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 (preferably version 4.08+) 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@galconn.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>$CVSROOT</literal> environment variable to
139 <literal>:pserver:anoncvs@glass.cse.ogi.edu:/cvs</literal></para>
142 <para>Run the command</para>
146 <para>The password is simply <literal>cvs</literal>. This
147 sets up a file in your home directory called
148 <literal>.cvspass</literal>, which squirrels away the
149 dummy password, so you only need to do this step once.</para>
153 <para>Now go to <xref linkend="cvs-first">.</para>
158 <sect3 id="cvs-read-write">
159 <title>Remote Read-Write CVS Access</title>
161 <para>We generally supply read-write access to folk doing
162 serious development on some part of the source tree, when
163 going through us would be a pain. If you're developing some
164 feature, or think you have the time and inclination to fix
165 bugs in our sources, feel free to ask for read-write
166 access. There is a certain amount of responsibility that goes
167 with commit privileges; we are more likely to grant you access
168 if you've demonstrated your competence by sending us patches
169 via mail in the past.</para>
171 <para>To get remote read-write CVS access, you need to do the
172 following steps.</para>
176 <para>Make sure that <literal>cvs</literal> and
177 <literal>ssh</literal> are both installed on your
182 <para>Generate a DSA private-key/public-key pair, thus:</para>
186 <para>(<literal>ssh-keygen</literal> comes with
187 <literal>ssh</literal>.) Running <literal>ssh-keygen
188 -d</literal> creates the private and public keys in
189 <literal>$HOME/.ssh/id_dsa</literal> and
190 <literal>$HOME/.ssh/id_dsa.pub</literal> respectively
191 (assuming you accept the standard defaults).</para>
193 <para><literal>ssh-keygen -d</literal> will only work if
194 you have Version 2 <literal>ssh</literal> installed; it
195 will fail harmlessly otherwise. If you only have Version
196 1 you can instead generate an RSA key pair using plain</para>
201 <para>Doing so creates the private and public RSA keys in
202 <literal>$HOME/.ssh/identity</literal> and
203 <literal>$HOME/.ssh/identity.pub</literal>
206 <para>[Deprecated.] Incidentally, you can force a Version
207 2 <literal>ssh</literal> to use the Version 1 protocol by
208 creating <literal>$HOME/config</literal> with the
209 following in it:</para>
217 <para>In both cases, <literal>ssh-keygen</literal> will
218 ask for a <firstterm>passphrase</firstterm>. The
219 passphrase is a password that protects your private key.
220 In response to the 'Enter passphrase' question, you can
224 <para>[Recommended.] Enter a passphrase, which you
225 will quote each time you use CVS.
226 <literal>ssh-agent</literal> makes this entirely
230 <para>[Deprecated.] Just hit return (i.e. use an empty
231 passphrase); then you won't need to quote the
232 passphrase when using CVS. The downside is that
233 anyone who can see into your <literal>.ssh</literal>
234 directory, and thereby get your private key, can mess
235 up the repository. So you must keep the
236 <literal>.ssh</literal> directory with draconian
237 no-access permissions.</para>
243 [Windows users.] The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
244 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
245 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
246 and run it as follows:
248 c:\tmp> set CYGWIN32=tty
249 c:\tmp> c:/user/local/bin/ssh-keygen1
252 <para>[Windows users.] To protect your
253 <literal>.ssh</literal> from access by anyone else,
254 right-click your <literal>.ssh</literal> directory, and
255 select <literal>Properties</literal>. If you are not on
256 the access control list, add yourself, and give yourself
257 full permissions (the second panel). Remove everyone else
258 from the access control list. Don't leave them there but
259 deny them access, because 'they' may be a list that
261 <para>[March 2003] In fact <command>ssh</command> 3.6.1 now seems to <emphasis>require</emphasis>
262 you to have Unix permissions 600 (read/write for owner only)
263 on the <literal>.ssh/identity</literal> file, else it
264 bombs out. For your local C drive, it seems that <literal>chmod 600 identity</literal> works,
265 but on Windows NT/XP, it doesn't work on a network drive (exact dteails obscure).
266 The solution seems to be to set the CYGWIN environment
267 variable to "<literal>ntsec neta</literal>". The CYGWIN environment variable is discussed
268 in <ulink url="http://cygwin.com/cygwin-ug-net/using-cygwinenv.html">the Cygwin User's Guide</ulink>,
269 and there are more details in <ulink url="http://cygwin.com/faq/faq_4.html#SEC44">the Cygwin FAQ</ulink>.
274 <para>Send a message to to the CVS repository
275 administrator (currently Jeff Lewis
276 <email>jeff@galconn.com</email>), containing:</para>
279 <para>Your desired user-name.</para>
282 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
283 <literal>.ssh/identity.pub</literal>).</para>
286 <para>He will set up your account.</para>
290 <para>Set the following environment variables:</para>
294 <constant>$HOME</constant>: points to your home directory. This is where CVS
295 will look for its <filename>.cvsrc</filename> file.
301 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
303 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
304 <literal>ssh</literal> assumes that your CVS client
305 understands how to execute shell script
306 ("#!"s,really), which is what
307 <literal>ssh</literal> is. This may not be the case on
308 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
309 <literal>ssh1</literal>.</para>
313 <para><literal>$CVSROOT</literal> to
314 <literal>:ext:</literal><replaceable>your-username</replaceable>
315 <literal>@cvs.haskell.org:/home/cvs/root</literal>
316 where <replaceable>your-username</replaceable> is your user name on
317 <literal>cvs.haskell.org</literal>.
319 <para>The <literal>CVSROOT</literal> environment variable will
320 be recorded in the checked-out tree, so you don't need to set
321 this every time. </para>
327 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
328 if you want to use an Emacs buffer for typing in those long commit messages.
334 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
335 set this to point to <filename>bash.exe</filename>.
346 Put the following in <filename>$HOME/.cvsrc</filename>:
357 These are the default options for the specified CVS commands,
358 and represent better defaults than the usual ones. (Feel
359 free to change them.)
363 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
364 the 8.3 DOS filesystem, but that restriction should have
365 been lifted by now (i.e., you're using VFAT or later filesystems.) If
366 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
374 <para>[Experts.] Once your account is set up, you can get
375 access from other machines without bothering Jeff, thus:</para>
378 <para>Generate a public/private key pair on the new
382 <para>Use ssh to log in to
383 <literal>cvs.haskell.org</literal>, from your old
387 <para>Add the public key for the new machine to the file
388 <literal>$HOME/ssh/authorized_keys</literal> on
389 <literal>cvs.haskell.org</literal>.
390 (<literal>authorized_keys2</literal>, I think, for Version
394 <para>Make sure that the new version of
395 <literal>authorized_keys</literal> still has 600 file
404 <sect2 id="cvs-first">
405 <title>Checking Out a Source Tree</title>
409 <para>Make sure you set your <literal>CVSROOT</literal>
410 environment variable according to either of the remote
411 methods above. The Approved Way to check out a source tree
412 is as follows:</para>
415 $ cvs checkout fpconfig
418 <para>At this point you have a new directory called
419 <literal>fptools</literal> which contains the basic stuff
420 for the fptools suite, including the configuration files and
421 some other junk. </para>
423 <para>[Windows users.] The following messages appear to be harmless:
425 setsockopt IPTOS_LOWDELAY: Invalid argument
426 setsockopt IPTOS_THROUGHPUT: Invalid argument
431 <para>You can call the fptools directory whatever you like,
432 CVS won't mind: </para>
435 $ mv fptools <replaceable>directory</replaceable>
438 <para> NB: after you've read the CVS manual you might be
439 tempted to try</para>
441 $ cvs checkout -d <replaceable>directory</replaceable> fpconfig
444 <para>instead of checking out <literal>fpconfig</literal>
445 and then renaming it. But this doesn't work, and will
446 result in checking out the entire repository instead of just
447 the <literal>fpconfig</literal> bit.</para>
449 $ cd <replaceable>directory</replaceable>
450 $ cvs checkout ghc hslibs libraries
453 <para>The second command here checks out the relevant
454 modules you want to work on. For a GHC build, for instance,
455 you need at least the <literal>ghc</literal>,
456 <literal>hslibs</literal> and <literal>libraries</literal>
457 modules (for a full list of the projects available, see
458 <xref linkend="projects">).</para>
460 <para>Remeber that if you do not have
461 <literal>happy</literal> installed, you need to check it out
467 <sect2 id="cvs-committing">
468 <title>Committing Changes</title>
470 <para>This is only if you have read-write access to the
471 repository. For anoncvs users, CVS will issue a "read-only
472 repository" error if you try to commit changes.</para>
476 <para>Build the software, if necessary. Unless you're just
477 working on documentation, you'll probably want to build the
478 software in order to test any changes you make.</para>
482 <para>Make changes. Preferably small ones first.</para>
486 <para>Test them. You can see exactly what changes you've
487 made by using the <literal>cvs diff</literal> command:</para>
491 <para>lists all the changes (using the
492 <literal>diff</literal> command) in and below the current
493 directory. In emacs, <literal>C-c C-v =</literal> runs
494 <literal>cvs diff</literal> on the current buffer and shows
495 you the results.</para>
499 <para>Before checking in a change, you need to update your
506 <para>This pulls in any changes that other people have made,
507 and merges them with yours. If there are any conflicts, CVS
508 will tell you, and you'll have to resolve them before you
509 can check your changes in. The documentation describes what
510 to do in the event of a conflict.</para>
512 <para>It's not always necessary to do a full cvs update
513 before checking in a change, since CVS will always tell you
514 if you try to check in a file that someone else has changed.
515 However, you should still update at regular intervals to
516 avoid making changes that don't work in conjuction with
517 changes that someone else made. Keeping an eye on what goes
518 by on the mailing list can help here.</para>
522 <para>When you're happy that your change isn't going to
523 break anything, check it in. For a one-file change:</para>
526 $ cvs commit <replaceable>filename</replaceable>
529 <para>CVS will then pop up an editor for you to enter a
530 "commit message", this is just a short description
531 of what your change does, and will be kept in the history of
534 <para>If you're using emacs, simply load up the file into a
535 buffer and type <literal>C-x C-q</literal>, and emacs will
536 prompt for a commit message and then check in the file for
539 <para>For a multiple-file change, things are a bit
540 trickier. There are several ways to do this, but this is the
541 way I find easiest. First type the commit message into a
542 temporary file. Then either</para>
545 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
548 <para>or, if nothing else has changed in this part of the
552 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
555 <para>where <replaceable>directory</replaceable> is a common
556 parent directory for all your changes, and
557 <replaceable>commit-message</replaceable> is the name of the
558 file containing the commit message.</para>
560 <para>Shortly afterwards, you'll get some mail from the
561 relevant mailing list saying which files changed, and giving
562 the commit message. For a multiple-file change, you should
563 still get only <emphasis>one</emphasis> message.</para>
568 <sect2 id="cvs-update">
569 <title>Updating Your Source Tree</title>
571 <para>It can be tempting to cvs update just part of a source
572 tree to bring in some changes that someone else has made, or
573 before committing your own changes. This is NOT RECOMMENDED!
574 Quite often changes in one part of the tree are dependent on
575 changes in another part of the tree (the
576 <literal>mk/*.mk</literal> files are a good example where
577 problems crop up quite often). Having an inconsistent tree is a
578 major cause of headaches. </para>
580 <para>So, to avoid a lot of hassle, follow this recipe for
581 updating your tree: </para>
585 $ cvs update -P 2>&1 | tee log</screen>
587 <para>Look at the log file, and fix any conflicts (denoted by a
588 <quote>C</quote> in the first column). New directories may have
589 appeared in the repository; CVS doesn't check these out by
590 default, so to get new directories you have to explicitly do
592 $ cvs update -d</screen>
593 in each project subdirectory. Don't do this at the top level,
594 because then <emphasis>all</emphasis> the projects will be
597 <para>If you're using multiple build trees, then for every build
598 tree you have pointing at this source tree, you need to update
599 the links in case any new files have appeared: </para>
602 $ cd <replaceable>build-tree</replaceable>
603 $ lndir <replaceable>source-tree</replaceable>
606 <para>Some files might have been removed, so you need to remove
607 the links pointing to these non-existent files:</para>
610 $ find . -xtype l -exec rm '{}' \;
613 <para>To be <emphasis>really</emphasis> safe, you should do
616 <screen>$ gmake all</screen>
618 <para>from the top-level, to update the dependencies and build
619 any changed files. </para>
622 <sect2 id="cvs-tags">
623 <title>GHC Tag Policy</title>
625 <para>If you want to check out a particular version of GHC,
626 you'll need to know how we tag versions in the repository. The
627 policy (as of 4.04) is:</para>
631 <para>The tree is branched before every major release. The
632 branch tag is <literal>ghc-x-xx-branch</literal>, where
633 <literal>x-xx</literal> is the version number of the release
634 with the <literal>'.'</literal> replaced by a
635 <literal>'-'</literal>. For example, the 4.04 release lives
636 on <literal>ghc-4-04-branch</literal>.</para>
640 <para>The release itself is tagged with
641 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
642 called <literal>ghc-4-06</literal>.</para>
646 <para>We didn't always follow these guidelines, so to see
647 what tags there are for previous versions, do <literal>cvs
648 log</literal> on a file that's been around for a while (like
649 <literal>fptools/ghc/README</literal>).</para>
653 <para>So, to check out a fresh GHC 4.06 tree you would
657 $ cvs co -r ghc-4-06 fpconfig
659 $ cvs co -r ghc-4-06 ghc hslibs
663 <sect2 id="cvs-hints">
664 <title>General Hints</title>
668 <para>As a general rule: commit changes in small units,
669 preferably addressing one issue or implementing a single
670 feature. Provide a descriptive log message so that the
671 repository records exactly which changes were required to
672 implement a given feature/fix a bug. I've found this
673 <emphasis>very</emphasis> useful in the past for finding out
674 when a particular bug was introduced: you can just wind back
675 the CVS tree until the bug disappears.</para>
679 <para>Keep the sources at least *buildable* at any given
680 time. No doubt bugs will creep in, but it's quite easy to
681 ensure that any change made at least leaves the tree in a
682 buildable state. We do nightly builds of GHC to keep an eye
683 on what things work/don't work each day and how we're doing
684 in relation to previous verions. This idea is truely wrecked
685 if the compiler won't build in the first place!</para>
689 <para>To check out extra bits into an already-checked-out
690 tree, use the following procedure. Suppose you have a
691 checked-out fptools tree containing just ghc, and you want
692 to add nofib to it:</para>
703 $ cvs update -d nofib
706 <para>(the -d flag tells update to create a new
707 directory). If you just want part of the nofib suite, you
712 $ cvs checkout nofib/spectral
715 <para>This works because <literal>nofib</literal> is a
716 module in its own right, and spectral is a subdirectory of
717 the nofib module. The path argument to checkout must always
718 start with a module name. There's no equivalent form of this
719 command using <literal>update</literal>.</para>
725 <sect1 id="projects">
726 <title>What projects are there?</title>
728 <para>The <literal>fptools</literal> suite consists of several
729 <firstterm>projects</firstterm>, most of which can be downloaded,
730 built and installed individually. Each project corresponds to a
731 subdirectory in the source tree, and if checking out from CVS then
732 each project can be checked out individually by sitting in the top
733 level of your source tree and typing <command>cvs checkout
734 <replaceable>project</replaceable></command>.</para>
736 <para>Here is a list of the projects currently available:</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>green-card</literal></term>
762 <indexterm><primary><literal>green-card</literal></primary><secondary>project</secondary></indexterm>
765 url="http://www.haskell.org/greencard/">Green Card</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),
1010 including native-code generator.</para>
1015 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
1016 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1018 <para>A registerised port is available for version 4.08,
1019 but GHC hasn't been built on that platform since (as far
1020 as we know). No native-code generator.</para>
1025 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1026 <indexterm><primary>i386-*-linux</primary></indexterm>
1028 <para>GHC works registerised and has a native code
1029 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1030 or later. NOTE about <literal>glibc</literal> versions:
1031 GHC binaries built on a system running <literal>glibc
1032 2.0</literal> won't work on a system running
1033 <literal>glibc 2.1</literal>, and vice versa. In general,
1034 don't expect compatibility between
1035 <literal>glibc</literal> versions, even if the shared
1036 library version hasn't changed.</para>
1041 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or
1043 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1045 <para>GHC works registerised. Pre-built packages are
1046 available in the native package format, so if you just
1047 need binaries you're better off just installing the
1048 package (it might even be on your installation
1054 <term>i386-unknown-openbsd (PCs running OpenBSD)</term>
1055 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1057 <para>Supported, with native code generator. Packages are
1058 available through the ports system in the native package
1064 <term>i386-unknown-netbsd (PCs running NetBSD and
1066 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1068 <para>Will require some minor porting effort, but should
1069 work registerised.</para>
1074 <term>i386-unknown-mingw32 (PCs running Windows)</term>
1075 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1077 <para>Fully supported under Win9x, WinNT, Win2k, and
1078 WinXP. Includes a native code generator. Building from
1079 source requires a recent <ulink
1080 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1081 to be installed.</para>
1086 <term>ia64-unknown-linux</term>
1087 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1089 <para>GHC currently works unregisterised. A registerised
1090 port is in progress.</para>
1095 <term>mips-sgi-irix5</term>
1096 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1098 <para>Port has worked in the past, but hasn't been tested
1099 for some time (and will certainly have rotted in various
1100 ways). As usual, we don't have access to machines and
1101 there hasn't been an overwhelming demand for this port,
1102 but feel free to get in touch.</para>
1107 <term>powerpc-ibm-aix</term>
1108 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1110 <para>Port currently doesn't work, needs some minimal
1111 porting effort. As usual, we don't have access to
1112 machines and there hasn't been an overwhelming demand for
1113 this port, but feel free to get in touch.</para>
1118 <term>powerpc-apple-darwin</term>
1119 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1121 <para>Supported registerised. No native code
1127 <term>powerpc-apple-linux</term>
1128 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1130 <para>Not supported (yet).</para>
1135 <para>Various other systems have had GHC ported to them in the
1136 distant past, including various Motorola 68k boxes. The 68k
1137 support still remains, but porting to one of these systems will
1138 certainly be a non-trivial task.</para>
1142 <title>What machines the other tools run on</title>
1144 <para>Unless you hear otherwise, the other tools work if GHC
1150 <sect1 id="sec-pre-supposed">
1151 <title>Installing pre-supposed utilities</title>
1153 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1154 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1156 <para>Here are the gory details about some utility programs you
1157 may need; <command>perl</command>, <command>gcc</command> and
1158 <command>happy</command> are the only important
1159 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1160 important if you're going for Parallel Haskell.) The
1161 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1162 script will tell you if you are missing something.</para>
1168 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1169 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1171 <para>GHC is required to build many of the tools, including
1172 GHC itself. If you need to port GHC to your platform
1173 because there isn't a binary distribution of GHC available,
1174 then see <xref linkend="sec-porting-ghc">.</para>
1176 <para>Which version of GHC you need will depend on the
1177 packages you intend to build. GHC itself will normally
1178 build using one of several older versions of itself - check
1179 the announcement or release notes for details.</para>
1185 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1186 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1188 <para><emphasis>You have to have Perl to proceed!</emphasis>
1189 Perl version 5 at least is required. GHC has been known to
1190 tickle bugs in Perl, so if you find that Perl crashes when
1191 running GHC try updating (or downgrading) your Perl
1192 installation. Versions of Perl that we use and are known to
1193 be fairly stable are 5.005 and 5.6.1.</para>
1195 <para>For Win32 platforms, you should use the binary
1196 supplied in the InstallShield (copy it to
1197 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1200 <para>Perl should be put somewhere so that it can be invoked
1201 by the <literal>#!</literal> script-invoking
1202 mechanism. The full pathname may need to be less than 32
1203 characters long on some systems.</para>
1208 <term>GNU C (<command>gcc</command>)</term>
1209 <indexterm><primary>pre-supposed: GCC (GNU C
1210 compiler)</primary></indexterm> <indexterm><primary>GCC (GNU C
1211 compiler), pre-supposed</primary></indexterm>
1213 <para>We recommend using GCC version 2.95.2 on all
1214 platforms. Failing that, version 2.7.2 is stable on most
1215 platforms. Earlier versions of GCC can be assumed not to
1216 work, and versions in between 2.7.2 and 2.95.2 (including
1217 <command>egcs</command>) have varying degrees of stability
1218 depending on the platform.</para>
1220 <para>GCC 3.2 is currently known to have problems building
1221 GHC on Sparc, but is stable on x86.</para>
1223 <para>GCC 3.3 currnetly cannot be used to build GHC, due to
1224 some problems with the new C preprocessor.</para>
1226 <para>If your GCC dies with “internal error” on
1227 some GHC source file, please let us know, so we can report
1228 it and get things improved. (Exception: on iX86
1229 boxes—you may need to fiddle with GHC's
1230 <option>-monly-N-regs</option> option; see the User's
1236 <term>GNU Make</term>
1237 <indexterm><primary>make</primary><secondary>GNU</secondary>
1240 <para>The fptools build system makes heavy use of features
1241 specific to GNU <command>make</command>, so you must have
1242 this installed in order to build any of the fptools
1249 <indexterm><primary>Happy</primary></indexterm>
1251 <para>Happy is a parser generator tool for Haskell, and is
1252 used to generate GHC's parsers. Happy is written in
1253 Haskell, and is a project in the CVS repository
1254 (<literal>fptools/happy</literal>). It can be built from
1255 source, but bear in mind that you'll need GHC installed in
1256 order to build it. To avoid the chicken/egg problem,
1257 install a binary distribtion of either Happy or GHC to get
1258 started. Happy distributions are available from <ulink
1259 url="http://www.haskell.org/happy/">Happy's Web
1260 Page</ulink>.</para>
1265 <term>Autoconf</term>
1266 <indexterm><primary>pre-supposed: Autoconf</primary></indexterm>
1267 <indexterm><primary>Autoconf, pre-supposed</primary></indexterm>
1269 <para>GNU Autoconf is needed if you intend to build from the
1270 CVS sources, it is <emphasis>not</emphasis> needed if you
1271 just intend to build a standard source distribution.</para>
1273 <para>Autoconf builds the <command>configure</command>
1274 script from <filename>configure.in</filename> and
1275 <filename>aclocal.m4</filename>. If you modify either of
1276 these files, you'll need <command>autoconf</command> to
1277 rebuild <filename>configure</filename>.</para>
1282 <term><command>sed</command></term>
1283 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1284 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1286 <para>You need a working <command>sed</command> if you are
1287 going to build from sources. The build-configuration stuff
1288 needs it. GNU sed version 2.0.4 is no good! It has a bug
1289 in it that is tickled by the build-configuration. 2.0.5 is
1290 OK. Others are probably OK too (assuming we don't create too
1291 elaborate configure scripts.)</para>
1296 <para>One <literal>fptools</literal> project is worth a quick note
1297 at this point, because it is useful for all the others:
1298 <literal>glafp-utils</literal> contains several utilities which
1299 aren't particularly Glasgow-ish, but Occasionally Indispensable.
1300 Like <command>lndir</command> for creating symbolic link
1303 <sect2 id="pre-supposed-gph-tools">
1304 <title>Tools for building parallel GHC (GPH)</title>
1308 <term>PVM version 3:</term>
1309 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1310 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1312 <para>PVM is the Parallel Virtual Machine on which
1313 Parallel Haskell programs run. (You only need this if you
1314 plan to run Parallel Haskell. Concurent Haskell, which
1315 runs concurrent threads on a uniprocessor doesn't need
1316 it.) Underneath PVM, you can have (for example) a network
1317 of workstations (slow) or a multiprocessor box
1320 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1321 It is readily available on the net; I think I got it from
1322 <literal>research.att.com</literal>, in
1323 <filename>netlib</filename>.</para>
1325 <para>A PVM installation is slightly quirky, but easy to
1326 do. Just follow the <filename>Readme</filename>
1327 instructions.</para>
1332 <term><command>bash</command>:</term>
1333 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1335 <para>Sadly, the <command>gr2ps</command> script, used to
1336 convert “parallelism profiles” to PostScript,
1337 is written in Bash (GNU's Bourne Again shell). This bug
1338 will be fixed (someday).</para>
1344 <sect2 id="pre-supposed-other-tools">
1345 <title>Other useful tools</title>
1350 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1351 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1353 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1354 to build a couple of utilities in
1355 <literal>glafp-utils</literal>. Depending on your
1356 operating system, the supplied <command>lex</command> may
1357 or may not work; you should get the GNU version.</para>
1362 <para>More tools are required if you want to format the documentation
1363 that comes with GHC and other fptools projects. See <xref
1364 linkend="building-docs">.</para>
1368 <sect1 id="sec-building-from-source">
1369 <title>Building from source</title>
1371 <indexterm><primary>Building from source</primary></indexterm>
1372 <indexterm><primary>Source, building from</primary></indexterm>
1374 <para>You've been rash enough to want to build some of the Glasgow
1375 Functional Programming tools (GHC, Happy, nofib, etc.) from
1376 source. You've slurped the source, from the CVS repository or
1377 from a source distribution, and now you're sitting looking at a
1378 huge mound of bits, wondering what to do next.</para>
1380 <para>Gingerly, you type <command>make</command>. Wrong
1383 <para>This rest of this guide is intended for duffers like me, who
1384 aren't really interested in Makefiles and systems configurations,
1385 but who need a mental model of the interlocking pieces so that
1386 they can make them work, extend them consistently when adding new
1387 software, and lay hands on them gently when they don't
1390 <sect2 id="quick-start">
1391 <title>Quick Start</title>
1393 <para>If you are starting from a source distribution, and just
1394 want a completely standard build, then the following should
1397 <screen>$ ./configure
1402 <para>For GHC, this will do a 2-stage bootstrap build of the
1403 compiler, with profiling libraries, and install the
1406 <para>If you want to do anything at all non-standard, or you
1407 want to do some development, read on...</para>
1410 <sect2 id="sec-source-tree">
1411 <title>Your source tree</title>
1413 <para>The source code is held in your <emphasis>source
1414 tree</emphasis>. The root directory of your source tree
1415 <emphasis>must</emphasis> contain the following directories and
1420 <para><filename>Makefile</filename>: the root
1425 <para><filename>mk/</filename>: the directory that contains
1426 the main Makefile code, shared by all the
1427 <literal>fptools</literal> software.</para>
1431 <para><filename>configure.in</filename>,
1432 <filename>config.sub</filename>,
1433 <filename>config.guess</filename>: these files support the
1434 configuration process.</para>
1438 <para><filename>install-sh</filename>.</para>
1442 <para>All the other directories are individual
1443 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1444 system—for example, the Glasgow Haskell Compiler
1445 (<literal>ghc</literal>), the Happy parser generator
1446 (<literal>happy</literal>), the <literal>nofib</literal>
1447 benchmark suite, and so on. You can have zero or more of these.
1448 Needless to say, some of them are needed to build others.</para>
1450 <para>The important thing to remember is that even if you want
1451 only one project (<literal>happy</literal>, say), you must have
1452 a source tree whose root directory contains
1453 <filename>Makefile</filename>, <filename>mk/</filename>,
1454 <filename>configure.in</filename>, and the project(s) you want
1455 (<filename>happy/</filename> in this case). You cannot get by
1456 with just the <filename>happy/</filename> directory.</para>
1460 <title>Build trees</title>
1461 <indexterm><primary>build trees</primary></indexterm>
1462 <indexterm><primary>link trees, for building</primary></indexterm>
1464 <para>If you just want to build the software once on a single
1465 platform, then your source tree can also be your build tree, and
1466 you can skip the rest of this section.</para>
1468 <para>We often want to build multiple versions of our software
1469 for different architectures, or with different options
1470 (e.g. profiling). It's very desirable to share a single copy of
1471 the source code among all these builds.</para>
1473 <para>So for every source tree we have zero or more
1474 <emphasis>build trees</emphasis>. Each build tree is initially
1475 an exact copy of the source tree, except that each file is a
1476 symbolic link to the source file, rather than being a copy of
1477 the source file. There are “standard” Unix
1478 utilities that make such copies, so standard that they go by
1480 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1481 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1482 are two (If you don't have either, the source distribution
1483 includes sources for the X11
1484 <command>lndir</command>—check out
1485 <filename>fptools/glafp-utils/lndir</filename>). See <Xref
1486 LinkEnd="sec-storysofar"> for a typical invocation.</para>
1488 <para>The build tree does not need to be anywhere near the
1489 source tree in the file system. Indeed, one advantage of
1490 separating the build tree from the source is that the build tree
1491 can be placed in a non-backed-up partition, saving your systems
1492 support people from backing up untold megabytes of
1493 easily-regenerated, and rapidly-changing, gubbins. The golden
1494 rule is that (with a single exception—<XRef
1495 LinkEnd="sec-build-config">) <emphasis>absolutely everything in
1496 the build tree is either a symbolic link to the source tree, or
1497 else is mechanically generated</emphasis>. It should be
1498 perfectly OK for your build tree to vanish overnight; an hour or
1499 two compiling and you're on the road again.</para>
1501 <para>You need to be a bit careful, though, that any new files
1502 you create (if you do any development work) are in the source
1503 tree, not a build tree!</para>
1505 <para>Remember, that the source files in the build tree are
1506 <emphasis>symbolic links</emphasis> to the files in the source
1507 tree. (The build tree soon accumulates lots of built files like
1508 <filename>Foo.o</filename>, as well.) You can
1509 <emphasis>delete</emphasis> a source file from the build tree
1510 without affecting the source tree (though it's an odd thing to
1511 do). On the other hand, if you <emphasis>edit</emphasis> a
1512 source file from the build tree, you'll edit the source-tree
1513 file directly. (You can set up Emacs so that if you edit a
1514 source file from the build tree, Emacs will silently create an
1515 edited copy of the source file in the build tree, leaving the
1516 source file unchanged; but the danger is that you think you've
1517 edited the source file whereas actually all you've done is edit
1518 the build-tree copy. More commonly you do want to edit the
1519 source file.)</para>
1521 <para>Like the source tree, the top level of your build tree
1522 must be (a linked copy of) the root directory of the
1523 <literal>fptools</literal> suite. Inside Makefiles, the root of
1524 your build tree is called
1525 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1526 In the rest of this document path names are relative to
1527 <constant>$(FPTOOLS_TOP)</constant> unless
1528 otherwise stated. For example, the file
1529 <filename>ghc/mk/target.mk</filename> is actually
1530 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</filename>.</para>
1533 <sect2 id="sec-build-config">
1534 <title>Getting the build you want</title>
1536 <para>When you build <literal>fptools</literal> you will be
1537 compiling code on a particular <emphasis>host
1538 platform</emphasis>, to run on a particular <emphasis>target
1539 platform</emphasis> (usually the same as the host
1540 platform)<indexterm><primary>platform</primary></indexterm>.
1541 The difficulty is that there are minor differences between
1542 different platforms; minor, but enough that the code needs to be
1543 a bit different for each. There are some big differences too:
1544 for a different architecture we need to build GHC with a
1545 different native-code generator.</para>
1547 <para>There are also knobs you can turn to control how the
1548 <literal>fptools</literal> software is built. For example, you
1549 might want to build GHC optimised (so that it runs fast) or
1550 unoptimised (so that you can compile it fast after you've
1551 modified it. Or, you might want to compile it with debugging on
1552 (so that extra consistency-checking code gets included) or off.
1555 <para>All of this stuff is called the
1556 <emphasis>configuration</emphasis> of your build. You set the
1557 configuration using a three-step process.</para>
1561 <term>Step 1: get ready for configuration.</term>
1563 <para>NOTE: if you're starting from a source distribution,
1564 rather than CVS sources, you can skip this step.</para>
1566 <para>Change directory to
1567 <constant>$(FPTOOLS_TOP)</constant> and
1569 <command>autoconf</command><indexterm><primary>autoconf</primary></indexterm>
1570 (with no arguments). This GNU program converts
1571 <filename><constant>$(FPTOOLS_TOP)</constant>/configure.in</filename>
1572 to a shell script called
1573 <filename><constant>$(FPTOOLS_TOP)</constant>/configure</filename>.
1576 <para>Some projects, including GHC, have their own
1577 configure script. If there's an
1578 <constant>$(FPTOOLS_TOP)/<project>/configure.in</constant>,
1579 then you need to run <command>autoconf</command> in that
1580 directory too.</para>
1582 <para>Both these steps are completely
1583 platform-independent; they just mean that the
1584 human-written file (<filename>configure.in</filename>) can
1585 be short, although the resulting shell script,
1586 <command>configure</command>, and
1587 <filename>mk/config.h.in</filename>, are long.</para>
1592 <term>Step 2: system configuration.</term>
1594 <para>Runs the newly-created <command>configure</command>
1595 script, thus:</para>
1598 ./configure <optional><parameter>args</parameter></optional>
1601 <para><command>configure</command>'s mission is to scurry
1602 round your computer working out what architecture it has,
1603 what operating system, whether it has the
1604 <Function>vfork</Function> system call, where
1605 <command>yacc</command> is kept, whether
1606 <command>gcc</command> is available, where various obscure
1607 <literal>#include</literal> files are, whether it's a
1608 leap year, and what the systems manager had for lunch. It
1609 communicates these snippets of information in two
1616 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1618 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1619 substituting for things between
1620 “<literal>@</literal>” brackets. So,
1621 “<literal>@HaveGcc@</literal>” will be
1622 replaced by “<literal>YES</literal>” or
1623 “<literal>NO</literal>” depending on what
1624 <command>configure</command> finds.
1625 <filename>mk/config.mk</filename> is included by every
1626 Makefile (directly or indirectly), so the
1627 configuration information is thereby communicated to
1628 all Makefiles.</para>
1632 <para> It translates
1633 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1635 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1636 The latter is <literal>#include</literal>d by
1637 various C programs, which can thereby make use of
1638 configuration information.</para>
1642 <para><command>configure</command> takes some optional
1643 arguments. Use <literal>./configure --help</literal> to
1644 get a list of the available arguments. Here are some of
1645 the ones you might need:</para>
1649 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1650 <indexterm><primary><literal>--with-ghc</literal></primary>
1653 <para>Specifies the path to an installed GHC which
1654 you would like to use. This compiler will be used
1655 for compiling GHC-specific code (eg. GHC itself).
1656 This option <emphasis>cannot</emphasis> be specified
1657 using <filename>build.mk</filename> (see later),
1658 because <command>configure</command> needs to
1659 auto-detect the version of GHC you're using. The
1660 default is to look for a compiler named
1661 <literal>ghc</literal> in your path.</para>
1666 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1667 <indexterm><primary><literal>--with-hc</literal></primary>
1670 <para>Specifies the path to any installed Haskell
1671 compiler. This compiler will be used for compiling
1672 generic Haskell code. The default is to use
1673 <literal>ghc</literal>.</para>
1678 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1679 <indexterm><primary><literal>--with-gcc</literal></primary>
1682 <para>Specifies the path to the installed GCC. This
1683 compiler will be used to compile all C files,
1684 <emphasis>except</emphasis> any generated by the
1685 installed Haskell compiler, which will have its own
1686 idea of which C compiler (if any) to use. The
1687 default is to use <literal>gcc</literal>.</para>
1692 <para><command>configure</command> caches the results of
1693 its run in <filename>config.cache</filename>. Quite often
1694 you don't want that; you're running
1695 <command>configure</command> a second time because
1696 something has changed. In that case, simply delete
1697 <filename>config.cache</filename>.</para>
1702 <term>Step 3: build configuration.</term>
1704 <para>Next, you say how this build of
1705 <literal>fptools</literal> is to differ from the standard
1706 defaults by creating a new file
1707 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1708 <emphasis>in the build tree</emphasis>. This file is the
1709 one and only file you edit in the build tree, precisely
1710 because it says how this build differs from the source.
1711 (Just in case your build tree does die, you might want to
1712 keep a private directory of <filename>build.mk</filename>
1713 files, and use a symbolic link in each build tree to point
1714 to the appropriate one.) So
1715 <filename>mk/build.mk</filename> never exists in the
1716 source tree—you create one in each build tree from
1717 the template. We'll discuss what to put in it
1723 <para>And that's it for configuration. Simple, eh?</para>
1725 <para>What do you put in your build-specific configuration file
1726 <filename>mk/build.mk</filename>? <emphasis>For almost all
1727 purposes all you will do is put make variable definitions that
1728 override those in</emphasis>
1729 <filename>mk/config.mk.in</filename>. The whole point of
1730 <filename>mk/config.mk.in</filename>—and its derived
1731 counterpart <filename>mk/config.mk</filename>—is to define
1732 the build configuration. It is heavily commented, as you will
1733 see if you look at it. So generally, what you do is look at
1734 <filename>mk/config.mk.in</filename>, and add definitions in
1735 <filename>mk/build.mk</filename> that override any of the
1736 <filename>config.mk</filename> definitions that you want to
1737 change. (The override occurs because the main boilerplate file,
1738 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1739 includes <filename>build.mk</filename> after
1740 <filename>config.mk</filename>.)</para>
1742 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1743 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1745 <para>For example, <filename>config.mk.in</filename> contains
1746 the definition:</para>
1749 GhcHcOpts=-O -Rghc-timing
1752 <para>The accompanying comment explains that this is the list of
1753 flags passed to GHC when building GHC itself. For doing
1754 development, it is wise to add <literal>-DDEBUG</literal>, to
1755 enable debugging code. So you would add the following to
1756 <filename>build.mk</filename>:</para>
1758 <para>or, if you prefer,</para>
1761 GhcHcOpts += -DDEBUG
1764 <para>GNU <command>make</command> allows existing definitions to
1765 have new text appended using the “<literal>+=</literal>”
1766 operator, which is quite a convenient feature.)</para>
1768 <para>If you want to remove the <literal>-O</literal> as well (a
1769 good idea when developing, because the turn-around cycle gets a
1770 lot quicker), you can just override
1771 <literal>GhcLibHcOpts</literal> altogether:</para>
1774 GhcHcOpts=-DDEBUG -Rghc-timing
1777 <para>When reading <filename>config.mk.in</filename>, remember
1778 that anything between “@...@” signs is going to be substituted
1779 by <command>configure</command> later. You
1780 <emphasis>can</emphasis> override the resulting definition if
1781 you want, but you need to be a bit surer what you are doing.
1782 For example, there's a line that says:</para>
1788 <para>This defines the Make variables <constant>YACC</constant>
1789 to the pathname for a <command>yacc</command> that
1790 <command>configure</command> finds somewhere. If you have your
1791 own pet <command>yacc</command> you want to use instead, that's
1792 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1798 <para>You do not <emphasis>have</emphasis> to have a
1799 <filename>mk/build.mk</filename> file at all; if you don't,
1800 you'll get all the default settings from
1801 <filename>mk/config.mk.in</filename>.</para>
1803 <para>You can also use <filename>build.mk</filename> to override
1804 anything that <command>configure</command> got wrong. One place
1805 where this happens often is with the definition of
1806 <constant>FPTOOLS_TOP_ABS</constant>: this
1807 variable is supposed to be the canonical path to the top of your
1808 source tree, but if your system uses an automounter then the
1809 correct directory is hard to find automatically. If you find
1810 that <command>configure</command> has got it wrong, just put the
1811 correct definition in <filename>build.mk</filename>.</para>
1815 <sect2 id="sec-storysofar">
1816 <title>The story so far</title>
1818 <para>Let's summarise the steps you need to carry to get
1819 yourself a fully-configured build tree from scratch.</para>
1823 <para> Get your source tree from somewhere (CVS repository
1824 or source distribution). Say you call the root directory
1825 <filename>myfptools</filename> (it does not have to be
1826 called <filename>fptools</filename>). Make sure that you
1827 have the essential files (see <XRef
1828 LinkEnd="sec-source-tree">).</para>
1833 <para>(Optional) Use <command>lndir</command> or
1834 <command>mkshadowdir</command> to create a build tree.</para>
1838 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1841 <para>(N.B. <command>mkshadowdir</command>'s first argument
1842 is taken relative to its second.) You probably want to give
1843 the build tree a name that suggests its main defining
1844 characteristic (in your mind at least), in case you later
1849 <para>Change directory to the build tree. Everything is
1850 going to happen there now.</para>
1853 $ cd /scratch/joe-bloggs/myfptools-sun4
1859 <para>Prepare for system configuration:</para>
1865 <para>(You can skip this step if you are starting from a
1866 source distribution, and you already have
1867 <filename>configure</filename> and
1868 <filename>mk/config.h.in</filename>.)</para>
1870 <para>Some projects, including GHC itself, have their own
1871 configure scripts, so it is necessary to run autoconf again
1872 in the appropriate subdirectories. eg:</para>
1875 $ (cd ghc; autoconf)
1880 <para>Do system configuration:</para>
1886 <para>Don't forget to check whether you need to add any
1887 arguments to <literal>configure</literal>; for example, a
1888 common requirement is to specify which GHC to use with
1889 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1893 <para>Create the file <filename>mk/build.mk</filename>,
1894 adding definitions for your desired configuration
1903 <para>You can make subsequent changes to
1904 <filename>mk/build.mk</filename> as often as you like. You do
1905 not have to run any further configuration programs to make these
1906 changes take effect. In theory you should, however, say
1907 <command>gmake clean</command>, <command>gmake all</command>,
1908 because configuration option changes could affect
1909 anything—but in practice you are likely to know what's
1914 <title>Making things</title>
1916 <para>At this point you have made yourself a fully-configured
1917 build tree, so you are ready to start building real
1920 <para>The first thing you need to know is that <emphasis>you
1921 must use GNU <command>make</command>, usually called
1922 <command>gmake</command>, not standard Unix
1923 <command>make</command></emphasis>. If you use standard Unix
1924 <command>make</command> you will get all sorts of error messages
1925 (but no damage) because the <literal>fptools</literal>
1926 <command>Makefiles</command> use GNU <command>make</command>'s
1927 facilities extensively.</para>
1929 <para>To just build the whole thing, <command>cd</command> to
1930 the top of your <literal>fptools</literal> tree and type
1931 <command>gmake</command>. This will prepare the tree and build
1932 the various projects in the correct order.</para>
1935 <sect2 id="sec-bootstrapping">
1936 <title>Bootstrapping GHC</title>
1938 <para>GHC requires a 2-stage bootstrap in order to provide
1939 full functionality, including GHCi. By a 2-stage bootstrap, we
1940 mean that the compiler is built once using the installed GHC,
1941 and then again using the compiler built in the first stage. You
1942 can also build a stage 3 compiler, but this normally isn't
1943 necessary except to verify that the stage 2 compiler is working
1946 <para>Note that when doing a bootstrap, the stage 1 compiler
1947 must be built, followed by the runtime system and libraries, and
1948 then the stage 2 compiler. The correct ordering is implemented
1949 by the top-level fptools <filename>Makefile</filename>, so if
1950 you want everything to work automatically it's best to start
1951 <command>make</command> from the top of the tree. When building
1952 GHC, the top-level fptools <filename>Makefile</filename> is set
1953 up to do a 2-stage bootstrap by default (when you say
1954 <command>make</command>). Some other targets it supports
1961 <para>Build everything as normal, including the stage 1
1969 <para>Build the stage 2 compiler only.</para>
1976 <para>Build the stage 3 compiler only.</para>
1981 <term>bootstrap</term> <term>bootstrap2</term>
1983 <para>Build stage 1 followed by stage 2.</para>
1988 <term>bootstrap3</term>
1990 <para>Build stages 1, 2 and 3.</para>
1995 <term>install</term>
1997 <para>Install everything, including the compiler built in
1998 stage 2. To override the stage, say <literal>make install
1999 stage=<replaceable>n</replaceable></literal> where
2000 <replaceable>n</replaceable> is the stage to install.</para>
2005 <para>The top-level <filename>Makefile</filename> also arranges
2006 to do the appropriate <literal>make boot</literal> steps (see
2007 below) before actually building anything.</para>
2009 <para>The <literal>stage1</literal>, <literal>stage2</literal>
2010 and <literal>stage3</literal> targets also work in the
2011 <literal>ghc/compiler</literal> directory, but don't forget that
2012 each stage requires its own <literal>make boot</literal> step:
2013 for example, you must do</para>
2015 <screen>$ make boot stage=2</screen>
2017 <para>before <literal>make stage2</literal> in
2018 <literal>ghc/compiler</literal>.</para>
2021 <sect2 id="sec-standard-targets">
2022 <title>Standard Targets</title>
2023 <indexterm><primary>targets, standard makefile</primary></indexterm>
2024 <indexterm><primary>makefile targets</primary></indexterm>
2026 <para>In any directory you should be able to make the following:</para>
2030 <term><literal>boot</literal></term>
2032 <para>does the one-off preparation required to get ready
2033 for the real work. Notably, it does <command>gmake
2034 depend</command> in all directories that contain programs.
2035 It also builds the necessary tools for compilation to
2038 <para>Invoking the <literal>boot</literal> target
2039 explicitly is not normally necessary. From the top-level
2040 <literal>fptools</literal> directory, invoking
2041 <literal>gmake</literal> causes <literal>gmake boot
2042 all</literal> to be invoked in each of the project
2043 subdirectories, in the order specified by
2044 <literal>$(AllTargets)</literal> in
2045 <literal>config.mk</literal>.</para>
2047 <para>If you're working in a subdirectory somewhere and
2048 need to update the dependencies, <literal>gmake
2049 boot</literal> is a good way to do it.</para>
2054 <term><literal>all</literal></term>
2056 <para>makes all the final target(s) for this Makefile.
2057 Depending on which directory you are in a “final
2058 target” may be an executable program, a library
2059 archive, a shell script, or a Postscript file. Typing
2060 <command>gmake</command> alone is generally the same as
2061 typing <command>gmake all</command>.</para>
2066 <term><literal>install</literal></term>
2068 <para>installs the things built by <literal>all</literal>
2069 (except for the documentation). Where does it install
2070 them? That is specified by
2071 <filename>mk/config.mk.in</filename>; you can override it
2072 in <filename>mk/build.mk</filename>, or by running
2073 <command>configure</command> with command-line arguments
2074 like <literal>--bindir=/home/simonpj/bin</literal>; see
2075 <literal>./configure --help</literal> for the full
2081 <term><literal>install-docs</literal></term>
2083 <para>installs the documentation. Otherwise behaves just
2084 like <literal>install</literal>.</para>
2089 <term><literal>uninstall</literal></term>
2091 <para>reverses the effect of
2092 <literal>install</literal>.</para>
2097 <term><literal>clean</literal></term>
2099 <para>Delete all files from the current directory that are
2100 normally created by building the program. Don't delete
2101 the files that record the configuration, or files
2102 generated by <command>gmake boot</command>. Also preserve
2103 files that could be made by building, but normally aren't
2104 because the distribution comes with them.</para>
2109 <term><literal>distclean</literal></term>
2111 <para>Delete all files from the current directory that are
2112 created by configuring or building the program. If you
2113 have unpacked the source and built the program without
2114 creating any other files, <literal>make
2115 distclean</literal> should leave only the files that were
2116 in the distribution.</para>
2121 <term><literal>mostlyclean</literal></term>
2123 <para>Like <literal>clean</literal>, but may refrain from
2124 deleting a few files that people normally don't want to
2130 <term><literal>maintainer-clean</literal></term>
2132 <para>Delete everything from the current directory that
2133 can be reconstructed with this Makefile. This typically
2134 includes everything deleted by
2135 <literal>distclean</literal>, plus more: C source files
2136 produced by Bison, tags tables, Info files, and so
2139 <para>One exception, however: <literal>make
2140 maintainer-clean</literal> should not delete
2141 <filename>configure</filename> even if
2142 <filename>configure</filename> can be remade using a rule
2143 in the <filename>Makefile</filename>. More generally,
2144 <literal>make maintainer-clean</literal> should not delete
2145 anything that needs to exist in order to run
2146 <filename>configure</filename> and then begin to build the
2152 <term><literal>check</literal></term>
2154 <para>run the test suite.</para>
2159 <para>All of these standard targets automatically recurse into
2160 sub-directories. Certain other standard targets do not:</para>
2164 <term><literal>configure</literal></term>
2166 <para>is only available in the root directory
2167 <constant>$(FPTOOLS_TOP)</constant>; it has
2168 been discussed in <XRef
2169 LinkEnd="sec-build-config">.</para>
2174 <term><literal>depend</literal></term>
2176 <para>make a <filename>.depend</filename> file in each
2177 directory that needs it. This <filename>.depend</filename>
2178 file contains mechanically-generated dependency
2179 information; for example, suppose a directory contains a
2180 Haskell source module <filename>Foo.lhs</filename> which
2181 imports another module <literal>Baz</literal>. Then the
2182 generated <filename>.depend</filename> file will contain
2183 the dependency:</para>
2189 <para>which says that the object file
2190 <filename>Foo.o</filename> depends on the interface file
2191 <filename>Baz.hi</filename> generated by compiling module
2192 <literal>Baz</literal>. The <filename>.depend</filename>
2193 file is automatically included by every Makefile.</para>
2198 <term><literal>binary-dist</literal></term>
2200 <para>make a binary distribution. This is the target we
2201 use to build the binary distributions of GHC and
2207 <term><literal>dist</literal></term>
2209 <para>make a source distribution. Note that this target
2210 does “make distclean” as part of its work;
2211 don't use it if you want to keep what you've built.</para>
2216 <para>Most <filename>Makefile</filename>s have targets other
2217 than these. You can discover them by looking in the
2218 <filename>Makefile</filename> itself.</para>
2222 <title>Using a project from the build tree</title>
2224 <para>If you want to build GHC (say) and just use it direct from
2225 the build tree without doing <literal>make install</literal>
2226 first, you can run the in-place driver script:
2227 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2229 <para> Do <emphasis>NOT</emphasis> use
2230 <filename>ghc/compiler/ghc</filename>, or
2231 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2232 scripts intended for installation, and contain hard-wired paths
2233 to the installed libraries, rather than the libraries in the
2236 <para>Happy can similarly be run from the build tree, using
2237 <filename>happy/src/happy-inplace</filename>.</para>
2241 <title>Fast Making</title>
2243 <indexterm><primary>fastmake</primary></indexterm>
2244 <indexterm><primary>dependencies, omitting</primary></indexterm>
2245 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2247 <para>Sometimes the dependencies get in the way: if you've made
2248 a small change to one file, and you're absolutely sure that it
2249 won't affect anything else, but you know that
2250 <command>make</command> is going to rebuild everything anyway,
2251 the following hack may be useful:</para>
2257 <para>This tells the make system to ignore dependencies and just
2258 build what you tell it to. In other words, it's equivalent to
2259 temporarily removing the <filename>.depend</filename> file in
2260 the current directory (where <command>mkdependHS</command> and
2261 friends store their dependency information).</para>
2263 <para>A bit of history: GHC used to come with a
2264 <command>fastmake</command> script that did the above job, but
2265 GNU make provides the features we need to do it without
2266 resorting to a script. Also, we've found that fastmaking is
2267 less useful since the advent of GHC's recompilation checker (see
2268 the User's Guide section on "Separate Compilation").</para>
2272 <sect1 id="sec-makefile-arch">
2273 <title>The <filename>Makefile</filename> architecture</title>
2274 <indexterm><primary>makefile architecture</primary></indexterm>
2276 <para><command>make</command> is great if everything
2277 works—you type <command>gmake install</command> and lo! the
2278 right things get compiled and installed in the right places. Our
2279 goal is to make this happen often, but somehow it often doesn't;
2280 instead some weird error message eventually emerges from the
2281 bowels of a directory you didn't know existed.</para>
2283 <para>The purpose of this section is to give you a road-map to
2284 help you figure out what is going right and what is going
2288 <title>Debugging</title>
2290 <para>Debugging <filename>Makefile</filename>s is something of a
2291 black art, but here's a couple of tricks that we find
2292 particularly useful. The following command allows you to see
2293 the contents of any make variable in the context of the current
2294 <filename>Makefile</filename>:</para>
2296 <screen>$ make show VALUE=HS_SRCS</screen>
2298 <para>where you can replace <literal>HS_SRCS</literal> with the
2299 name of any variable you wish to see the value of.</para>
2301 <para>GNU make has a <option>-d</option> option which generates
2302 a dump of the decision procedure used to arrive at a conclusion
2303 about which files should be recompiled. Sometimes useful for
2304 tracking down problems with superfluous or missing
2305 recompilations.</para>
2309 <title>A small project</title>
2311 <para>To get started, let us look at the
2312 <filename>Makefile</filename> for an imaginary small
2313 <literal>fptools</literal> project, <literal>small</literal>.
2314 Each project in <literal>fptools</literal> has its own directory
2315 in <constant>FPTOOLS_TOP</constant>, so the
2316 <literal>small</literal> project will have its own directory
2317 <constant>FPOOLS_TOP/small/</constant>. Inside the
2318 <filename>small/</filename> directory there will be a
2319 <filename>Makefile</filename>, looking something like
2322 <indexterm><primary>Makefile, minimal</primary></indexterm>
2325 # Makefile for fptools project "small"
2328 include $(TOP)/mk/boilerplate.mk
2330 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2333 include $(TOP)/target.mk
2336 <para>this <filename>Makefile</filename> has three
2341 <para>The first section includes
2344 One of the most important
2345 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2346 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2351 a file of “boilerplate” code from the level
2352 above (which in this case will be
2353 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2354 As its name suggests, <filename>boilerplate.mk</filename>
2355 consists of a large quantity of standard
2356 <filename>Makefile</filename> code. We discuss this
2357 boilerplate in more detail in <XRef LinkEnd="sec-boiler">.
2358 <indexterm><primary>include, directive in
2359 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2360 inclusion</primary></indexterm></para>
2362 <para>Before the <literal>include</literal> statement, you
2363 must define the <command>make</command> variable
2364 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2365 to be the directory containing the <filename>mk</filename>
2366 directory in which the <filename>boilerplate.mk</filename>
2367 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2370 include ../mk/boilerplate.mk # NO NO NO
2374 <para>Why? Because the <filename>boilerplate.mk</filename>
2375 file needs to know where it is, so that it can, in turn,
2376 <literal>include</literal> other files. (Unfortunately,
2377 when an <literal>include</literal>d file does an
2378 <literal>include</literal>, the filename is treated relative
2379 to the directory in which <command>gmake</command> is being
2380 run, not the directory in which the
2381 <literal>include</literal>d sits.) In general,
2382 <emphasis>every file <filename>foo.mk</filename> assumes
2384 <filename><constant>$(TOP)</constant>/mk/foo.mk</filename>
2385 refers to itself.</emphasis> It is up to the
2386 <filename>Makefile</filename> doing the
2387 <literal>include</literal> to ensure this is the case.</para>
2389 <para>Files intended for inclusion in other
2390 <filename>Makefile</filename>s are written to have the
2391 following property: <emphasis>after
2392 <filename>foo.mk</filename> is <literal>include</literal>d,
2393 it leaves <constant>TOP</constant> containing the same value
2394 as it had just before the <literal>include</literal>
2395 statement</emphasis>. In our example, this invariant
2396 guarantees that the <literal>include</literal> for
2397 <filename>target.mk</filename> will look in the same
2398 directory as that for <filename>boilerplate.mk</filename>.</para>
2402 <para> The second section defines the following standard
2403 <command>make</command> variables:
2404 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2405 (the source files from which is to be built), and
2406 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2407 (the executable binary to be built). We will discuss in
2408 more detail what the “standard variables” are,
2409 and how they affect what happens, in <XRef
2410 LinkEnd="sec-targets">.</para>
2412 <para>The definition for <constant>SRCS</constant> uses the
2413 useful GNU <command>make</command> construct
2414 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2415 which expands to a list of all the files matching the
2416 pattern <literal>pat</literal> in the current directory. In
2417 this example, <constant>SRCS</constant> is set to the list
2418 of all the <filename>.lhs</filename> and
2419 <filename>.c</filename> files in the directory. (Let's
2420 suppose there is one of each, <filename>Foo.lhs</filename>
2421 and <filename>Baz.c</filename>.)</para>
2425 <para>The last section includes a second file of standard
2427 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2428 It contains the rules that tell <command>gmake</command> how
2429 to make the standard targets (<Xref
2430 LinkEnd="sec-standard-targets">). Why, you ask, can't this
2431 standard code be part of
2432 <filename>boilerplate.mk</filename>? Good question. We
2433 discuss the reason later, in <Xref
2434 LinkEnd="sec-boiler-arch">.</para>
2436 <para>You do not <emphasis>have</emphasis> to
2437 <literal>include</literal> the
2438 <filename>target.mk</filename> file. Instead, you can write
2439 rules of your own for all the standard targets. Usually,
2440 though, you will find quite a big payoff from using the
2441 canned rules in <filename>target.mk</filename>; the price
2442 tag is that you have to understand what canned rules get
2443 enabled, and what they do (<Xref
2444 LinkEnd="sec-targets">).</para>
2448 <para>In our example <filename>Makefile</filename>, most of the
2449 work is done by the two <literal>include</literal>d files. When
2450 you say <command>gmake all</command>, the following things
2455 <para><command>gmake</command> figures out that the object
2456 files are <filename>Foo.o</filename> and
2457 <filename>Baz.o</filename>.</para>
2461 <para>It uses a boilerplate pattern rule to compile
2462 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2463 using a Haskell compiler. (Which one? That is set in the
2464 build configuration.)</para>
2468 <para>It uses another standard pattern rule to compile
2469 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2470 using a C compiler. (Ditto.)</para>
2474 <para>It links the resulting <filename>.o</filename> files
2475 together to make <literal>small</literal>, using the Haskell
2476 compiler to do the link step. (Why not use
2477 <command>ld</command>? Because the Haskell compiler knows
2478 what standard libraries to link in. How did
2479 <command>gmake</command> know to use the Haskell compiler to
2480 do the link, rather than the C compiler? Because we set the
2481 variable <constant>HS_PROG</constant> rather than
2482 <constant>C_PROG</constant>.)</para>
2486 <para>All <filename>Makefile</filename>s should follow the above
2487 three-section format.</para>
2491 <title>A larger project</title>
2493 <para>Larger projects are usually structured into a number of
2494 sub-directories, each of which has its own
2495 <filename>Makefile</filename>. (In very large projects, this
2496 sub-structure might be iterated recursively, though that is
2497 rare.) To give you the idea, here's part of the directory
2498 structure for the (rather large) GHC project:</para>
2508 ...source files for documentation...
2511 ...source files for driver...
2514 parser/...source files for parser...
2515 renamer/...source files for renamer...
2519 <para>The sub-directories <filename>docs</filename>,
2520 <filename>driver</filename>, <filename>compiler</filename>, and
2521 so on, each contains a sub-component of GHC, and each has its
2522 own <filename>Makefile</filename>. There must also be a
2523 <filename>Makefile</filename> in
2524 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>.
2525 It does most of its work by recursively invoking
2526 <command>gmake</command> on the <filename>Makefile</filename>s
2527 in the sub-directories. We say that
2528 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2529 <filename>Makefile</filename></emphasis>, because it does little
2530 except organise its children, while the
2531 <filename>Makefile</filename>s in the sub-directories are all
2532 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2533 principle the sub-directories might themselves contain a
2534 non-leaf <filename>Makefile</filename> and several
2535 sub-sub-directories, but that does not happen in GHC.)</para>
2537 <para>The <filename>Makefile</filename> in
2538 <filename>ghc/compiler</filename> is considered a leaf
2539 <filename>Makefile</filename> even though the
2540 <filename>ghc/compiler</filename> has sub-directories, because
2541 these sub-directories do not themselves have
2542 <filename>Makefile</filename>s in them. They are just used to
2543 structure the collection of modules that make up GHC, but all
2544 are managed by the single <filename>Makefile</filename> in
2545 <filename>ghc/compiler</filename>.</para>
2547 <para>You will notice that <filename>ghc/</filename> also
2548 contains a directory <filename>ghc/mk/</filename>. It contains
2549 GHC-specific <filename>Makefile</filename> boilerplate code.
2550 More precisely:</para>
2554 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2555 at the top of <filename>ghc/Makefile</filename>, and of all
2556 the leaf <filename>Makefile</filename>s in the
2557 sub-directories. It in turn <literal>include</literal>s the
2558 main boilerplate file
2559 <filename>mk/boilerplate.mk</filename>.</para>
2563 <para><filename>ghc/mk/target.mk</filename> is
2564 <literal>include</literal>d at the bottom of
2565 <filename>ghc/Makefile</filename>, and of all the leaf
2566 <filename>Makefile</filename>s in the sub-directories. It
2567 in turn <literal>include</literal>s the file
2568 <filename>mk/target.mk</filename>.</para>
2572 <para>So these two files are the place to look for GHC-wide
2573 customisation of the standard boilerplate.</para>
2576 <sect2 id="sec-boiler-arch">
2577 <title>Boilerplate architecture</title>
2578 <indexterm><primary>boilerplate architecture</primary></indexterm>
2580 <para>Every <filename>Makefile</filename> includes a
2581 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2582 file at the top, and
2583 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2584 file at the bottom. In this section we discuss what is in these
2585 files, and why there have to be two of them. In general:</para>
2589 <para><filename>boilerplate.mk</filename> consists of:</para>
2593 <para><emphasis>Definitions of millions of
2594 <command>make</command> variables</emphasis> that
2595 collectively specify the build configuration. Examples:
2596 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2597 the options to feed to the Haskell compiler;
2598 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2599 the sub-directories to enable within the
2600 <literal>nofib</literal> project;
2601 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2602 the name of the Haskell compiler to use when compiling
2603 GHC in the <literal>ghc</literal> project.</para>
2607 <para><emphasis>Standard pattern rules</emphasis> that
2608 tell <command>gmake</command> how to construct one file
2609 from another.</para>
2613 <para><filename>boilerplate.mk</filename> needs to be
2614 <literal>include</literal>d at the <emphasis>top</emphasis>
2615 of each <filename>Makefile</filename>, so that the user can
2616 replace the boilerplate definitions or pattern rules by
2617 simply giving a new definition or pattern rule in the
2618 <filename>Makefile</filename>. <command>gmake</command>
2619 simply takes the last definition as the definitive one.</para>
2621 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2622 definitions, it is also quite common to
2623 <emphasis>augment</emphasis> them. For example, a
2624 <filename>Makefile</filename> might say:</para>
2630 <para>thereby adding “<option>-O</option>” to
2632 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2636 <para><filename>target.mk</filename> contains
2637 <command>make</command> rules for the standard targets
2638 described in <Xref LinkEnd="sec-standard-targets">. These
2639 rules are selectively included, depending on the setting of
2640 certain <command>make</command> variables. These variables
2641 are usually set in the middle section of the
2642 <filename>Makefile</filename> between the two
2643 <literal>include</literal>s.</para>
2645 <para><filename>target.mk</filename> must be included at the
2646 end (rather than being part of
2647 <filename>boilerplate.mk</filename>) for several tiresome
2653 <para><command>gmake</command> commits target and
2654 dependency lists earlier than it should. For example,
2655 <FIlename>target.mk</FIlename> has a rule that looks
2659 $(HS_PROG) : $(OBJS)
2660 $(HC) $(LD_OPTS) $< -o $@
2663 <para>If this rule was in
2664 <filename>boilerplate.mk</filename> then
2665 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2667 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2668 would not have their final values at the moment
2669 <command>gmake</command> encountered the rule. Alas,
2670 <command>gmake</command> takes a snapshot of their
2671 current values, and wires that snapshot into the rule.
2672 (In contrast, the commands executed when the rule
2673 “fires” are only substituted at the moment
2674 of firing.) So, the rule must follow the definitions
2675 given in the <filename>Makefile</filename> itself.</para>
2679 <para>Unlike pattern rules, ordinary rules cannot be
2680 overriden or replaced by subsequent rules for the same
2681 target (at least, not without an error message).
2682 Including ordinary rules in
2683 <filename>boilerplate.mk</filename> would prevent the
2684 user from writing rules for specific targets in specific
2689 <para>There are a couple of other reasons I've
2690 forgotten, but it doesn't matter too much.</para>
2697 <sect2 id="sec-boiler">
2698 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2699 <indexterm><primary>boilerplate.mk</primary></indexterm>
2701 <para>If you look at
2702 <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename>
2703 you will find that it consists of the following sections, each
2704 held in a separate file:</para>
2708 <term><filename>config.mk</filename></term>
2709 <indexterm><primary>config.mk</primary></indexterm>
2711 <para>is the build configuration file we discussed at
2712 length in <Xref LinkEnd="sec-build-config">.</para>
2717 <term><filename>paths.mk</filename></term>
2718 <indexterm><primary>paths.mk</primary></indexterm>
2720 <para>defines <command>make</command> variables for
2721 pathnames and file lists. This file contains code for
2722 automatically compiling lists of source files and deriving
2723 lists of object files from those. The results can be
2724 overriden in the <filename>Makefile</filename>, but in
2725 most cases the automatic setup should do the right
2728 <para>The following variables may be set in the
2729 <filename>Makefile</filename> to affect how the automatic
2730 source file search is done:</para>
2734 <term><literal>ALL_DIRS</literal></term>
2735 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2738 <para>Set to a list of directories to search in
2739 addition to the current directory for source
2745 <term><literal>EXCLUDE_SRCS</literal></term>
2746 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2749 <para>Set to a list of source files (relative to the
2750 current directory) to omit from the automatic
2751 search. The source searching machinery is clever
2752 enough to know that if you exclude a source file
2753 from which other sources are derived, then the
2754 derived sources should also be excluded. For
2755 example, if you set <literal>EXCLUDED_SRCS</literal>
2756 to include <filename>Foo.y</filename>, then
2757 <filename>Foo.hs</filename> will also be
2763 <term><literal>EXTRA_SRCS</literal></term>
2764 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2767 <para>Set to a list of extra source files (perhaps
2768 in directories not listed in
2769 <literal>ALL_DIRS</literal>) that should be
2775 <para>The results of the automatic source file search are
2776 placed in the following make variables:</para>
2780 <term><literal>SRCS</literal></term>
2781 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2783 <para>All source files found, sorted and without
2784 duplicates, including those which might not exist
2785 yet but will be derived from other existing sources.
2786 <literal>SRCS</literal> <emphasis>can</emphasis> be
2787 overriden if necessary, in which case the variables
2788 below will follow suit.</para>
2793 <term><literal>HS_SRCS</literal></term>
2794 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2796 <para>all Haskell source files in the current
2797 directory, including those derived from other source
2798 files (eg. Happy sources also give rise to Haskell
2804 <term><literal>HS_OBJS</literal></term>
2805 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2807 <para>Object files derived from
2808 <literal>HS_SRCS</literal>.</para>
2813 <term><literal>HS_IFACES</literal></term>
2814 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2816 <para>Interface files (<literal>.hi</literal> files)
2817 derived from <literal>HS_SRCS</literal>.</para>
2822 <term><literal>C_SRCS</literal></term>
2823 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2825 <para>All C source files found.</para>
2830 <term><literal>C_OBJS</literal></term>
2831 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2833 <para>Object files derived from
2834 <literal>C_SRCS</literal>.</para>
2839 <term><literal>SCRIPT_SRCS</literal></term>
2840 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2842 <para>All script source files found
2843 (<literal>.lprl</literal> files).</para>
2848 <term><literal>SCRIPT_OBJS</literal></term>
2849 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2851 <para><quote>object</quote> files derived from
2852 <literal>SCRIPT_SRCS</literal>
2853 (<literal>.prl</literal> files).</para>
2858 <term><literal>HSC_SRCS</literal></term>
2859 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2861 <para>All <literal>hsc2hs</literal> source files
2862 (<literal>.hsc</literal> files).</para>
2867 <term><literal>HAPPY_SRCS</literal></term>
2868 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2870 <para>All <literal>happy</literal> source files
2871 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2876 <term><literal>OBJS</literal></term>
2877 <indexterm><primary>OBJS</primary></indexterm>
2879 <para>the concatenation of
2880 <literal>$(HS_OBJS)</literal>,
2881 <literal>$(C_OBJS)</literal>, and
2882 <literal>$(SCRIPT_OBJS)</literal>.</para>
2887 <para>Any or all of these definitions can easily be
2888 overriden by giving new definitions in your
2889 <filename>Makefile</filename>.</para>
2891 <para>What, exactly, does <filename>paths.mk</filename>
2892 consider a <quote>source file</quote> to be? It's based
2893 on the file's suffix (e.g. <filename>.hs</filename>,
2894 <filename>.lhs</filename>, <filename>.c</filename>,
2895 <filename>.hy</filename>, etc), but this is the kind of
2896 detail that changes, so rather than enumerate the source
2897 suffices here the best thing to do is to look in
2898 <filename>paths.mk</filename>.</para>
2903 <term><filename>opts.mk</filename></term>
2904 <indexterm><primary>opts.mk</primary></indexterm>
2906 <para>defines <command>make</command> variables for option
2907 strings to pass to each program. For example, it defines
2908 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2909 the option strings to pass to the Haskell compiler. See
2910 <Xref LinkEnd="sec-suffix">.</para>
2915 <term><filename>suffix.mk</filename></term>
2916 <indexterm><primary>suffix.mk</primary></indexterm>
2918 <para>defines standard pattern rules—see <Xref
2919 LinkEnd="sec-suffix">.</para>
2924 <para>Any of the variables and pattern rules defined by the
2925 boilerplate file can easily be overridden in any particular
2926 <filename>Makefile</filename>, because the boilerplate
2927 <literal>include</literal> comes first. Definitions after this
2928 <literal>include</literal> directive simply override the default
2929 ones in <filename>boilerplate.mk</filename>.</para>
2932 <sect2 id="sec-suffix">
2933 <title>Pattern rules and options</title>
2934 <indexterm><primary>Pattern rules</primary></indexterm>
2937 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2938 defines standard <emphasis>pattern rules</emphasis> that say how
2939 to build one kind of file from another, for example, how to
2940 build a <filename>.o</filename> file from a
2941 <filename>.c</filename> file. (GNU <command>make</command>'s
2942 <emphasis>pattern rules</emphasis> are more powerful and easier
2943 to use than Unix <command>make</command>'s <emphasis>suffix
2944 rules</emphasis>.)</para>
2946 <para>Almost all the rules look something like this:</para>
2951 $(CC) $(CC_OPTS) -c $< -o $@
2954 <para>Here's how to understand the rule. It says that
2955 <emphasis>something</emphasis><filename>.o</filename> (say
2956 <filename>Foo.o</filename>) can be built from
2957 <emphasis>something</emphasis><filename>.c</filename>
2958 (<filename>Foo.c</filename>), by invoking the C compiler (path
2959 name held in <constant>$(CC)</constant>), passing to it
2960 the options <constant>$(CC_OPTS)</constant> and
2961 the rule's dependent file of the rule
2962 <literal>$<</literal> (<filename>Foo.c</filename> in
2963 this case), and putting the result in the rule's target
2964 <literal>$@</literal> (<filename>Foo.o</filename> in this
2967 <para>Every program is held in a <command>make</command>
2968 variable defined in <filename>mk/config.mk</filename>—look
2969 in <filename>mk/config.mk</filename> for the complete list. One
2970 important one is the Haskell compiler, which is called
2971 <constant>$(HC)</constant>.</para>
2973 <para>Every program's options are are held in a
2974 <command>make</command> variables called
2975 <constant><prog>_OPTS</constant>. the
2976 <constant><prog>_OPTS</constant> variables are
2977 defined in <filename>mk/opts.mk</filename>. Almost all of them
2978 are defined like this:</para>
2981 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
2984 <para>The four variables from which
2985 <constant>CC_OPTS</constant> is built have the following
2990 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
2992 <para>options passed to all C compilations.</para>
2997 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
2999 <para>options passed to C compilations for way
3000 <literal><way></literal>. For example,
3001 <constant>WAY_mp_CC_OPTS</constant>
3002 gives options to pass to the C compiler when compiling way
3003 <literal>mp</literal>. The variable
3004 <constant>WAY_CC_OPTS</constant> holds
3005 options to pass to the C compiler when compiling the
3006 standard way. (<Xref LinkEnd="sec-ways"> dicusses
3007 multi-way compilation.)</para>
3012 <term><constant><module>_CC_OPTS</constant>:</term>
3014 <para>options to pass to the C compiler that are specific
3015 to module <literal><module></literal>. For example,
3016 <constant>SMap_CC_OPTS</constant> gives the
3017 specific options to pass to the C compiler when compiling
3018 <filename>SMap.c</filename>.</para>
3023 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3025 <para>extra options to pass to all C compilations. This
3026 is intended for command line use, thus:</para>
3029 gmake libHS.a EXTRA_CC_OPTS="-v"
3036 <sect2 id="sec-targets">
3037 <title>The main <filename>mk/target.mk</filename> file</title>
3038 <indexterm><primary>target.mk</primary></indexterm>
3040 <para><filename>target.mk</filename> contains canned rules for
3041 all the standard targets described in <Xref
3042 LinkEnd="sec-standard-targets">. It is complicated by the fact
3043 that you don't want all of these rules to be active in every
3044 <filename>Makefile</filename>. Rather than have a plethora of
3045 tiny files which you can include selectively, there is a single
3046 file, <filename>target.mk</filename>, which selectively includes
3047 rules based on whether you have defined certain variables in
3048 your <filename>Makefile</filename>. This section explains what
3049 rules you get, what variables control them, and what the rules
3050 do. Hopefully, you will also get enough of an idea of what is
3051 supposed to happen that you can read and understand any weird
3052 special cases yourself.</para>
3056 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3058 <para>If <constant>HS_PROG</constant> is defined,
3059 you get rules with the following targets:</para>
3063 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3065 <para>itself. This rule links
3066 <constant>$(OBJS)</constant> with the Haskell
3067 runtime system to get an executable called
3068 <constant>$(HS_PROG)</constant>.</para>
3073 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3076 <constant>$(HS_PROG)</constant> in
3077 <constant>$(bindir)</constant>.</para>
3086 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3088 <para>is similar to <constant>HS_PROG</constant>,
3089 except that the link step links
3090 <constant>$(C_OBJS)</constant> with the C
3091 runtime system.</para>
3096 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3098 <para>is similar to <constant>HS_PROG</constant>,
3099 except that it links
3100 <constant>$(LIB_OBJS)</constant> to make the
3101 library archive <constant>$(LIBRARY)</constant>,
3102 and <literal>install</literal> installs it in
3103 <constant>$(libdir)</constant>.</para>
3108 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3110 <para>…</para>
3115 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3117 <para>…</para>
3122 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3124 <para>If <constant>HS_SRCS</constant> is defined
3125 and non-empty, a rule for the target
3126 <literal>depend</literal> is included, which generates
3127 dependency information for Haskell programs. Similarly
3128 for <constant>C_SRCS</constant>.</para>
3133 <para>All of these rules are “double-colon” rules,
3137 install :: $(HS_PROG)
3138 ...how to install it...
3141 <para>GNU <command>make</command> treats double-colon rules as
3142 separate entities. If there are several double-colon rules for
3143 the same target it takes each in turn and fires it if its
3144 dependencies say to do so. This means that you can, for
3145 example, define both <constant>HS_PROG</constant> and
3146 <constant>LIBRARY</constant>, which will generate two rules for
3147 <literal>install</literal>. When you type <command>gmake
3148 install</command> both rules will be fired, and both the program
3149 and the library will be installed, just as you wanted.</para>
3152 <sect2 id="sec-subdirs">
3153 <title>Recursion</title>
3154 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3155 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3157 <para>In leaf <filename>Makefile</filename>s the variable
3158 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3159 is undefined. In non-leaf <filename>Makefile</filename>s,
3160 <constant>SUBDIRS</constant> is set to the list of
3161 sub-directories that contain subordinate
3162 <filename>Makefile</filename>s. <emphasis>It is up to you to
3163 set <constant>SUBDIRS</constant> in the
3164 <filename>Makefile</filename>.</emphasis> There is no automation
3165 here—<constant>SUBDIRS</constant> is too important to
3168 <para>When <constant>SUBDIRS</constant> is defined,
3169 <filename>target.mk</filename> includes a rather neat rule for
3170 the standard targets (<Xref LinkEnd="sec-standard-targets"> that
3171 simply invokes <command>make</command> recursively in each of
3172 the sub-directories.</para>
3174 <para><emphasis>These recursive invocations are guaranteed to
3175 occur in the order in which the list of directories is specified
3176 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3177 be important. For example, when you say <command>gmake
3178 boot</command> it can be important that the recursive invocation
3179 of <command>make boot</command> is done in one sub-directory
3180 (the include files, say) before another (the source files).
3181 Generally, put the most independent sub-directory first, and the
3182 most dependent last.</para>
3185 <sect2 id="sec-ways">
3186 <title>Way management</title>
3187 <indexterm><primary>way management</primary></indexterm>
3189 <para>We sometimes want to build essentially the same system in
3190 several different “ways”. For example, we want to build GHC's
3191 <literal>Prelude</literal> libraries with and without profiling,
3192 so that there is an appropriately-built library archive to link
3193 with when the user compiles his program. It would be possible
3194 to have a completely separate build tree for each such “way”,
3195 but it would be horribly bureaucratic, especially since often
3196 only parts of the build tree need to be constructed in multiple
3200 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3201 contains some clever magic to allow you to build several
3202 versions of a system; and to control locally how many versions
3203 are built and how they differ. This section explains the
3206 <para>The files for a particular way are distinguished by
3207 munging the suffix. The <quote>normal way</quote> is always
3208 built, and its files have the standard suffices
3209 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3210 In addition, you can build one or more extra ways, each
3211 distinguished by a <emphasis>way tag</emphasis>. The object
3212 files and interface files for one of these extra ways are
3213 distinguished by their suffix. For example, way
3214 <literal>mp</literal> has files
3215 <filename>.mp_o</filename> and
3216 <filename>.mp_hi</filename>. Library archives have their
3217 way tag the other side of the dot, for boring reasons; thus,
3218 <filename>libHS_mp.a</filename>.</para>
3220 <para>A <command>make</command> variable called
3221 <constant>way</constant> holds the current way tag.
3222 <emphasis><constant>way</constant> is only ever set on the
3223 command line of <command>gmake</command></emphasis> (usually in
3224 a recursive invocation of <command>gmake</command> by the
3225 system). It is never set inside a
3226 <filename>Makefile</filename>. So it is a global constant for
3227 any one invocation of <command>gmake</command>. Two other
3228 <command>make</command> variables,
3229 <constant>way_</constant> and
3230 <constant>_way</constant> are immediately derived from
3231 <constant>$(way)</constant> and never altered. If
3232 <constant>way</constant> is not set, then neither are
3233 <constant>way_</constant> and
3234 <constant>_way</constant>, and the invocation of
3235 <command>make</command> will build the <quote>normal
3236 way</quote>. If <constant>way</constant> is set, then the other
3237 two variables are set in sympathy. For example, if
3238 <constant>$(way)</constant> is “<literal>mp</literal>”,
3239 then <constant>way_</constant> is set to
3240 “<literal>mp_</literal>” and
3241 <constant>_way</constant> is set to
3242 “<literal>_mp</literal>”. These three variables are
3243 then used when constructing file names.</para>
3245 <para>So how does <command>make</command> ever get recursively
3246 invoked with <constant>way</constant> set? There are two ways
3247 in which this happens:</para>
3251 <para>For some (but not all) of the standard targets, when
3252 in a leaf sub-directory, <command>make</command> is
3253 recursively invoked for each way tag in
3254 <constant>$(WAYS)</constant>. You set
3255 <constant>WAYS</constant> in the
3256 <filename>Makefile</filename> to the list of way tags you
3257 want these targets built for. The mechanism here is very
3258 much like the recursive invocation of
3259 <command>make</command> in sub-directories (<Xref
3260 LinkEnd="sec-subdirs">). It is up to you to set
3261 <constant>WAYS</constant> in your
3262 <filename>Makefile</filename>; this is how you control what
3263 ways will get built.</para>
3267 <para>For a useful collection of targets (such as
3268 <filename>libHS_mp.a</filename>,
3269 <filename>Foo.mp_o</filename>) there is a rule which
3270 recursively invokes <command>make</command> to make the
3271 specified target, setting the <constant>way</constant>
3272 variable. So if you say <command>gmake
3273 Foo.mp_o</command> you should see a recursive
3274 invocation <command>gmake Foo.mp_o way=mp</command>,
3275 and <emphasis>in this recursive invocation the pattern rule
3276 for compiling a Haskell file into a <filename>.o</filename>
3277 file will match</emphasis>. The key pattern rules (in
3278 <filename>suffix.mk</filename>) look like this:
3282 $(HC) $(HC_OPTS) $< -o $@
3289 <para>You can invoke <command>make</command> with a
3290 particular <literal>way</literal> setting yourself, in order
3291 to build files related to a particular
3292 <literal>way</literal> in the current directory. eg.
3298 will build files for the profiling way only in the current
3305 <title>When the canned rule isn't right</title>
3307 <para>Sometimes the canned rule just doesn't do the right thing.
3308 For example, in the <literal>nofib</literal> suite we want the
3309 link step to print out timing information. The thing to do here
3310 is <emphasis>not</emphasis> to define
3311 <constant>HS_PROG</constant> or
3312 <constant>C_PROG</constant>, and instead define a special
3313 purpose rule in your own <filename>Makefile</filename>. By
3314 using different variable names you will avoid the canned rules
3315 being included, and conflicting with yours.</para>
3319 <sect1 id="building-docs">
3320 <title>Building the documentation</title>
3322 <sect2 id="pre-supposed-doc-tools">
3323 <title>Tools for building the Documentation</title>
3325 <para>The following additional tools are required if you want to
3326 format the documentation that comes with the
3327 <literal>fptools</literal> projects:</para>
3331 <term>DocBook</term>
3332 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3333 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3335 <para>Much of our documentation is written in SGML, using
3336 the DocBook DTD. Instructions on installing and
3337 configuring the DocBook tools are below.</para>
3343 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3344 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3346 <para>A decent TeX distribution is required if you want to
3347 produce printable documentation. We recomment teTeX,
3348 which includes just about everything you need.</para>
3353 <term>Haddock</term>
3354 <indexterm><primary>Haddock</primary>
3357 <para>Haddock is a Haskell documentation tool that we use
3358 for automatically generating documentation from the
3359 library source code. It is an <literal>fptools</literal>
3360 project in itself. To build documentation for the
3361 libraries (<literal>fptools/libraries</literal>) you
3362 should check out and build Haddock in
3363 <literal>fptools/haddock</literal>. Haddock requires GHC
3371 <title>Installing the DocBook tools</title>
3374 <title>Installing the DocBook tools on Linux</title>
3376 <para>If you're on a recent RedHat system (7.0+), you probably
3377 have working DocBook tools already installed. The configure
3378 script should detect your setup and you're away.</para>
3380 <para>If you don't have DocBook tools installed, and you are
3381 using a system that can handle RedHat RPM packages, you can
3382 probably use the <ULink
3383 URL="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3384 DocBook tools</ULink>, which is the most shrink-wrapped SGML
3385 suite that we could find. You need all the RPMs except for
3386 psgml (i.e. <Filename>docbook</Filename>,
3387 <Filename>jade</Filename>, <Filename>jadetex</Filename>,
3388 <Filename>sgmlcommon</Filename> and
3389 <Filename>stylesheets</Filename>). Note that most of these
3390 RPMs are architecture neutral, so are likely to be found in a
3391 <Filename>noarch</Filename> directory. The SuSE RPMs also
3392 work; the RedHat ones <Emphasis>don't</Emphasis> in RedHat 6.2
3393 (7.0 and later should be OK), but they are easy to fix: just
3395 <Filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</Filename>
3396 to <Filename>/usr/lib/sgml/lib/dblib.dsl</Filename>. </para>
3400 <title>Installing DocBook on FreeBSD</title>
3402 <para>On FreeBSD systems, the easiest way to get DocBook up
3403 and running is to install it from the ports tree or a
3404 pre-compiled package (packages are available from your local
3405 FreeBSD mirror site).</para>
3407 <para>To use the ports tree, do this:
3409 $ cd /usr/ports/textproc/docproj
3412 This installs the FreeBSD documentation project tools, which
3413 includes everything needed to format the GHC
3414 documentation.</para>
3418 <title>Installing from binaries on Windows</title>
3420 <Para>It's a good idea to use Norman Walsh's <ULink
3421 URL="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3422 notes</ULink> as a guide. You should get version 3.1 of
3423 DocBook, and note that his file <Filename>test.sgm</Filename>
3424 won't work, as it needs version 3.0. You should unpack Jade
3425 into <Filename>\Jade</Filename>, along with the entities,
3426 DocBook into <Filename>\docbook</Filename>, and the DocBook
3427 stylesheets into <Filename>\docbook\stylesheets</Filename> (so
3428 they actually end up in
3429 <Filename>\docbook\stylesheets\docbook</Filename>).</para>
3434 <title>Installing the DocBook tools from source</title>
3439 <para>Install <ULink
3440 URL="http://openjade.sourceforge.net/">OpenJade</ULink>
3441 (Windows binaries are available as well as sources). If you
3442 want DVI, PS, or PDF then install JadeTeX from the
3443 <Filename>dsssl</Filename> subdirectory. (If you get the
3447 ! LaTeX Error: Unknown option implicit=false' for package hyperref'.
3450 your version of <Command>hyperref</Command> is out of date;
3451 download it from CTAN
3452 (<Filename>macros/latex/contrib/supported/hyperref</Filename>),
3453 and make it, ensuring that you have first removed or renamed
3454 your old copy. If you start getting file not found errors
3455 when making the test for <Command>hyperref</Command>, you
3456 can abort at that point and proceed straight to
3457 <Command>make install</Command>, or enter them as
3458 <Filename>../</Filename><Emphasis>filename</Emphasis>.)</para>
3460 <para>Make links from <Filename>virtex</Filename> to
3461 <Filename>jadetex</Filename> and
3462 <Filename>pdfvirtex</Filename> to
3463 <Filename>pdfjadetex</Filename> (otherwise DVI, PostScript
3464 and PDF output will not work). Copy
3465 <Filename>dsssl/*.{dtd,dsl}</Filename> and
3466 <Filename>catalog</Filename> to
3467 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3471 <title>DocBook and the DocBook stylesheets</title>
3473 <para>Get a Zip of <ULink
3474 URL="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ULink>
3475 and install the contents in
3476 <Filename>/usr/[local/]/lib/sgml</Filename>.</para>
3478 <para>Get the <ULink
3479 URL="http://nwalsh.com/docbook/dsssl/">DocBook
3480 stylesheets</ULink> and install in
3481 <Filename>/usr/[local/]lib/sgml/stylesheets</Filename>
3482 (thereby creating a subdirectory docbook). For indexing,
3483 copy or link <Filename>collateindex.pl</Filename> from the
3484 DocBook stylesheets archive in <Filename>bin</Filename> into
3485 a directory on your <Constant>PATH</Constant>.</para>
3487 <para>Download the <ULink
3488 URL="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3489 entities</ULink> into
3490 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3496 <title>Configuring the DocBook tools</title>
3498 <Para>Once the DocBook tools are installed, the configure script
3499 will detect them and set up the build system accordingly. If you
3500 have a system that isn't supported, let us know, and we'll try
3505 <title>Remaining problems</title>
3507 <para>If you install from source, you'll get a pile of warnings
3510 <Screen>DTDDECL catalog entries are not supported</Screen>
3512 every time you build anything. These can safely be ignored, but
3513 if you find them tedious you can get rid of them by removing all
3514 the <Constant>DTDDECL</Constant> entries from
3515 <Filename>docbook.cat</Filename>.</para>
3519 <title>Building the documentation</title>
3521 <para>To build documentation in a certain format, you can
3522 say, for example,</para>
3528 <para>to build HTML documentation below the current directory.
3529 The available formats are: <literal>dvi</literal>,
3530 <literal>ps</literal>, <literal>pdf</literal>,
3531 <literal>html</literal>, and <literal>rtf</literal>. Note that
3532 not all documentation can be built in all of these formats: HTML
3533 documentation is generally supported everywhere, and DocBook
3534 documentation might support the other formats (depending on what
3535 other tools you have installed).</para>
3537 <para>All of these targets are recursive; that is, saying
3538 <literal>make html</literal> will make HTML docs for all the
3539 documents recursively below the current directory.</para>
3541 <para>Because there are many different formats that the DocBook
3542 documentation can be generated in, you have to select which ones
3543 you want by setting the <literal>SGMLDocWays</literal> variable
3544 to a list of them. For example, in
3545 <filename>build.mk</filename> you might have a line:</para>
3548 SGMLDocWays = html ps
3551 <para>This will cause the documentation to be built in the requested
3552 formats as part of the main build (the default is not to build
3553 any documentation at all).</para>
3557 <title>Installing the documentation</title>
3559 <para>To install the documentation, use:</para>
3565 <para>This will install the documentation into
3566 <literal>$(datadir)</literal> (which defaults to
3567 <literal>$(prefix)/share</literal>). The exception is HTML
3568 documentation, which goes into
3569 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3571 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3572 to a list of formats, the <literal>install-docs</literal> target
3573 won't do anything for SGML documentation.</para>
3579 <sect1 id="sec-porting-ghc">
3580 <title>Porting GHC</title>
3582 <para>This section describes how to port GHC to a currenly
3583 unsupported platform. There are two distinct
3584 possibilities:</para>
3588 <para>The hardware architecture for your system is already
3589 supported by GHC, but you're running an OS that isn't
3590 supported (or perhaps has been supported in the past, but
3591 currently isn't). This is the easiest type of porting job,
3592 but it still requires some careful bootstrapping. Proceed to
3593 <xref linkend="sec-booting-from-hc">.</para>
3597 <para>Your system's hardware architecture isn't supported by
3598 GHC. This will be a more difficult port (though by comparison
3599 perhaps not as difficult as porting gcc). Proceed to <xref
3600 linkend="unregisterised-porting">.</para>
3604 <sect2 id="sec-booting-from-hc">
3605 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3607 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3608 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3609 <indexterm><primary>porting GHC</primary></indexterm>
3611 <para>Bootstrapping GHC on a system without GHC already
3612 installed is achieved by taking the intermediate C files (known
3613 as HC files) from a GHC compilation on a supported system to the
3614 target machine, and compiling them using gcc to get a working
3617 <para><emphasis>NOTE: GHC versions 5.xx and later are
3618 significantly harder to bootstrap from C than earlier versions.
3619 We recommend starting from version 4.08.2 if you need to
3620 bootstrap in this way.</emphasis></para>
3622 <para>HC files are architecture-dependent (but not
3623 OS-dependent), so you have to get a set that were generated on
3624 similar hardware. There may be some supplied on the GHC
3625 download page, otherwise you'll have to compile some up
3626 yourself, or start from <emphasis>unregisterised</emphasis> HC
3627 files - see <xref linkend="unregisterised-porting">.</para>
3629 <para>The following steps should result in a working GHC build
3630 with full libraries:</para>
3634 <para>Unpack the HC files on top of a fresh source tree
3635 (make sure the source tree version matches the version of
3636 the HC files <emphasis>exactly</emphasis>!). This will
3637 place matching <filename>.hc</filename> files next to the
3638 corresponding Haskell source (<filename>.hs</filename> or
3639 <filename>.lhs</filename>) in the compiler subdirectory
3640 <filename>ghc/compiler</filename> and in the libraries
3641 (subdirectories of <filename>hslibs</filename> and
3642 <literal>libraries</literal>).</para>
3646 <para>The actual build process is fully automated by the
3647 <filename>hc-build</filename> script located in the
3648 <filename>distrib</filename> directory. If you eventually
3649 want to install GHC into the directory
3650 <replaceable>dir</replaceable>, the following
3651 command will execute the whole build process (it won't
3652 install yet):</para>
3655 foo% distrib/hc-build --prefix=<replaceable>dir</replaceable>
3657 <indexterm><primary>--hc-build</primary></indexterm>
3659 <para>By default, the installation directory is
3660 <filename>/usr/local</filename>. If that is what you want,
3661 you may omit the argument to <filename>hc-build</filename>.
3662 Generally, any option given to <filename>hc-build</filename>
3663 is passed through to the configuration script
3664 <filename>configure</filename>. If
3665 <filename>hc-build</filename> successfully completes the
3666 build process, you can install the resulting system, as
3676 <sect2 id="unregisterised-porting">
3677 <title>Porting GHC to a new architecture</title>
3679 <para>The first step in porting to a new architecture is to get
3680 an <firstterm>unregisterised</firstterm> build working. An
3681 unregisterised build is one that compiles via vanilla C only.
3682 By contrast, a registerised build uses the following
3683 architecture-specific hacks for speed:</para>
3687 <para>Global register variables: certain abstract machine
3688 <quote>registers</quote> are mapped to real machine
3689 registers, depending on how many machine registers are
3691 <filename>ghc/includes/MachRegs.h</filename>).</para>
3695 <para>Assembly-mangling: when compiling via C, we feed the
3696 assembly generated by gcc though a Perl script known as the
3697 <firstterm>mangler</firstterm> (see
3698 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3699 mangler rearranges the assembly to support tail-calls and
3700 various other optimisations.</para>
3704 <para>In an unregisterised build, neither of these hacks are
3705 used — the idea is that the C code generated by the
3706 compiler should compile using gcc only. The lack of these
3707 optimisations costs about a factor of two in performance, but
3708 since unregisterised compilation is usually just a step on the
3709 way to a full registerised port, we don't mind too much.</para>
3712 <title>Building an unregisterised port</title>
3714 <para>The first step is to get some unregisterised HC files.
3715 Either (a) download them from the GHC site (if there are
3716 some available for the right version of GHC), or
3717 (b) build them yourself on any machine with a working
3718 GHC. If at all possible this should be a machine with the
3719 same word size as the target.</para>
3721 <para>There is a script available which should automate the
3722 process of doing the 2-stage bootstrap necessary to get the
3723 unregisterised HC files - it's available in <ulink
3724 url="http://cvs.haskell.org/cgi-bin/cvsweb.cgi/fptools/distrib/cross-port"><filename>fptools/distrib/cross-port</filename></ulink>
3727 <para>Now take these unregisterised HC files to the target
3728 platform and bootstrap a compiler from them as per the
3729 instructions in <xref linkend="sec-booting-from-hc">. In
3730 <filename>build.mk</filename>, you need to tell the build
3731 system that the compiler you're building is
3732 (a) unregisterised itself, and (b) builds
3733 unregisterised binaries. This varies depending on the GHC
3734 version you're bootstraping:</para>
3737 # build.mk for GHC 4.08.x
3738 GhcWithRegisterised=NO
3742 # build.mk for GHC 5.xx and 6.x
3743 GhcUnregisterised=YES
3746 <para>Versions 5.xx and 6.x only: use the option
3747 <option>--enable-hc-boot-unregisterised</option> instead of
3748 <option>--enable-hc-boot</option> when running
3749 <filename>./configure</filename>.</para>
3751 <para>The build may not go through cleanly. We've tried to
3752 stick to writing portable code in most parts of the compiler,
3753 so it should compile on any POSIXish system with gcc, but in
3754 our experience most systems differ from the standards in one
3755 way or another. Deal with any problems as they arise - if you
3756 get stuck, ask the experts on
3757 <email>glasgow-haskell-users@haskell.org</email>.</para>
3759 <para>Once you have the unregisterised compiler up and
3760 running, you can use it to start a registerised port. The
3761 following sections describe the various parts of the system
3762 that will need architecture-specific tweaks in order to get a
3763 registerised build going.</para>
3765 <para>Lots of useful information about the innards of GHC is
3766 available in the <ulink
3767 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3768 Commentary</ulink>, which might be helpful if you run into
3769 some code which needs tweaking for your system.</para>
3773 <title>Porting the RTS</title>
3775 <para>The following files need architecture-specific code for a
3776 registerised build:</para>
3780 <term><filename>ghc/includes/MachRegs.h</filename></term>
3781 <indexterm><primary><filename>MachRegs.h</filename></primary>
3784 <para>Defines the STG-register to machine-register
3785 mapping. You need to know your platform's C calling
3786 convention, and which registers are generally available
3787 for mapping to global register variables. There are
3788 plenty of useful comments in this file.</para>
3792 <term><filename>ghc/includes/TailCalls.h</filename></term>
3793 <indexterm><primary><filename>TailCalls.h</filename></primary>
3796 <para>Macros that cooperate with the mangler (see <xref
3797 linkend="sec-mangler">) to make proper tail-calls
3802 <term><filename>ghc/rts/Adjustor.c</filename></term>
3803 <indexterm><primary><filename>Adjustor.c</filename></primary>
3807 <literal>foreign import "wrapper"</literal>
3809 <literal>foreign export dynamic</literal>).
3810 Not essential for getting GHC bootstrapped, so this file
3811 can be deferred until later if necessary.</para>
3815 <term><filename>ghc/rts/StgCRun.c</filename></term>
3816 <indexterm><primary><filename>StgCRun.c</filename></primary>
3819 <para>The little assembly layer between the C world and
3820 the Haskell world. See the comments and code for the
3821 other architectures in this file for pointers.</para>
3825 <term><filename>ghc/rts/MBlock.h</filename></term>
3826 <term><filename>ghc/rts/MBlock.c</filename></term>
3827 <indexterm><primary><filename>MBlock.h</filename></primary>
3829 <indexterm><primary><filename>MBlock.c</filename></primary>
3832 <para>These files are really OS-specific rather than
3833 architecture-specific. In <filename>MBlock.h</filename>
3834 is specified the absolute location at which the RTS
3835 should try to allocate memory on your platform (try to
3836 find an area which doesn't conflict with code or dynamic
3837 libraries). In <filename>Mblock.c</filename> you might
3838 need to tweak the call to <literal>mmap()</literal> for
3845 <sect3 id="sec-mangler">
3846 <title>The mangler</title>
3848 <para>The mangler is an evil Perl-script that rearranges the
3849 assembly code output from gcc to do two main things:</para>
3853 <para>Remove function prologues and epilogues, and all
3854 movement of the C stack pointer. This is to support
3855 tail-calls: every code block in Haskell code ends in an
3856 explicit jump, so we don't want the C-stack overflowing
3857 while we're jumping around between code blocks.</para>
3860 <para>Move the <firstterm>info table</firstterm> for a
3861 closure next to the entry code for that closure. In
3862 unregisterised code, info tables contain a pointer to the
3863 entry code, but in registerised compilation we arrange
3864 that the info table is shoved right up against the entry
3865 code, and addressed backwards from the entry code pointer
3866 (this saves a word in the info table and an extra
3867 indirection when jumping to the closure entry
3872 <para>The mangler is abstracted to a certain extent over some
3873 architecture-specific things such as the particular assembler
3874 directives used to herald symbols. Take a look at the
3875 definitions for other architectures and use these as a
3876 starting point.</para>
3880 <title>The native code generator</title>
3882 <para>The native code generator isn't essential to getting a
3883 registerised build going, but it's a desirable thing to have
3884 because it can cut compilation times in half. The native code
3885 generator is described in some detail in the <ulink
3886 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3887 commentary</ulink>.</para>
3893 <para>To support GHCi, you need to port the dynamic linker
3894 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
3895 currently supports the ELF and PEi386 object file formats - if
3896 your platform uses one of these then you probably don't have
3897 to do anything except fiddle with the
3898 <literal>#ifdef</literal>s at the top of
3899 <filename>Linker.c</filename> to tell it about your OS.</para>
3901 <para>If your system uses a different object file format, then
3902 you have to write a linker — good luck!</para>
3908 <sect1 id="sec-build-pitfalls">
3909 <title>Known pitfalls in building Glasgow Haskell
3911 <indexterm><primary>problems, building</primary></indexterm>
3912 <indexterm><primary>pitfalls, in building</primary></indexterm>
3913 <indexterm><primary>building pitfalls</primary></indexterm></title>
3916 WARNINGS about pitfalls and known “problems”:
3925 One difficulty that comes up from time to time is running out of space
3926 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
3927 compensate for the vagaries of different sysadmin approaches to temp
3929 <indexterm><primary>tmp, running out of space in</primary></indexterm>
3931 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
3932 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
3935 The best way around it is to say
3938 export TMPDIR=<dir>
3941 in your <filename>build.mk</filename> file.
3942 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
3951 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
3952 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
3960 When compiling via C, you'll sometimes get “warning: assignment from
3961 incompatible pointer type” out of GCC. Harmless.
3968 Similarly, <command>ar</command>chiving warning messages like the following are not
3972 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
3973 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
3983 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
3984 get an “Out of heap space” error message. These can vary with the
3985 vagaries of different systems, it seems. The solution is simple:
3992 If you're compiling with GHC 4.00 or later, then the
3993 <emphasis>maximum</emphasis> heap size must have been reached. This
3994 is somewhat unlikely, since the maximum is set to 64M by default.
3995 Anyway, you can raise it with the
3996 <option>-optCrts-M<size></option> flag (add this flag to
3997 <constant><module>_HC_OPTS</constant>
3998 <command>make</command> variable in the appropriate
3999 <filename>Makefile</filename>).
4006 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
4015 and try again: <command>gmake</command>. (see <Xref LinkEnd="sec-suffix"> for information about
4016 <constant><module>_HC_OPTS</constant>.)
4018 Alternatively, just cut to the chase:
4022 % make EXTRA_HC_OPTS=-optCrts-M128M
4031 If you try to compile some Haskell, and you get errors from GCC about
4032 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4033 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4035 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4036 this bug also suggests that you have an old GCC.
4044 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4048 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4049 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4051 ? # or, on some machines: ar s $i
4056 We'd be interested to know if this is still necessary.
4064 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4065 a bit from one Unix to another. One particular gotcha is macro calls
4070 SLIT("Hello, world")
4074 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4075 arguments, so you get
4079 :731: macro `SLIT' used with too many (2) args
4083 Alas, <command>cpp</command> doesn't tell you the offending file!
4085 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4096 <Sect1 id="winbuild"><Title>Notes for building under Windows</Title>
4099 This section summarises how to get the utilities you need on your
4100 Win95/98/NT/2000 machine to use CVS and build GHC. Similar notes for
4101 installing and running GHC may be found in the user guide. In general,
4102 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4103 You should read the GHC installation guide sections on Windows (in the user
4104 guide) before continuing to read these notes.
4108 <sect2 id="cygwin-and-mingw"><Title>Cygwin and MinGW</Title>
4110 <para> The Windows situation for building GHC is rather confusing. This section
4111 tries to clarify, and to establish terminology.</para>
4113 <sect3 id="ghc-mingw"><title>GHC-mingw</title>
4115 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4116 is a collection of header
4117 files and import libraries that allow one to use <command>gcc</command> and produce
4118 native Win32 programs that do not rely on any third-party DLLs. The
4119 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4120 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4123 <para>The GHC that we distribute includes, inside the distribution itself, the MinGW <command>gcc</command>,
4124 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4125 GHC compiles Haskell to C (or to
4126 assembly code), and then invokes these MinGW tools to generate an executable binary.
4127 The resulting binaries can run on any Win32 system.
4129 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4131 <para> The down-side of GHC-mingw is that the MinGW libraries do not support anything like the full
4132 Posix interface. So programs compiled with GHC-mingw cannot import the (Haskell) Posix
4133 library; they have to do
4134 their input output using standard Haskell I/O libraries, or native Win32 bindings.
4138 <sect3 id="ghc-cygwin"><title>GHC-cygwin</title>
4140 <para>There <emphasis>is</emphasis> a way to get the full Posix interface, which is to use Cygwin.
4141 <ulink url="http://www.cygwin.com">Cygwin</ulink> is a complete Unix simulation that runs on Win32.
4142 Cygwin comes with a shell, and all the usual Unix commands: <command>mv</command>, <command>rm</command>,
4143 <command>ls</command>, plus of course <command>gcc</command>, <command>ld</command> and so on.
4144 A C program compiled with the Cygwin <command>gcc</command> certainly can use all of Posix.
4146 <para>So why doesn't GHC use the Cygwin <command>gcc</command> and libraries? Because
4147 Cygwin comes with a DLL <emphasis>that must be linked with every runnable Cygwin-compiled program</emphasis>.
4148 A program compiled by the Cygwin tools cannot run at all unless Cygwin is installed.
4149 If GHC targeted Cygwin, users would have to install Cygwin just to run the Haskell programs
4150 that GHC compiled; and the Cygwin DLL would have to be in the DLL load path.
4151 Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4152 does not change, but the implementation certainly does. Even the interfaces to functions
4153 it exports seem to change occasionally. So programs compiled by GHC might only run with
4154 particular versions of Cygwin. All of this seems very undesirable.
4157 Nevertheless, it is certainly possible to build a version of GHC that targets Cygwin;
4158 we will call that <emphasis>GHC-cygwin</emphasis>. The up-side of GHC-cygwin is
4159 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4163 <sect3><title>HOST_OS vs TARGET_OS</title>
4166 In the source code you'll find various ifdefs looking like:
4168 #ifdef mingw32_HOST_OS
4174 #ifdef mingw32_TARGET_OS
4178 These macros are set by the configure script (via the file config.h).
4179 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4182 The "host" system is the one on which GHC itself will be run.
4185 The "target" system is the one for which the program compiled by GHC will be run.
4188 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4189 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4194 <sect3><title>Summary</title>
4196 <para>Notice that "GHC-mingw" means "GHC that <emphasis>targets</emphasis> MinGW". It says nothing about
4197 how that GHC was <emphasis>built</emphasis>. It is entirely possible to have a GHC-mingw that was built
4198 by compiling GHC's Haskell sources with a GHC-cygwin, or vice versa.</para>
4200 <para>We distribute only a GHC-mingw built by a GHC-mingw; supporting
4201 GHC-cygwin too is beyond our resources. The GHC we distribute
4202 therefore does not require Cygwin to run, nor do the programs it
4203 compiles require Cygwin.</para>
4205 <para>The instructions that follow describe how to build GHC-mingw. It is
4206 possible to build GHC-cygwin, but it's not a supported route, and the build system might
4209 <para>In your build tree, you build a compiler called <Command>ghc-inplace</Command>. It
4210 uses the <Command>gcc</Command> that you specify using the
4211 <option>--with-gcc</option> flag when you run
4212 <Command>configure</Command> (see below).
4213 The makefiles are careful to use <Command>ghc-inplace</Command> (not <Command>gcc</Command>)
4214 to compile any C files, so that it will in turn invoke the right <Command>gcc</Command> rather that
4215 whatever one happens to be in your path. However, the makefiles do use whatever <Command>ld</Command>
4216 and <Command>ar</Command> happen to be in your path. This is a bit naughty, but (a) they are only
4217 used to glom together .o files into a bigger .o file, or a .a file,
4218 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4219 Cygwin and Mingw use the same .o file format. So its ok.
4224 <Sect2><Title>Installing and configuring Cygwin</Title>
4226 <para>You don't need Cygwin to <emphasis>use</emphasis> GHC,
4227 but you do need it to <emphasis>build</emphasis> GHC.</para>
4229 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4230 The installation process is straightforward; we install it in <Filename>c:/cygwin</Filename>.
4231 During the installation dialogue, make sure that you select:
4232 <command>cvs</command>, <command>openssh</command>,
4233 <command>autoconf</command>,
4234 <command>binutils</command> (includes ld and (I think) ar),
4235 <command>gcc</command>,
4236 <command>flex</command>,
4237 <command>make</command>.
4240 <para> Now set the following user environment variables:
4243 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4244 <constant>PATH</constant></para></listitem>
4248 Set <constant>MAKE_MODE</constant> to <Literal>UNIX</Literal>. If you
4249 don't do this you get very weird messages when you type
4250 <Command>make</Command>, such as:
4252 /c: /c: No such file or directory
4257 <listitem><para> Set <constant>SHELL</constant> to
4258 <Filename>c:/cygwin/bin/sh</Filename>. When you invoke a shell in Emacs, this
4259 <constant>SHELL</constant> is what you get.
4262 <listitem><para> Set <constant>HOME</constant> to point to your
4263 home directory. This is where, for example,
4264 <command>bash</command> will look for your <filename>.bashrc</filename>
4265 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4271 There are a few other things to do:
4275 By default, cygwin provides the command shell <filename>ash</filename>
4276 as <filename>sh.exe</filename>. We have often seen build-system problems that
4277 turn out to be due to bugs in <filename>ash</filename>
4279 and length of command lines). On the other hand <filename>bash</filename> seems
4281 So, in <filename>cygwin/bin</filename>
4282 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4283 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4284 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4285 you can't rename a running program!
4291 Some script files used in the make system start with "<Command>#!/bin/perl</Command>",
4292 (and similarly for <Command>sh</Command>). Notice the hardwired path!
4293 So you need to ensure that your <Filename>/bin</Filename> directory has the following
4296 <listitem> <para><Command>sh</Command></para></listitem>
4297 <listitem> <para><Command>perl</Command></para></listitem>
4298 <listitem> <para><Command>cat</Command></para></listitem>
4300 All these come in Cygwin's <Filename>bin</Filename> directory, which you probably have
4301 installed as <Filename>c:/cygwin/bin</Filename>. By default Cygwin mounts "<Filename>/</Filename>" as
4302 <Filename>c:/cygwin</Filename>, so if you just take the defaults it'll all work ok.
4303 (You can discover where your Cygwin
4304 root directory <Filename>/</Filename> is by typing <Command>mount</Command>.)
4305 Provided <Filename>/bin</Filename> points to the Cygwin <Filename>bin</Filename>
4306 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4307 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4313 <para>Finally, here are some things to be aware of when using Cygwin:
4315 <listitem> <para>Cygwin doesn't deal well with filenames that include
4316 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4320 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4321 magical text in it. So other programs that don't use Cygwin's
4322 I/O libraries won't recognise such files as symlinks.
4323 In particular, programs compiled by GHC are meant to be runnable
4324 without having Cygwin, so they don't use the Cygwin library, so
4325 they don't recognise symlinks.
4329 Win32 has a <command>find</command> command which is not the same as Cygwin's find.
4330 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4331 before the Cygwin one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4332 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4333 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4340 <Sect2><Title>Other things you need to install</Title>
4342 <para>You have to install the following other things to build GHC:
4346 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4347 This is what you will use to compile GHC. Add it in your
4348 <constant>PATH</constant>: the installer tells you the path element
4349 you need to add upon completion.
4355 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4356 Happy is a parser generator used to compile the Haskell grammar. Add it in your
4357 <constant>PATH</constant>.
4363 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4364 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw">).
4365 Just pick up a mingw bundle at
4366 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4367 We install it in <filename>c:/mingw</filename>.
4369 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4370 They are only going to get used by explicit access (via the --with-gcc flag you
4371 give to <Command>configure</Command> later). If you do add them to your path
4372 you are likely to get into a mess because their names overlap with Cygwin binaries.
4378 <para>We use <command>emacs</command> a lot, so we install that too.
4379 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4380 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4381 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4382 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4383 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4384 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4390 <para> Finally, check out a copy of GHC sources from
4391 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
4398 <Sect2><Title>Building GHC</Title>
4401 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source">);
4402 the bullets below only tell
4403 you about Windows-specific wrinkles.</para>
4407 Run <Command>autoconf</Command> both in <filename>fptools</filename>
4408 and in <filename>fptools/ghc</filename>. If you omit the latter step you'll
4409 get an error when you run <filename>./configure</filename>:
4412 creating mk/config.h
4413 mk/config.h is unchanged
4415 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4416 ./configure: ./configure: No such file or directory
4417 configure: error: ./configure failed for ghc
4422 <listitem> <para><command>autoconf</command> seems to create the file <filename>configure</filename>
4423 read-only. So if you need to run autoconf again (which I sometimes do for safety's sake),
4426 /usr/bin/autoconf: cannot create configure: permission denied
4428 Solution: delete <filename>configure</filename> first.
4433 You either need to add <filename>ghc</filename> to your
4434 <constant>PATH</constant> before you invoke
4435 <Command>configure</Command>, or use the <Command>configure</Command>
4436 option <option>--with-ghc=c:/ghc/ghc-some-version/bin/ghc</option>.
4441 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4442 This file occasionally remembers out-of-date configuration information, which
4443 can be really confusing.
4449 After <command>autoconf</command> run <command>./configure</command> in
4450 <filename>fptools/</filename> thus:
4453 ./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc
4455 This is the point at which you specify that you are building GHC-mingw
4456 (see <xref linkend="ghc-mingw">). </para>
4458 <para> Both these options are important! It's possible to get into
4459 trouble using the wrong C compiler!</para>
4461 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4462 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4463 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4464 understand a Cygwin path. For example, if you
4465 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4466 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4467 time it tries to invoke it. Worse, the failure comes with
4468 no error message whatsoever. GHC simply fails silently when first invoked,
4469 typically leaving you with this:
4471 make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4472 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4473 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4474 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4475 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4476 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4477 -optc-fomit-frame-pointer -O2 -static
4478 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4479 make[2]: *** [Adjustor.o] Error 1
4480 make[1]: *** [all] Error 1
4481 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4482 make: *** [all] Error 1
4488 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin">)
4489 you'll have to do something more like:
4491 ./configure --with-gcc=...the Cygwin gcc...
4496 <listitem><para> You almost certainly want to set
4500 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config">).
4501 This tells the build system not to split each library into a myriad of little object files, one
4502 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4503 it dramatically increases the time taken to build the libraries in the first place.
4507 <listitem><para> Do not attempt to build the documentation.
4508 It needs all kinds of wierd Jade stuff that we haven't worked out for
4509 Win32.</para></listitem>