special cases yourself.</para>
3099 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3101 <para>If <constant>HS_PROG</constant> is defined,
3102 you get rules with the following targets:</para>
3106 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3108 <para>itself. This rule links
3109 <constant>$(OBJS)</constant> with the Haskell
3110 runtime system to get an executable called
3111 <constant>$(HS_PROG)</constant>.</para>
3116 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3119 <constant>$(HS_PROG)</constant> in
3120 <constant>$(bindir)</constant>.</para>
3129 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3131 <para>is similar to <constant>HS_PROG</constant>,
3132 except that the link step links
3133 <constant>$(C_OBJS)</constant> with the C
3134 runtime system.</para>
3139 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3141 <para>is similar to <constant>HS_PROG</constant>,
3142 except that it links
3143 <constant>$(LIB_OBJS)</constant> to make the
3144 library archive <constant>$(LIBRARY)</constant>,
3145 and <literal>install</literal> installs it in
3146 <constant>$(libdir)</constant>.</para>
3151 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3153 <para>…</para>
3158 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3160 <para>…</para>
3165 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3167 <para>If <constant>HS_SRCS</constant> is defined
3168 and non-empty, a rule for the target
3169 <literal>depend</literal> is included, which generates
3170 dependency information for Haskell programs. Similarly
3171 for <constant>C_SRCS</constant>.</para>
3176 <para>All of these rules are “double-colon” rules,
3180 install :: $(HS_PROG)
3181 ...how to install it...
3184 <para>GNU <command>make</command> treats double-colon rules as
3185 separate entities. If there are several double-colon rules for
3186 the same target it takes each in turn and fires it if its
3187 dependencies say to do so. This means that you can, for
3188 example, define both <constant>HS_PROG</constant> and
3189 <constant>LIBRARY</constant>, which will generate two rules for
3190 <literal>install</literal>. When you type <command>gmake
3191 install</command> both rules will be fired, and both the program
3192 and the library will be installed, just as you wanted.</para>
3195 <sect2 id="sec-subdirs">
3196 <title>Recursion</title>
3197 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3198 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3200 <para>In leaf <filename>Makefile</filename>s the variable
3201 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3202 is undefined. In non-leaf <filename>Makefile</filename>s,
3203 <constant>SUBDIRS</constant> is set to the list of
3204 sub-directories that contain subordinate
3205 <filename>Makefile</filename>s. <emphasis>It is up to you to
3206 set <constant>SUBDIRS</constant> in the
3207 <filename>Makefile</filename>.</emphasis> There is no automation
3208 here—<constant>SUBDIRS</constant> is too important to
3211 <para>When <constant>SUBDIRS</constant> is defined,
3212 <filename>target.mk</filename> includes a rather neat rule for
3213 the standard targets (<Xref LinkEnd="sec-standard-targets"> that
3214 simply invokes <command>make</command> recursively in each of
3215 the sub-directories.</para>
3217 <para><emphasis>These recursive invocations are guaranteed to
3218 occur in the order in which the list of directories is specified
3219 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3220 be important. For example, when you say <command>gmake
3221 boot</command> it can be important that the recursive invocation
3222 of <command>make boot</command> is done in one sub-directory
3223 (the include files, say) before another (the source files).
3224 Generally, put the most independent sub-directory first, and the
3225 most dependent last.</para>
3228 <sect2 id="sec-ways">
3229 <title>Way management</title>
3230 <indexterm><primary>way management</primary></indexterm>
3232 <para>We sometimes want to build essentially the same system in
3233 several different “ways”. For example, we want to build GHC's
3234 <literal>Prelude</literal> libraries with and without profiling,
3235 so that there is an appropriately-built library archive to link
3236 with when the user compiles his program. It would be possible
3237 to have a completely separate build tree for each such “way”,
3238 but it would be horribly bureaucratic, especially since often
3239 only parts of the build tree need to be constructed in multiple
3243 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3244 contains some clever magic to allow you to build several
3245 versions of a system; and to control locally how many versions
3246 are built and how they differ. This section explains the
3249 <para>The files for a particular way are distinguished by
3250 munging the suffix. The <quote>normal way</quote> is always
3251 built, and its files have the standard suffices
3252 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3253 In addition, you can build one or more extra ways, each
3254 distinguished by a <emphasis>way tag</emphasis>. The object
3255 files and interface files for one of these extra ways are
3256 distinguished by their suffix. For example, way
3257 <literal>mp</literal> has files
3258 <filename>.mp_o</filename> and
3259 <filename>.mp_hi</filename>. Library archives have their
3260 way tag the other side of the dot, for boring reasons; thus,
3261 <filename>libHS_mp.a</filename>.</para>
3263 <para>A <command>make</command> variable called
3264 <constant>way</constant> holds the current way tag.
3265 <emphasis><constant>way</constant> is only ever set on the
3266 command line of <command>gmake</command></emphasis> (usually in
3267 a recursive invocation of <command>gmake</command> by the
3268 system). It is never set inside a
3269 <filename>Makefile</filename>. So it is a global constant for
3270 any one invocation of <command>gmake</command>. Two other
3271 <command>make</command> variables,
3272 <constant>way_</constant> and
3273 <constant>_way</constant> are immediately derived from
3274 <constant>$(way)</constant> and never altered. If
3275 <constant>way</constant> is not set, then neither are
3276 <constant>way_</constant> and
3277 <constant>_way</constant>, and the invocation of
3278 <command>make</command> will build the <quote>normal
3279 way</quote>. If <constant>way</constant> is set, then the other
3280 two variables are set in sympathy. For example, if
3281 <constant>$(way)</constant> is “<literal>mp</literal>”,
3282 then <constant>way_</constant> is set to
3283 “<literal>mp_</literal>” and
3284 <constant>_way</constant> is set to
3285 “<literal>_mp</literal>”. These three variables are
3286 then used when constructing file names.</para>
3288 <para>So how does <command>make</command> ever get recursively
3289 invoked with <constant>way</constant> set? There are two ways
3290 in which this happens:</para>
3294 <para>For some (but not all) of the standard targets, when
3295 in a leaf sub-directory, <command>make</command> is
3296 recursively invoked for each way tag in
3297 <constant>$(WAYS)</constant>. You set
3298 <constant>WAYS</constant> in the
3299 <filename>Makefile</filename> to the list of way tags you
3300 want these targets built for. The mechanism here is very
3301 much like the recursive invocation of
3302 <command>make</command> in sub-directories (<Xref
3303 LinkEnd="sec-subdirs">). It is up to you to set
3304 <constant>WAYS</constant> in your
3305 <filename>Makefile</filename>; this is how you control what
3306 ways will get built.</para>
3310 <para>For a useful collection of targets (such as
3311 <filename>libHS_mp.a</filename>,
3312 <filename>Foo.mp_o</filename>) there is a rule which
3313 recursively invokes <command>make</command> to make the
3314 specified target, setting the <constant>way</constant>
3315 variable. So if you say <command>gmake
3316 Foo.mp_o</command> you should see a recursive
3317 invocation <command>gmake Foo.mp_o way=mp</command>,
3318 and <emphasis>in this recursive invocation the pattern rule
3319 for compiling a Haskell file into a <filename>.o</filename>
3320 file will match</emphasis>. The key pattern rules (in
3321 <filename>suffix.mk</filename>) look like this:
3325 $(HC) $(HC_OPTS) $< -o $@
3332 <para>You can invoke <command>make</command> with a
3333 particular <literal>way</literal> setting yourself, in order
3334 to build files related to a particular
3335 <literal>way</literal> in the current directory. eg.
3341 will build files for the profiling way only in the current
3348 <title>When the canned rule isn't right</title>
3350 <para>Sometimes the canned rule just doesn't do the right thing.
3351 For example, in the <literal>nofib</literal> suite we want the
3352 link step to print out timing information. The thing to do here
3353 is <emphasis>not</emphasis> to define
3354 <constant>HS_PROG</constant> or
3355 <constant>C_PROG</constant>, and instead define a special
3356 purpose rule in your own <filename>Makefile</filename>. By
3357 using different variable names you will avoid the canned rules
3358 being included, and conflicting with yours.</para>
3362 <sect1 id="building-docs">
3363 <title>Building the documentation</title>
3365 <sect2 id="pre-supposed-doc-tools">
3366 <title>Tools for building the Documentation</title>
3368 <para>The following additional tools are required if you want to
3369 format the documentation that comes with the
3370 <literal>fptools</literal> projects:</para>
3374 <term>DocBook</term>
3375 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3376 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3378 <para>Much of our documentation is written in SGML, using
3379 the DocBook DTD. Instructions on installing and
3380 configuring the DocBook tools are below.</para>
3386 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3387 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3389 <para>A decent TeX distribution is required if you want to
3390 produce printable documentation. We recomment teTeX,
3391 which includes just about everything you need.</para>
3396 <term>Haddock</term>
3397 <indexterm><primary>Haddock</primary>
3400 <para>Haddock is a Haskell documentation tool that we use
3401 for automatically generating documentation from the
3402 library source code. It is an <literal>fptools</literal>
3403 project in itself. To build documentation for the
3404 libraries (<literal>fptools/libraries</literal>) you
3405 should check out and build Haddock in
3406 <literal>fptools/haddock</literal>. Haddock requires GHC
3414 <title>Installing the DocBook tools</title>
3417 <title>Installing the DocBook tools on Linux</title>
3419 <para>If you're on a recent RedHat system (7.0+), you probably
3420 have working DocBook tools already installed. The configure
3421 script should detect your setup and you're away.</para>
3423 <para>If you don't have DocBook tools installed, and you are
3424 using a system that can handle RedHat RPM packages, you can
3425 probably use the <ULink
3426 URL="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3427 DocBook tools</ULink>, which is the most shrink-wrapped SGML
3428 suite that we could find. You need all the RPMs except for
3429 psgml (i.e. <Filename>docbook</Filename>,
3430 <Filename>jade</Filename>, <Filename>jadetex</Filename>,
3431 <Filename>sgmlcommon</Filename> and
3432 <Filename>stylesheets</Filename>). Note that most of these
3433 RPMs are architecture neutral, so are likely to be found in a
3434 <Filename>noarch</Filename> directory. The SuSE RPMs also
3435 work; the RedHat ones <Emphasis>don't</Emphasis> in RedHat 6.2
3436 (7.0 and later should be OK), but they are easy to fix: just
3438 <Filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</Filename>
3439 to <Filename>/usr/lib/sgml/lib/dblib.dsl</Filename>. </para>
3443 <title>Installing DocBook on FreeBSD</title>
3445 <para>On FreeBSD systems, the easiest way to get DocBook up
3446 and running is to install it from the ports tree or a
3447 pre-compiled package (packages are available from your local
3448 FreeBSD mirror site).</para>
3450 <para>To use the ports tree, do this:
3452 $ cd /usr/ports/textproc/docproj
3455 This installs the FreeBSD documentation project tools, which
3456 includes everything needed to format the GHC
3457 documentation.</para>
3461 <title>Installing from binaries on Windows</title>
3463 <Para>It's a good idea to use Norman Walsh's <ULink
3464 URL="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3465 notes</ULink> as a guide. You should get version 3.1 of
3466 DocBook, and note that his file <Filename>test.sgm</Filename>
3467 won't work, as it needs version 3.0. You should unpack Jade
3468 into <Filename>\Jade</Filename>, along with the entities,
3469 DocBook into <Filename>\docbook</Filename>, and the DocBook
3470 stylesheets into <Filename>\docbook\stylesheets</Filename> (so
3471 they actually end up in
3472 <Filename>\docbook\stylesheets\docbook</Filename>).</para>
3477 <title>Installing the DocBook tools from source</title>
3482 <para>Install <ULink
3483 URL="http://openjade.sourceforge.net/">OpenJade</ULink>
3484 (Windows binaries are available as well as sources). If you
3485 want DVI, PS, or PDF then install JadeTeX from the
3486 <Filename>dsssl</Filename> subdirectory. (If you get the
3490 ! LaTeX Error: Unknown option implicit=false' for package hyperref'.
3493 your version of <Command>hyperref</Command> is out of date;
3494 download it from CTAN
3495 (<Filename>macros/latex/contrib/supported/hyperref</Filename>),
3496 and make it, ensuring that you have first removed or renamed
3497 your old copy. If you start getting file not found errors
3498 when making the test for <Command>hyperref</Command>, you
3499 can abort at that point and proceed straight to
3500 <Command>make install</Command>, or enter them as
3501 <Filename>../</Filename><Emphasis>filename</Emphasis>.)</para>
3503 <para>Make links from <Filename>virtex</Filename> to
3504 <Filename>jadetex</Filename> and
3505 <Filename>pdfvirtex</Filename> to
3506 <Filename>pdfjadetex</Filename> (otherwise DVI, PostScript
3507 and PDF output will not work). Copy
3508 <Filename>dsssl/*.{dtd,dsl}</Filename> and
3509 <Filename>catalog</Filename> to
3510 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3514 <title>DocBook and the DocBook stylesheets</title>
3516 <para>Get a Zip of <ULink
3517 URL="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ULink>
3518 and install the contents in
3519 <Filename>/usr/[local/]/lib/sgml</Filename>.</para>
3521 <para>Get the <ULink
3522 URL="http://nwalsh.com/docbook/dsssl/">DocBook
3523 stylesheets</ULink> and install in
3524 <Filename>/usr/[local/]lib/sgml/stylesheets</Filename>
3525 (thereby creating a subdirectory docbook). For indexing,
3526 copy or link <Filename>collateindex.pl</Filename> from the
3527 DocBook stylesheets archive in <Filename>bin</Filename> into
3528 a directory on your <Constant>PATH</Constant>.</para>
3530 <para>Download the <ULink
3531 URL="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3532 entities</ULink> into
3533 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3539 <title>Configuring the DocBook tools</title>
3541 <Para>Once the DocBook tools are installed, the configure script
3542 will detect them and set up the build system accordingly. If you
3543 have a system that isn't supported, let us know, and we'll try
3548 <title>Remaining problems</title>
3550 <para>If you install from source, you'll get a pile of warnings
3553 <Screen>DTDDECL catalog entries are not supported</Screen>
3555 every time you build anything. These can safely be ignored, but
3556 if you find them tedious you can get rid of them by removing all
3557 the <Constant>DTDDECL</Constant> entries from
3558 <Filename>docbook.cat</Filename>.</para>
3562 <title>Building the documentation</title>
3564 <para>To build documentation in a certain format, you can
3565 say, for example,</para>
3571 <para>to build HTML documentation below the current directory.
3572 The available formats are: <literal>dvi</literal>,
3573 <literal>ps</literal>, <literal>pdf</literal>,
3574 <literal>html</literal>, and <literal>rtf</literal>. Note that
3575 not all documentation can be built in all of these formats: HTML
3576 documentation is generally supported everywhere, and DocBook
3577 documentation might support the other formats (depending on what
3578 other tools you have installed).</para>
3580 <para>All of these targets are recursive; that is, saying
3581 <literal>make html</literal> will make HTML docs for all the
3582 documents recursively below the current directory.</para>
3584 <para>Because there are many different formats that the DocBook
3585 documentation can be generated in, you have to select which ones
3586 you want by setting the <literal>SGMLDocWays</literal> variable
3587 to a list of them. For example, in
3588 <filename>build.mk</filename> you might have a line:</para>
3591 SGMLDocWays = html ps
3594 <para>This will cause the documentation to be built in the requested
3595 formats as part of the main build (the default is not to build
3596 any documentation at all).</para>
3600 <title>Installing the documentation</title>
3602 <para>To install the documentation, use:</para>
3608 <para>This will install the documentation into
3609 <literal>$(datadir)</literal> (which defaults to
3610 <literal>$(prefix)/share</literal>). The exception is HTML
3611 documentation, which goes into
3612 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3614 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3615 to a list of formats, the <literal>install-docs</literal> target
3616 won't do anything for SGML documentation.</para>
3622 <sect1 id="sec-porting-ghc">
3623 <title>Porting GHC</title>
3625 <para>This section describes how to port GHC to a currenly
3626 unsupported platform. There are two distinct
3627 possibilities:</para>
3631 <para>The hardware architecture for your system is already
3632 supported by GHC, but you're running an OS that isn't
3633 supported (or perhaps has been supported in the past, but
3634 currently isn't). This is the easiest type of porting job,
3635 but it still requires some careful bootstrapping. Proceed to
3636 <xref linkend="sec-booting-from-hc">.</para>
3640 <para>Your system's hardware architecture isn't supported by
3641 GHC. This will be a more difficult port (though by comparison
3642 perhaps not as difficult as porting gcc). Proceed to <xref
3643 linkend="unregisterised-porting">.</para>
3647 <sect2 id="sec-booting-from-hc">
3648 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3650 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3651 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3652 <indexterm><primary>porting GHC</primary></indexterm>
3654 <para>Bootstrapping GHC on a system without GHC already
3655 installed is achieved by taking the intermediate C files (known
3656 as HC files) from a GHC compilation on a supported system to the
3657 target machine, and compiling them using gcc to get a working
3660 <para><emphasis>NOTE: GHC versions 5.xx were hard to bootstrap
3661 from C. We recommend using GHC 6.0.1 or
3662 later.</emphasis></para>
3664 <para>HC files are platform-dependent, so you have to get a set
3665 that were generated on similar hardware. There may be some
3666 supplied on the GHC download page, otherwise you'll have to
3667 compile some up yourself, or start from
3668 <emphasis>unregisterised</emphasis> HC files - see <xref
3669 linkend="unregisterised-porting">.</para>
3671 <para>The following steps should result in a working GHC build
3672 with full libraries:</para>
3676 <para>Unpack the HC files on top of a fresh source tree
3677 (make sure the source tree version matches the version of
3678 the HC files <emphasis>exactly</emphasis>!). This will
3679 place matching <filename>.hc</filename> files next to the
3680 corresponding Haskell source (<filename>.hs</filename> or
3681 <filename>.lhs</filename>) in the compiler subdirectory
3682 <filename>ghc/compiler</filename> and in the libraries
3683 (subdirectories of <filename>hslibs</filename> and
3684 <literal>libraries</literal>).</para>
3688 <para>The actual build process is fully automated by the
3689 <filename>hc-build</filename> script located in the
3690 <filename>distrib</filename> directory. If you eventually
3691 want to install GHC into the directory
3692 <replaceable>dir</replaceable>, the following
3693 command will execute the whole build process (it won't
3694 install yet):</para>
3697 foo% distrib/hc-build --prefix=<replaceable>dir</replaceable>
3699 <indexterm><primary>--hc-build</primary></indexterm>
3701 <para>By default, the installation directory is
3702 <filename>/usr/local</filename>. If that is what you want,
3703 you may omit the argument to <filename>hc-build</filename>.
3704 Generally, any option given to <filename>hc-build</filename>
3705 is passed through to the configuration script
3706 <filename>configure</filename>. If
3707 <filename>hc-build</filename> successfully completes the
3708 build process, you can install the resulting system, as
3718 <sect2 id="unregisterised-porting">
3719 <title>Porting GHC to a new architecture</title>
3721 <para>The first step in porting to a new architecture is to get
3722 an <firstterm>unregisterised</firstterm> build working. An
3723 unregisterised build is one that compiles via vanilla C only.
3724 By contrast, a registerised build uses the following
3725 architecture-specific hacks for speed:</para>
3729 <para>Global register variables: certain abstract machine
3730 <quote>registers</quote> are mapped to real machine
3731 registers, depending on how many machine registers are
3733 <filename>ghc/includes/MachRegs.h</filename>).</para>
3737 <para>Assembly-mangling: when compiling via C, we feed the
3738 assembly generated by gcc though a Perl script known as the
3739 <firstterm>mangler</firstterm> (see
3740 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3741 mangler rearranges the assembly to support tail-calls and
3742 various other optimisations.</para>
3746 <para>In an unregisterised build, neither of these hacks are
3747 used — the idea is that the C code generated by the
3748 compiler should compile using gcc only. The lack of these
3749 optimisations costs about a factor of two in performance, but
3750 since unregisterised compilation is usually just a step on the
3751 way to a full registerised port, we don't mind too much.</para>
3753 <para>Notes on GHC portability in general: we've tried to stick
3754 to writing portable code in most parts of the system, so it
3755 should compile on any POSIXish system with gcc, but in our
3756 experience most systems differ from the standards in one way or
3757 another. Deal with any problems as they arise - if you get
3758 stuck, ask the experts on
3759 <email>glasgow-haskell-users@haskell.org</email>.</para>
3761 <para>Lots of useful information about the innards of GHC is
3762 available in the <ulink
3763 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3764 Commentary</ulink>, which might be helpful if you run into some
3765 code which needs tweaking for your system.</para>
3768 <title>Cross-compiling to produce an unregisterised GHC</title>
3770 <para>In this section, we explain how to bootstrap GHC on a
3771 new platform, using unregisterised intermediate C files. We
3772 haven't put a great deal of effort into automating this
3773 process, for two reasons: it is done very rarely, and the
3774 process usually requires human intervention to cope with minor
3775 porting issues anyway.</para>
3777 <para>The following step-by-step instructions should result in
3778 a fully working, albeit unregisterised, GHC. Firstly, you
3779 need a machine that already has a working GHC (we'll call this
3780 the <firstterm>host</firstterm> machine), in order to
3781 cross-compile the intermediate C files that we will use to
3782 bootstrap the compiler on the <firstterm>target</firstterm>
3787 <para>On the target machine:</para>
3791 <para>Unpack a source tree (preferably a released
3792 version). We will call the path to the root of this
3793 tree <replaceable>T</replaceable>.</para>
3798 $ cd <replaceable>T</replaceable>
3799 $ ./configure --enable-hc-boot --enable-hc-boot-unregisterised
3802 <para>You might need to update
3803 <filename>configure.in</filename> to recognise the new
3804 architecture, and re-generate
3805 <filename>configure</filename> with
3806 <literal>autoreconf</literal>.</para>
3811 $ cd <replaceable>T</replaceable>/ghc/includes
3819 <para>On the host machine:</para>
3823 <para>Unpack a source tree (same released version). Call
3824 this directory <replaceable>H</replaceable>.</para>
3829 $ cd <replaceable>H</replaceable>
3836 <filename><replaceable>H</replaceable>/mk/build.mk</filename>,
3837 with the following contents:</para>
3840 GhcUnregisterised = YES
3841 GhcLibHcOpts = -O -H32m -keep-hc-files
3844 GhcWithNativeCodeGen = NO
3845 GhcWithInterpreter = NO
3846 GhcStage1HcOpts = -O -H32m -fasm
3847 GhcStage2HcOpts = -O -fvia-C -keep-hc-files
3853 <filename><replaceable>H</replaceable>/mk/config.mk</filename>:</para>
3856 <para>change <literal>TARGETPLATFORM</literal>
3857 appropriately, and set the variables involving
3858 <literal>TARGET</literal> to the correct values for
3859 the target platform. This step is necessary because
3860 currently <literal>configure</literal> doesn't cope
3861 with specifying different values for the
3862 <literal>--host</literal> and
3863 <literal>--target</literal> flags.</para>
3866 <para>copy <literal>LeadingUnderscore</literal>
3867 setting from target.</para>
3874 <filename><replaceable>T</replaceable>/ghc/includes/config.h</filename>
3876 <filename><replaceable>H</replaceable>/ghc/includes</filename>.
3877 Note that we are building on the host machine, using the
3878 target machine's <literal>config.h</literal> file. This
3879 is so that the intermediate C files generated here will
3880 be suitable for compiling on the target system.</para>
3885 <para>Touch <literal>config.h</literal>, just to make
3886 sure it doesn't get replaced during the build:</para>
3888 $ touch <replaceable>H</replaceable>/ghc/includes/config.h</screen>
3892 <para>Now build the compiler:</para>
3894 $ cd <replaceable>H</replaceable>/glafp-utils && make boot && make
3895 $ cd <replaceable>H</replaceable>/ghc && make boot && make
3897 <para>Don't worry if the build falls over in the RTS, we
3898 don't need the RTS yet.</para>
3903 $ cd <replaceable>H</replaceable>/libraries
3904 $& make boot && make
3910 $ cd <replaceable>H</replaceable>/ghc
3911 $ make boot stage=2 && make stage=2
3917 $ cd <replaceable>H</replaceable>/ghc/utils
3919 $ make -k HC=<replaceable>H</replaceable>/ghc/compiler/stage1/ghc-inplace \
3920 EXTRA_HC_OPTS='-O -fvia-C -keep-hc-files'
3926 $ cd <replaceable>H</replaceable>
3927 $ make hc-file-bundle Project=Ghc
3933 <filename><replaceable>H</replaceable>/*-hc.tar.gz</filename>
3934 to <filename><replaceable>T</replaceable>/..</filename>.</para>
3940 <para>On the target machine:</para>
3942 <para>At this stage we simply need to bootstrap a compiler
3943 from the intermediate C files we generated above. The
3944 process of bootstrapping from C files is automated by the
3945 script in <literal>distrib/hc-build</literal>, and is
3946 described in <xref linkend="sec-booting-from-hc">.</para>
3949 $ ./distrib/hc-build --enable-hc-boot-unregisterised
3952 <para>However, since this is a bootstrap on a new machine,
3953 the automated process might not run to completion the
3954 first time. For that reason, you might want to treat the
3955 <literal>hc-build</literal> script as a list of
3956 instructions to follow, rather than as a fully automated
3957 script. This way you'll be able to restart the process
3958 part-way through if you need to fix anything on the
3961 <para>Don't bother with running
3962 <literal>make install</literal> in the newly
3963 bootstrapped tree; just use the compiler in that tree to
3964 build a fresh compiler from scratch, this time without
3965 booting from C files. Before doing this, you might want
3966 to check that the bootstrapped compiler is generating
3967 working binaries:</para>
3971 main = putStrLn "Hello World!\n"
3973 $ <replaceable>T</replaceable>/ghc/compiler/ghc-inplace hello.hs -o hello
3978 <para>Once you have the unregisterised compiler up and
3979 running, you can use it to start a registerised port. The
3980 following sections describe the various parts of the
3981 system that will need architecture-specific tweaks in
3982 order to get a registerised build going.</para>
3989 <title>Porting the RTS</title>
3991 <para>The following files need architecture-specific code for a
3992 registerised build:</para>
3996 <term><filename>ghc/includes/MachRegs.h</filename></term>
3997 <indexterm><primary><filename>MachRegs.h</filename></primary>
4000 <para>Defines the STG-register to machine-register
4001 mapping. You need to know your platform's C calling
4002 convention, and which registers are generally available
4003 for mapping to global register variables. There are
4004 plenty of useful comments in this file.</para>
4008 <term><filename>ghc/includes/TailCalls.h</filename></term>
4009 <indexterm><primary><filename>TailCalls.h</filename></primary>
4012 <para>Macros that cooperate with the mangler (see <xref
4013 linkend="sec-mangler">) to make proper tail-calls
4018 <term><filename>ghc/rts/Adjustor.c</filename></term>
4019 <indexterm><primary><filename>Adjustor.c</filename></primary>
4023 <literal>foreign import "wrapper"</literal>
4025 <literal>foreign export dynamic</literal>).
4026 Not essential for getting GHC bootstrapped, so this file
4027 can be deferred until later if necessary.</para>
4031 <term><filename>ghc/rts/StgCRun.c</filename></term>
4032 <indexterm><primary><filename>StgCRun.c</filename></primary>
4035 <para>The little assembly layer between the C world and
4036 the Haskell world. See the comments and code for the
4037 other architectures in this file for pointers.</para>
4041 <term><filename>ghc/rts/MBlock.h</filename></term>
4042 <term><filename>ghc/rts/MBlock.c</filename></term>
4043 <indexterm><primary><filename>MBlock.h</filename></primary>
4045 <indexterm><primary><filename>MBlock.c</filename></primary>
4048 <para>These files are really OS-specific rather than
4049 architecture-specific. In <filename>MBlock.h</filename>
4050 is specified the absolute location at which the RTS
4051 should try to allocate memory on your platform (try to
4052 find an area which doesn't conflict with code or dynamic
4053 libraries). In <filename>Mblock.c</filename> you might
4054 need to tweak the call to <literal>mmap()</literal> for
4061 <sect3 id="sec-mangler">
4062 <title>The mangler</title>
4064 <para>The mangler is an evil Perl-script that rearranges the
4065 assembly code output from gcc to do two main things:</para>
4069 <para>Remove function prologues and epilogues, and all
4070 movement of the C stack pointer. This is to support
4071 tail-calls: every code block in Haskell code ends in an
4072 explicit jump, so we don't want the C-stack overflowing
4073 while we're jumping around between code blocks.</para>
4076 <para>Move the <firstterm>info table</firstterm> for a
4077 closure next to the entry code for that closure. In
4078 unregisterised code, info tables contain a pointer to the
4079 entry code, but in registerised compilation we arrange
4080 that the info table is shoved right up against the entry
4081 code, and addressed backwards from the entry code pointer
4082 (this saves a word in the info table and an extra
4083 indirection when jumping to the closure entry
4088 <para>The mangler is abstracted to a certain extent over some
4089 architecture-specific things such as the particular assembler
4090 directives used to herald symbols. Take a look at the
4091 definitions for other architectures and use these as a
4092 starting point.</para>
4096 <title>The native code generator</title>
4098 <para>The native code generator isn't essential to getting a
4099 registerised build going, but it's a desirable thing to have
4100 because it can cut compilation times in half. The native code
4101 generator is described in some detail in the <ulink
4102 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
4103 commentary</ulink>.</para>
4109 <para>To support GHCi, you need to port the dynamic linker
4110 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
4111 currently supports the ELF and PEi386 object file formats - if
4112 your platform uses one of these then things will be
4113 significantly easier. The majority of Unix platforms use the
4114 ELF format these days. Even so, there are some
4115 machine-specific parts of the ELF linker: for example, the
4116 code for resolving particular relocation types is
4117 machine-specific, so some porting of this code to your
4118 architecture will probaly be necessary.</para>
4120 <para>If your system uses a different object file format, then
4121 you have to write a linker — good luck!</para>
4127 <sect1 id="sec-build-pitfalls">
4128 <title>Known pitfalls in building Glasgow Haskell
4130 <indexterm><primary>problems, building</primary></indexterm>
4131 <indexterm><primary>pitfalls, in building</primary></indexterm>
4132 <indexterm><primary>building pitfalls</primary></indexterm></title>
4135 WARNINGS about pitfalls and known “problems”:
4144 One difficulty that comes up from time to time is running out of space
4145 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
4146 compensate for the vagaries of different sysadmin approaches to temp
4148 <indexterm><primary>tmp, running out of space in</primary></indexterm>
4150 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
4151 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
4154 The best way around it is to say
4157 export TMPDIR=<dir>
4160 in your <filename>build.mk</filename> file.
4161 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
4170 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
4171 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
4179 When compiling via C, you'll sometimes get “warning: assignment from
4180 incompatible pointer type” out of GCC. Harmless.
4187 Similarly, <command>ar</command>chiving warning messages like the following are not
4191 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
4192 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
4202 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
4203 get an “Out of heap space” error message. These can vary with the
4204 vagaries of different systems, it seems. The solution is simple:
4211 If you're compiling with GHC 4.00 or later, then the
4212 <emphasis>maximum</emphasis> heap size must have been reached. This
4213 is somewhat unlikely, since the maximum is set to 64M by default.
4214 Anyway, you can raise it with the
4215 <option>-optCrts-M<size></option> flag (add this flag to
4216 <constant><module>_HC_OPTS</constant>
4217 <command>make</command> variable in the appropriate
4218 <filename>Makefile</filename>).
4225 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
4234 and try again: <command>gmake</command>. (see <Xref LinkEnd="sec-suffix"> for information about
4235 <constant><module>_HC_OPTS</constant>.)
4237 Alternatively, just cut to the chase:
4241 % make EXTRA_HC_OPTS=-optCrts-M128M
4250 If you try to compile some Haskell, and you get errors from GCC about
4251 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4252 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4254 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4255 this bug also suggests that you have an old GCC.
4263 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4267 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4268 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4270 ? # or, on some machines: ar s $i
4275 We'd be interested to know if this is still necessary.
4283 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4284 a bit from one Unix to another. One particular gotcha is macro calls
4289 SLIT("Hello, world")
4293 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4294 arguments, so you get
4298 :731: macro `SLIT' used with too many (2) args
4302 Alas, <command>cpp</command> doesn't tell you the offending file!
4304 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4315 <Sect1 id="platforms"><Title>Platforms, scripts, and file names</Title>
4317 GHC is designed both to be built, and to run, on both Unix and Windows. This flexibility
4318 gives rise to a good deal of brain-bending detail, which we have tried to collect in this chapter.
4321 <sect2 id="cygwin-and-mingw"><Title>Windows platforms: Cygwin, MSYS, and MinGW</Title>
4323 <para> The build system is built around Unix-y makefiles. Because it's not native,
4324 the Windows situation for building GHC is particularly confusing. This section
4325 tries to clarify, and to establish terminology.</para>
4327 <sect3 id="ghc-mingw"><title>MinGW</title>
4329 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4330 is a collection of header
4331 files and import libraries that allow one to use <command>gcc</command> and produce
4332 native Win32 programs that do not rely on any third-party DLLs. The
4333 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4334 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4338 <para> The down-side of MinGW is that the MinGW libraries do not support anything like the full
4343 <sect3 id="ghc-cygwin"><title>Cygwin and MSYS</title>
4345 <para>You can't use the MinGW to <emphasis>build</emphasis> GHC, because MinGW doesn't have a shell,
4346 or the standard Unix commands such as <command>mv</command>, <command>rm</command>,
4347 <command>ls</command>, nor build-system stuff such as <command>make</command> and <command>cvs</command>.
4348 For that, there are two choices: <ulink url="http://www.cygwin.com">Cygwin</ulink>
4349 and <ulink url="http://www.mingw.org/msys.shtml">MSYS</ulink>:
4353 Cygwin comes with compilation tools (<command>gcc</command>, <command>ld</command> and so on), which
4354 compile code that has access to all of Posix. The price is that the executables must be
4355 dynamically linked with the Cygwin DLL, so that <emphasis>you cannot run a Cywin-compiled program on a machine
4356 that doesn't have Cygwin</emphasis>. Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4357 does not change, but the implementation certainly does. Even the interfaces to functions
4358 it exports seem to change occasionally. </para>
4362 MSYS is a fork of the Cygwin tree, so they
4363 are fundamentally similar. However, MSYS is by design much smaller and simpler. Access to the file system goes
4364 through fewer layers, so MSYS is quite a bit faster too.
4367 <para>Furthermore, MSYS provides no compilation tools; it relies instead on the MinGW tools. These
4368 compile binaries that run with no DLL support, on any Win32 system.
4369 However, MSYS does come with all the make-system tools, such as <command>make</command>, <command>autoconf</command>,
4370 <command>cvs</command>, <command>ssh</command> etc. To get these, you have to download the
4371 MsysDTK (Developer Tool Kit) package, as well as the base MSYS package.
4373 <para>MSYS does have a DLL, but it's only used by MSYS commands (<command>sh</command>, <command>rm</command>,
4374 <command>ssh</command> and so on),
4375 not by programs compiled under MSYS.
4383 <sect3><title>Targeting MinGW</title>
4385 <para>We want GHC to compile programs that work on any Win32 system. Hence:
4388 GHC does invoke a C compiler, assembler, linker and so on, but we ensure that it only
4389 invokes the MinGW tools, not the Cygwin ones. That means that the programs GHC compiles
4390 will work on any system, but it also means that the programs GHC compiles do not have access
4391 to all of Posix. In particular, they cannot import the (Haskell) Posix
4392 library; they have to do
4393 their input output using standard Haskell I/O libraries, or native Win32 bindings.</para>
4394 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4398 To make the GHC distribution self-contained, the GHC distribution includes the MinGW <command>gcc</command>,
4399 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4402 So <emphasis>GHC targets MinGW</emphasis>, not Cygwin.
4403 It is in principle possible to build a version of GHC, <emphasis>GHC-cygwin</emphasis>,
4404 that targets Cygwin instead. The up-side of GHC-cygwin is
4405 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4406 <emphasis>We do not support GHC-cygwin, however; it is beyond our resources.</emphasis>
4409 <para>While GHC <emphasis>targets</emphasis> MinGW, that says nothing about
4410 how GHC is <emphasis>built</emphasis>. We use both MSYS and Cygwin as build environments for
4411 GHC; both work fine, though MSYS is rather lighter weight.</para>
4413 <para>In your build tree, you build a compiler called <Command>ghc-inplace</Command>. It
4414 uses the <Command>gcc</Command> that you specify using the
4415 <option>--with-gcc</option> flag when you run
4416 <Command>configure</Command> (see below).
4417 The makefiles are careful to use <Command>ghc-inplace</Command> (not <Command>gcc</Command>)
4418 to compile any C files, so that it will in turn invoke the correct <Command>gcc</Command> rather that
4419 whatever one happens to be in your path. However, the makefiles do use whatever <Command>ld</Command>
4420 and <Command>ar</Command> happen to be in your path. This is a bit naughty, but (a) they are only
4421 used to glom together .o files into a bigger .o file, or a .a file,
4422 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4423 Cygwin and MinGW use the same .o file format. So its ok.
4427 <sect3><title> File names </title>
4429 <para>Cygwin, MSYS, and the underlying Windows file system all understand file paths of form <literal>c:/tmp/foo</literal>.
4433 MSYS programs understand <filename>/bin</filename>, <filename>/usr/bin</filename>, and map Windows's lettered drives as
4434 <filename>/c/tmp/foo</filename> etc. The exact mount table is given in the doc subdirectory of the MSYS distribution.
4436 <para> When it invokes a command, the MSYS shell sees whether the invoked binary lives in the MSYS <filename>/bin</filename>
4437 directory. If so, it just invokes it. If not, it assumes the program is no an MSYS program, and walks over the command-line
4438 arguments changing MSYS paths into native-compatible paths. It does this inside sub-arguments and inside quotes. For example,
4443 the MSYS shell will actually call <literal>foogle</literal> with argument <literal>-Bc:/tmp/baz</literal>.
4447 Cygwin programs have a more complicated mount table, and map the lettered drives as <filename>/cygdrive/c/tmp/foo</filename>.
4449 <para>The Cygwin shell does no argument processing when invoking non-Cygwin programs.
4455 <sect3><title>HOST_OS vs TARGET_OS</title>
4458 In the source code you'll find various ifdefs looking like:
4460 #ifdef mingw32_HOST_OS
4466 #ifdef mingw32_TARGET_OS
4470 These macros are set by the configure script (via the file config.h).
4471 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4474 The "host" system is the one on which GHC itself will be run.
4477 The "target" system is the one for which the program compiled by GHC will be run.
4480 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4481 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4488 <sect2><title>Wrapper scripts</title>
4491 Many programs, including GHC itself and hsc2hs, need to find associated binaries and libraries.
4492 For <emphasis>installed</emphasis> programs, the strategy depends on the platform. We'll use
4493 GHC itself as an example:
4496 On Unix, the command <command>ghc</command> is a shell script, generated by adding installation
4497 paths to the front of the source file <filename>ghc.sh</filename>,
4498 that invokes the real binary, passing "-B<emphasis>path</emphasis>" as an argument to tell <command>ghc</command>
4499 where to find its supporting files.
4503 On vanilla Windows, it turns out to be much harder to make reliable script to be run by the
4504 native Windows shell <command>cmd</command> (e.g. limits on the length
4505 of the command line). So instead we invoke the GHC binary directly, with no -B flag.
4506 GHC uses the Windows <literal>getExecDir</literal> function to find where the executable is,
4507 and from that figures out where the supporting files are.
4510 (You can find the layout of GHC's supporting files in the
4511 section "Layout of installed files" of Section 2 of the GHC user guide.)
4514 Things work differently for <emphasis>in-place</emphasis> execution, where you want to
4515 execute a program that has just been built in a build tree. The difference is that the
4516 layout of the supporting files is different.
4517 In this case, whether on Windows or Unix, we always use a shell script. This works OK
4518 on Windows because the script is executed by MSYS or Cygwin, which don't have the
4519 shortcomings of the native Windows <command>cmd</command> shell.
4526 <Sect1 id="winbuild"><Title>Instructions for building under Windows</Title>
4529 This section gives detailed instructions for how to build
4530 GHC from source on your Windows machine. Similar instructions for
4531 installing and running GHC may be found in the user guide. In general,
4532 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4535 Make sure you read the preceding section on platforms (<xref linkend="platforms">)
4536 before reading section.
4537 You don't need Cygwin or MSYS to <emphasis>use</emphasis> GHC,
4538 but you do need one or the other to <emphasis>build</emphasis> GHC.</para>
4541 <Sect2 id=msys-install><Title>Installing and configuring MSYS</Title>
4544 MSYS is a lightweight alternative to Cygwin.
4545 You don't need MSYS to <emphasis>use</emphasis> GHC,
4546 but you do need it or Cygwin to <emphasis>build</emphasis> GHC.
4547 Here's how to install MSYS.
4550 Go to <ulink url="http://www.mingw.org/download.shtml">http://www.mingw.org/download.shtml</ulink> and
4551 download the following (of course, the version numbers will differ):
4553 <listitem><para>The main MSYS package (binary is sufficient): <literal>MSYS-1.0.9.exe</literal>
4555 <listitem><para>The MSYS developer's toolkit (binary is sufficient): <literal>msysDTK-1.0.1.exe</literal>.
4556 This provides <command>make</command>, <command>autoconf</command>,
4557 <command>ssh</command>, <command>cvs</command> and probably more besides.
4560 Run both executables (in the order given above) to install them. I put them in <literal>c:/msys</literal>
4564 Set the following environment variables
4566 <listitem><para><literal>PATH</literal>: add <literal>c:/msys/1.0/bin</literal> to your path. (Of course, the version number may differ.)
4569 <listitem><para><literal>HOME</literal>: set to your home directory (e.g. <literal>c:/userid</literal>).
4570 This is where, among other things, <command>ssh</command> will look for your <literal>.ssh</literal> directory.
4573 <listitem><para><literal>SHELL</literal>: set to <literal>c:/msys/1.0/bin/sh.exe</literal>
4576 <listitem><para><literal>CVS_RSH</literal>: set to <literal>c:/msys/1.0/bin/ssh.exe</literal>. Only necessary if
4580 <listitem><para><literal>MAKE_MODE</literal>: set to <literal>UNIX</literal>. (I'm not certain this is necessary for MSYS.)
4587 Check that the <literal>CYGWIN</literal> environment variable is <emphasis>not</emphasis> set. It's a bad bug
4588 that MSYS is affected by this, but if you have CYGWIN set to "ntsec ntea", which is right for Cygwin, it
4589 causes the MSYS <command>ssh</command> to bogusly fail complaining that your <filename>.ssh/identity</filename>
4590 file has too-liberal permissinos.
4595 <para>Here are some points to bear in mind when using MSYS:
4597 <listitem> <para> MSYS does some kind of special magic to binaries stored in
4598 <filename>/bin</filename> and <filename>/usr/bin</filename>, which are by default both mapped
4599 to <filename>c:/msys/1.0/bin</filename> (assuming you installed MSYS in <filename>c:/msys</filename>).
4600 Do not put any other binaries (such as GHC or Alex) in this directory or its sub-directories:
4601 they fail in mysterious ways. However, it's fine to put other binaries in <filename>/usr/local/bin</filename>,
4602 which maps to <filename>c:/msys/1.0/local/bin</filename>.</para></listitem>
4604 <listitem> <para> MSYS seems to implement symbolic links by copying, so sharing is lost.
4608 Win32 has a <command>find</command> command which is not the same as MSYS's find.
4609 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4610 before the MSYS one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4611 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4612 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4616 MSYS comes with <command>bzip</command>, and MSYS's <command>tar</command>'s <literal>-j</literal>
4617 will bunzip an archive (e.g. <literal>tar xvjf foo.tar.bz2</literal>). Useful when you get a
4618 bzip'd dump.</para></listitem>
4624 <Sect2><Title>Installing and configuring Cygwin</Title>
4626 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4627 The installation process is straightforward; we install it in <Filename>c:/cygwin</Filename>.
4628 During the installation dialogue, make sure that you select all of the following:
4629 <command>cvs</command>,
4630 <command>openssh</command>,
4631 <command>autoconf</command>,
4632 <command>binutils</command> (includes ld and (I think) ar),
4633 <command>gcc</command>,
4634 <command>flex</command>,
4635 <command>make</command>.
4636 If you miss out any of these, strange things will happen to you. To see thse packages,
4637 click on the "View" button in the "Select Packages"
4638 stage of Cygwin's installation dialogue, until the view says "Full". The default view, which is
4639 "Category" isn't very helpful, and the "View" button is rather unobtrousive.
4641 <para> Now set the following user environment variables:
4644 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4645 <constant>PATH</constant></para></listitem>
4649 Set <constant>MAKE_MODE</constant> to <Literal>UNIX</Literal>. If you
4650 don't do this you get very weird messages when you type
4651 <Command>make</Command>, such as:
4653 /c: /c: No such file or directory
4658 <listitem><para> Set <constant>SHELL</constant> to
4659 <Filename>c:/cygwin/bin/bash</Filename>. When you invoke a shell in Emacs, this
4660 <constant>SHELL</constant> is what you get.
4663 <listitem><para> Set <constant>HOME</constant> to point to your
4664 home directory. This is where, for example,
4665 <command>bash</command> will look for your <filename>.bashrc</filename>
4666 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4672 There are a few other things to do:
4676 By default, cygwin provides the command shell <filename>ash</filename>
4677 as <filename>sh.exe</filename>. We have often seen build-system problems that
4678 turn out to be due to bugs in <filename>ash</filename>
4680 and length of command lines). On the other hand <filename>bash</filename> seems
4682 So, in <filename>cygwin/bin</filename>
4683 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4684 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4685 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4686 you can't rename a running program!
4692 Some script files used in the make system start with "<Command>#!/bin/perl</Command>",
4693 (and similarly for <Command>sh</Command>). Notice the hardwired path!
4694 So you need to ensure that your <Filename>/bin</Filename> directory has the following
4697 <listitem> <para><Command>sh</Command></para></listitem>
4698 <listitem> <para><Command>perl</Command></para></listitem>
4699 <listitem> <para><Command>cat</Command></para></listitem>
4701 All these come in Cygwin's <Filename>bin</Filename> directory, which you probably have
4702 installed as <Filename>c:/cygwin/bin</Filename>. By default Cygwin mounts "<Filename>/</Filename>" as
4703 <Filename>c:/cygwin</Filename>, so if you just take the defaults it'll all work ok.
4704 (You can discover where your Cygwin
4705 root directory <Filename>/</Filename> is by typing <Command>mount</Command>.)
4706 Provided <Filename>/bin</Filename> points to the Cygwin <Filename>bin</Filename>
4707 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4708 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4714 <para>Finally, here are some things to be aware of when using Cygwin:
4716 <listitem> <para>Cygwin doesn't deal well with filenames that include
4717 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4721 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4722 magical text in it. So other programs that don't use Cygwin's
4723 I/O libraries won't recognise such files as symlinks.
4724 In particular, programs compiled by GHC are meant to be runnable
4725 without having Cygwin, so they don't use the Cygwin library, so
4726 they don't recognise symlinks.
4730 See the notes in <xref linkend="msys-install"> about <command>find</command> and <command>bzip</command>,
4731 which apply to Cygwin too.
4739 <Sect2 id="configure-ssh"><Title>Configuring SSH</Title>
4741 <para><command>ssh</command> comes with Cygwin, provided you remember to ask for it when
4742 you install Cygwin. (If not, the installer lets you update easily.) Look for <command>openssh</command>
4743 (not ssh) in the Cygwin list of applications!</para>
4745 <para>There are several strange things about <command>ssh</command> on Windows that you need to know.
4749 The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
4750 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
4751 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
4752 and run it as follows:
4754 c:\tmp> set CYGWIN32=tty
4755 c:\tmp> c:/user/local/bin/ssh-keygen1
4759 <listitem><para> (Cygwin-only problem, I think.)
4760 <command>ssh</command> needs to access your directory <filename>.ssh</filename>, in your home directory.
4761 To determine your home directory <command>ssh</command> first looks in
4762 <filename>c:/cygwin/etc/passwd</filename> (or wherever you have Cygwin installed). If there's an entry
4763 there with your userid, it'll use that entry to determine your home directory, <emphasis>ignoring
4764 the setting of the environment variable $HOME</emphasis>. If the home directory is
4765 bogus, <command>ssh</command> fails horribly. The best way to see what is going on is to say
4767 ssh -v cvs.haskell.org
4769 which makes <command>ssh</command> print out information about its activity.
4771 <para> You can fix this problem, either by correcting the home-directory field in
4772 <filename>c:/cygwin/etc/passwd</filename>, or by simply deleting the entire entry for your userid. If
4773 you do that, <command>ssh</command> uses the $HOME environment variable instead.
4779 <para>To protect your
4780 <literal>.ssh</literal> from access by anyone else,
4781 right-click your <literal>.ssh</literal> directory, and
4782 select <literal>Properties</literal>. If you are not on
4783 the access control list, add yourself, and give yourself
4784 full permissions (the second panel). Remove everyone else
4785 from the access control list. Don't leave them there but
4786 deny them access, because 'they' may be a list that
4787 includes you!</para>
4791 <para>In fact <command>ssh</command> 3.6.1 now seems to <emphasis>require</emphasis>
4792 you to have Unix permissions 600 (read/write for owner only)
4793 on the <literal>.ssh/identity</literal> file, else it
4794 bombs out. For your local C drive, it seems that <literal>chmod 600 identity</literal> works,
4795 but on Windows NT/XP, it doesn't work on a network drive (exact dteails obscure).
4796 The solution seems to be to set the $CYGWIN environment
4797 variable to "<literal>ntsec neta</literal>". The $CYGWIN environment variable is discussed
4798 in <ulink url="http://cygwin.com/cygwin-ug-net/using-cygwinenv.html">the Cygwin User's Guide</ulink>,
4799 and there are more details in <ulink url="http://cygwin.com/faq/faq_4.html#SEC44">the Cygwin FAQ</ulink>.
4806 <Sect2><Title>Other things you need to install</Title>
4808 <para>You have to install the following other things to build GHC, listed below.</para>
4810 <para>On Windows you often install executables in directories with spaces, such as
4811 "<filename>Program Files</filename>". However, the <literal>make</literal> system for fptools doesn't
4812 deal with this situation (it'd have to do more quoting of binaries), so you are strongly advised
4813 to put binaries for all tools in places with no spaces in their path.
4814 On both MSYS and Cygwin, it's perfectly OK to install such programs in the standard Unixy places,
4815 <filename>/usr/local/bin</filename> and <filename>/usr/local/lib</filename>. But it doesn't matter,
4816 provided they are in your path.
4820 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4821 This is what you will use to compile GHC. Add it in your
4822 <constant>PATH</constant>: the installer tells you the path element
4823 you need to add upon completion.
4829 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4830 Happy is a parser generator used to compile the Haskell grammar. Under MSYS or Cygwin you can easily
4831 build it from the source distribution using
4837 This should install it in <filename>/usr/local/bin</filename> (which maps to <filename>c:/msys/1.0/local/bin</filename>
4839 Make sure the installation directory is in your
4840 <constant>PATH</constant>.
4845 <para>Install Alex. This can be done by building from the
4846 source distribution in the same way as Happy. Sources are
4847 available from <ulink
4848 url="http://www.haskell.org/alex">http://www.haskell.org/alex</ulink>.</para>
4852 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4853 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw">).
4854 Just pick up a mingw bundle at
4855 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4856 We install it in <filename>c:/mingw</filename>.
4858 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4859 They are only going to get used by explicit access (via the --with-gcc flag you
4860 give to <Command>configure</Command> later). If you do add them to your path
4861 you are likely to get into a mess because their names overlap with Cygwin binaries.
4867 <para>We use <command>emacs</command> a lot, so we install that too.
4868 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4869 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4870 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4871 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4872 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4873 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4879 <para> Finally, check out a copy of GHC sources from
4880 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
4887 <Sect2><Title>Building GHC</Title>
4890 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source">);
4891 the bullets below only tell
4892 you about Windows-specific wrinkles.</para>
4896 If you used <Command>autoconf</Command> instead of <Command>autoreconf</Command>,
4897 you'll get an error when you run <filename>./configure</filename>:
4900 creating mk/config.h
4901 mk/config.h is unchanged
4903 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4904 ./configure: ./configure: No such file or directory
4905 configure: error: ./configure failed for ghc
4910 <listitem> <para><command>autoreconf</command> seems to create the file <filename>configure</filename>
4911 read-only. So if you need to run autoreconf again (which I sometimes do for safety's sake),
4914 /usr/bin/autoconf: cannot create configure: permission denied
4916 Solution: delete <filename>configure</filename> first.
4921 After <command>autoreconf</command> run <command>./configure</command> in
4922 <filename>fptools/</filename> thus:
4925 ./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc
4927 This is the point at which you specify that you are building GHC-mingw
4928 (see <xref linkend="ghc-mingw">). </para>
4930 <para> Both these options are important! It's possible to get into
4931 trouble using the wrong C compiler!</para>
4933 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4934 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4935 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4936 understand a Cygwin path. For example, if you
4937 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4938 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4939 time it tries to invoke it. Worse, the failure comes with
4940 no error message whatsoever. GHC simply fails silently when first invoked,
4941 typically leaving you with this:
4943 make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4944 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4945 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4946 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4947 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4948 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4949 -optc-fomit-frame-pointer -O2 -static
4950 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4951 make[2]: *** [Adjustor.o] Error 1
4952 make[1]: *** [all] Error 1
4953 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4954 make: *** [all] Error 1
4960 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin">)
4961 you'll have to do something more like:
4963 ./configure --with-gcc=...the Cygwin gcc...
4969 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4970 This file occasionally remembers out-of-date configuration information, which
4971 can be really confusing.
4975 <listitem><para> You almost certainly want to set
4979 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config">).
4980 This tells the build system not to split each library into a myriad of little object files, one
4981 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4982 it dramatically increases the time taken to build the libraries in the first place.
4986 <listitem><para> Do not attempt to build the documentation.
4987 It needs all kinds of wierd Jade stuff that we haven't worked out for
4988 Win32.</para></listitem>