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>
463 <sect2 id="cvs-committing">
464 <title>Committing Changes</title>
466 <para>This is only if you have read-write access to the
467 repository. For anoncvs users, CVS will issue a "read-only
468 repository" error if you try to commit changes.</para>
472 <para>Build the software, if necessary. Unless you're just
473 working on documentation, you'll probably want to build the
474 software in order to test any changes you make.</para>
478 <para>Make changes. Preferably small ones first.</para>
482 <para>Test them. You can see exactly what changes you've
483 made by using the <literal>cvs diff</literal> command:</para>
487 <para>lists all the changes (using the
488 <literal>diff</literal> command) in and below the current
489 directory. In emacs, <literal>C-c C-v =</literal> runs
490 <literal>cvs diff</literal> on the current buffer and shows
491 you the results.</para>
495 <para>Before checking in a change, you need to update your
502 <para>This pulls in any changes that other people have made,
503 and merges them with yours. If there are any conflicts, CVS
504 will tell you, and you'll have to resolve them before you
505 can check your changes in. The documentation describes what
506 to do in the event of a conflict.</para>
508 <para>It's not always necessary to do a full cvs update
509 before checking in a change, since CVS will always tell you
510 if you try to check in a file that someone else has changed.
511 However, you should still update at regular intervals to
512 avoid making changes that don't work in conjuction with
513 changes that someone else made. Keeping an eye on what goes
514 by on the mailing list can help here.</para>
518 <para>When you're happy that your change isn't going to
519 break anything, check it in. For a one-file change:</para>
522 $ cvs commit <replaceable>filename</replaceable>
525 <para>CVS will then pop up an editor for you to enter a
526 "commit message", this is just a short description
527 of what your change does, and will be kept in the history of
530 <para>If you're using emacs, simply load up the file into a
531 buffer and type <literal>C-x C-q</literal>, and emacs will
532 prompt for a commit message and then check in the file for
535 <para>For a multiple-file change, things are a bit
536 trickier. There are several ways to do this, but this is the
537 way I find easiest. First type the commit message into a
538 temporary file. Then either</para>
541 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
544 <para>or, if nothing else has changed in this part of the
548 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
551 <para>where <replaceable>directory</replaceable> is a common
552 parent directory for all your changes, and
553 <replaceable>commit-message</replaceable> is the name of the
554 file containing the commit message.</para>
556 <para>Shortly afterwards, you'll get some mail from the
557 relevant mailing list saying which files changed, and giving
558 the commit message. For a multiple-file change, you should
559 still get only <emphasis>one</emphasis> message.</para>
564 <sect2 id="cvs-update">
565 <title>Updating Your Source Tree</title>
567 <para>It can be tempting to cvs update just part of a source
568 tree to bring in some changes that someone else has made, or
569 before committing your own changes. This is NOT RECOMMENDED!
570 Quite often changes in one part of the tree are dependent on
571 changes in another part of the tree (the
572 <literal>mk/*.mk</literal> files are a good example where
573 problems crop up quite often). Having an inconsistent tree is a
574 major cause of headaches. </para>
576 <para>So, to avoid a lot of hassle, follow this recipe for
577 updating your tree: </para>
581 $ cvs update -Pd 2>&1 | tee log</screen>
583 <para>Look at the log file, and fix any conflicts (denoted by a
584 <quote>C</quote> in the first column). If you're using multiple
585 build trees, then for every build tree you have pointing at this
586 source tree, you need to update the links in case any new files
587 have appeared: </para>
590 $ cd <replaceable>build-tree</replaceable>
591 $ lndir <replaceable>source-tree</replaceable>
594 <para>Some files might have been removed, so you need to remove
595 the links pointing to these non-existent files:</para>
598 $ find . -xtype l -exec rm '{}' \;
601 <para>To be <emphasis>really</emphasis> safe, you should do
604 <screen>$ gmake all</screen>
606 <para>from the top-level, to update the dependencies and build
607 any changed files. </para>
610 <sect2 id="cvs-tags">
611 <title>GHC Tag Policy</title>
613 <para>If you want to check out a particular version of GHC,
614 you'll need to know how we tag versions in the repository. The
615 policy (as of 4.04) is:</para>
619 <para>The tree is branched before every major release. The
620 branch tag is <literal>ghc-x-xx-branch</literal>, where
621 <literal>x-xx</literal> is the version number of the release
622 with the <literal>'.'</literal> replaced by a
623 <literal>'-'</literal>. For example, the 4.04 release lives
624 on <literal>ghc-4-04-branch</literal>.</para>
628 <para>The release itself is tagged with
629 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
630 called <literal>ghc-4-06</literal>.</para>
634 <para>We didn't always follow these guidelines, so to see
635 what tags there are for previous versions, do <literal>cvs
636 log</literal> on a file that's been around for a while (like
637 <literal>fptools/ghc/README</literal>).</para>
641 <para>So, to check out a fresh GHC 4.06 tree you would
645 $ cvs co -r ghc-4-06 fpconfig
647 $ cvs co -r ghc-4-06 ghc hslibs
651 <sect2 id="cvs-hints">
652 <title>General Hints</title>
656 <para>As a general rule: commit changes in small units,
657 preferably addressing one issue or implementing a single
658 feature. Provide a descriptive log message so that the
659 repository records exactly which changes were required to
660 implement a given feature/fix a bug. I've found this
661 <emphasis>very</emphasis> useful in the past for finding out
662 when a particular bug was introduced: you can just wind back
663 the CVS tree until the bug disappears.</para>
667 <para>Keep the sources at least *buildable* at any given
668 time. No doubt bugs will creep in, but it's quite easy to
669 ensure that any change made at least leaves the tree in a
670 buildable state. We do nightly builds of GHC to keep an eye
671 on what things work/don't work each day and how we're doing
672 in relation to previous verions. This idea is truely wrecked
673 if the compiler won't build in the first place!</para>
677 <para>To check out extra bits into an already-checked-out
678 tree, use the following procedure. Suppose you have a
679 checked-out fptools tree containing just ghc, and you want
680 to add nofib to it:</para>
691 $ cvs update -d nofib
694 <para>(the -d flag tells update to create a new
695 directory). If you just want part of the nofib suite, you
700 $ cvs checkout nofib/spectral
703 <para>This works because <literal>nofib</literal> is a
704 module in its own right, and spectral is a subdirectory of
705 the nofib module. The path argument to checkout must always
706 start with a module name. There's no equivalent form of this
707 command using <literal>update</literal>.</para>
713 <sect1 id="projects">
714 <title>What projects are there?</title>
716 <para>The <literal>fptools</literal> suite consists of several
717 <firstterm>projects</firstterm>, most of which can be downloaded,
718 built and installed individually. Each project corresponds to a
719 subdirectory in the source tree, and if checking out from CVS then
720 each project can be checked out individually by sitting in the top
721 level of your source tree and typing <command>cvs checkout
722 <replaceable>project</replaceable></command>.</para>
724 <para>Here is a list of the projects currently available:</para>
728 <term><literal>ghc</literal></term>
729 <indexterm><primary><literal>ghc</literal></primary>
730 <secondary>project</secondary></indexterm>
732 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
733 Haskell Compiler</ulink> (minus libraries). Absolutely
734 required for building GHC.</para>
739 <term><literal>glafp-utils</literal></term>
740 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
742 <para>Utility programs, some of which are used by the
743 build/installation system. Required for pretty much
749 <term><literal>green-card</literal></term>
750 <indexterm><primary><literal>green-card</literal></primary><secondary>project</secondary></indexterm>
753 url="http://www.haskell.org/greencard/">Green Card</ulink>
754 system for generating Haskell foreign function
760 <term><literal>haggis</literal></term>
761 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
764 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
765 Haskell GUI framework.</para>
770 <term><literal>haddock</literal></term>
771 <indexterm><primary><literal>haddock</literal></primary><secondary>project</secondary></indexterm>
774 url="http://www.haskell.org/haddock/">Haddock</ulink>
775 documentation tool.</para>
780 <term><literal>happy</literal></term>
781 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
784 url="http://www.haskell.org/happy/">Happy</ulink> Parser
790 <term><literal>hdirect</literal></term>
791 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
794 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
795 Haskell interoperability tool.</para>
800 <term><literal>hood</literal></term>
801 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
803 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
804 Object Observation Debugger</ulink>.</para>
809 <term><literal>hslibs</literal></term>
810 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
812 <para>Supplemental libraries for GHC
813 (<emphasis>required</emphasis> for building GHC).</para>
818 <term><literal>libraries</literal></term>
819 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
821 <para>Hierarchical Haskell library suite
822 (<emphasis>required</emphasis> for building GHC).</para>
827 <term><literal>mhms</literal></term>
828 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
830 <para>The Modular Haskell Metric System.</para>
835 <term><literal>nofib</literal></term>
836 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
838 <para>The NoFib suite: A collection of Haskell programs used
839 primarily for benchmarking.</para>
844 <term><literal>testsuite</literal></term>
845 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
847 <para>A testing framework, including GHC's regression test
853 <para>So, to build GHC you need at least the
854 <literal>ghc</literal>, <literal>libraries</literal> and
855 <literal>hslibs</literal> projects (a GHC source distribution will
856 already include the bits you need).</para>
859 <sect1 id="sec-build-checks">
860 <title>Things to check before you start</title>
862 <para>Here's a list of things to check before you get
868 <indexterm><primary>Disk space needed</primary></indexterm>
869 <para>Disk space needed: from about 100Mb for a basic GHC
870 build, up to probably 500Mb for a GHC build with everything
871 included (libraries built several different ways,
876 <para>Use an appropriate machine / operating system. <xref
877 linkend="sec-port-info"> lists the supported platforms; if
878 yours isn't amongst these then you can try porting GHC (see
879 <xref linkend="sec-porting-ghc">).</para>
883 <para>Be sure that the “pre-supposed” utilities are
884 installed. <Xref LinkEnd="sec-pre-supposed">
889 <para>If you have any problem when building or installing the
890 Glasgow tools, please check the “known pitfalls” (<Xref
891 LinkEnd="sec-build-pitfalls">). Also check the FAQ for the
892 version you're building, which is part of the User's Guide and
893 available on the <ulink URL="http://www.haskell.org/ghc/" >GHC web
896 <indexterm><primary>bugs</primary><secondary>known</secondary></indexterm>
898 <para>If you feel there is still some shortcoming in our
899 procedure or instructions, please report it.</para>
901 <para>For GHC, please see the <ulink
902 url="http://www.haskell.org/ghc/docs/latest/set/bug-reporting.html">bug-reporting
903 section of the GHC Users' Guide</ulink>, to maximise the
904 usefulness of your report.</para>
906 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
907 <para>If in doubt, please send a message to
908 <email>glasgow-haskell-bugs@haskell.org</email>.
909 <indexterm><primary>bugs</primary><secondary>mailing
910 list</secondary></indexterm></para>
915 <sect1 id="sec-port-info">
916 <title>What machines the Glasgow tools run on</title>
918 <indexterm><primary>ports</primary><secondary>GHC</secondary></indexterm>
919 <indexterm><primary>GHC</primary><secondary>ports</secondary></indexterm>
920 <indexterm><primary>platforms</primary><secondary>supported</secondary></indexterm>
922 <para>The main question is whether or not the Haskell compiler
923 (GHC) runs on your platform.</para>
925 <para>A “platform” is a
926 architecture/manufacturer/operating-system combination, such as
927 <literal>sparc-sun-solaris2</literal>. Other common ones are
928 <literal>alpha-dec-osf2</literal>,
929 <literal>hppa1.1-hp-hpux9</literal>,
930 <literal>i386-unknown-linux</literal>,
931 <literal>i386-unknown-solaris2</literal>,
932 <literal>i386-unknown-freebsd</literal>,
933 <literal>i386-unknown-cygwin32</literal>,
934 <literal>m68k-sun-sunos4</literal>,
935 <literal>mips-sgi-irix5</literal>,
936 <literal>sparc-sun-sunos4</literal>,
937 <literal>sparc-sun-solaris2</literal>,
938 <literal>powerpc-ibm-aix</literal>.</para>
940 <para>Some libraries may only work on a limited number of
941 platforms; for example, a sockets library is of no use unless the
942 operating system supports the underlying BSDisms.</para>
945 <title>What platforms the Haskell compiler (GHC) runs on</title>
947 <indexterm><primary>fully-supported platforms</primary></indexterm>
948 <indexterm><primary>native-code generator</primary></indexterm>
949 <indexterm><primary>registerised ports</primary></indexterm>
950 <indexterm><primary>unregisterised ports</primary></indexterm>
952 <para>The GHC hierarchy of Porting Goodness: (a) Best is a
953 native-code generator; (b) next best is a
954 “registerised” port; (c) the bare minimum is an
955 “unregisterised” port.
956 (“Unregisterised” is so terrible that we won't say
957 more about it).</para>
959 <para>We use Sparcs running Solaris 2.7 and x86 boxes running
960 FreeBSD and Linux, so those are the best supported platforms,
961 unsurprisingly.</para>
963 <para>Here's everything that's known about GHC ports. We
964 identify platforms by their “canonical”
965 CPU/Manufacturer/OS triple.</para>
969 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:</term>
970 <indexterm><primary>alpha-dec-osf</primary></indexterm>
971 <indexterm><primary>alpha-dec-linux</primary></indexterm>
972 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
973 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
974 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
977 <para>The OSF port is currently working (as of GHC version
978 5.02.1) and well supported. The native code generator is
979 currently non-working. Other operating systems will
980 require some minor porting.</para>
985 <term>sparc-sun-sunos4</term>
986 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
988 <para>Probably works with minor tweaks, hasn't been tested
994 <term>sparc-sun-solaris2</term>
995 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
997 <para>Fully supported (at least for Solaris 2.7),
998 including native-code generator.</para>
1003 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
1004 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1006 <para>A registerised port is available for version 4.08,
1007 but GHC hasn't been built on that platform since (as far
1008 as we know). No native-code generator.</para>
1013 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1014 <indexterm><primary>i386-*-linux</primary></indexterm>
1016 <para>GHC works registerised and has a native code
1017 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1018 or later. NOTE about <literal>glibc</literal> versions:
1019 GHC binaries built on a system running <literal>glibc
1020 2.0</literal> won't work on a system running
1021 <literal>glibc 2.1</literal>, and vice versa. In general,
1022 don't expect compatibility between
1023 <literal>glibc</literal> versions, even if the shared
1024 library version hasn't changed.</para>
1029 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or
1031 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1033 <para>GHC works registerised. Pre-built packages are
1034 available in the native package format, so if you just
1035 need binaries you're better off just installing the
1036 package (it might even be on your installation
1042 <term>i386-unknown-openbsd (PCs running OpenBSD)</term>
1043 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1045 <para>Supported, with native code generator. Packages are
1046 available through the ports system in the native package
1052 <term>i386-unknown-netbsd (PCs running NetBSD and
1054 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1056 <para>Will require some minor porting effort, but should
1057 work registerised.</para>
1062 <term>i386-unknown-mingw32 (PCs running Windows)</term>
1063 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1065 <para>Fully supported under Win9x, WinNT, Win2k, and
1066 WinXP. Includes a native code generator. Building from
1067 source requires a recent <ulink
1068 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1069 to be installed.</para>
1074 <term>ia64-unknown-linux</term>
1075 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1077 <para>GHC currently works unregisterised. A registerised
1078 port is in progress.</para>
1083 <term>mips-sgi-irix5</term>
1084 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1086 <para>Port has worked in the past, but hasn't been tested
1087 for some time (and will certainly have rotted in various
1088 ways). As usual, we don't have access to machines and
1089 there hasn't been an overwhelming demand for this port,
1090 but feel free to get in touch.</para>
1095 <term>powerpc-ibm-aix</term>
1096 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1098 <para>Port currently doesn't work, needs some minimal
1099 porting effort. As usual, we don't have access to
1100 machines and there hasn't been an overwhelming demand for
1101 this port, but feel free to get in touch.</para>
1106 <term>powerpc-apple-darwin</term>
1107 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1109 <para>Supported registerised. No native code
1115 <term>powerpc-apple-linux</term>
1116 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1118 <para>Not supported (yet).</para>
1123 <para>Various other systems have had GHC ported to them in the
1124 distant past, including various Motorola 68k boxes. The 68k
1125 support still remains, but porting to one of these systems will
1126 certainly be a non-trivial task.</para>
1130 <title>What machines the other tools run on</title>
1132 <para>Unless you hear otherwise, the other tools work if GHC
1138 <sect1 id="sec-pre-supposed">
1139 <title>Installing pre-supposed utilities</title>
1141 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1142 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1144 <para>Here are the gory details about some utility programs you
1145 may need; <command>perl</command>, <command>gcc</command> and
1146 <command>happy</command> are the only important
1147 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1148 important if you're going for Parallel Haskell.) The
1149 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1150 script will tell you if you are missing something.</para>
1156 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1157 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1159 <para>GHC is required to build many of the tools, including
1160 GHC itself. If you need to port GHC to your platform
1161 because there isn't a binary distribution of GHC available,
1162 then see <xref linkend="sec-porting-ghc">.</para>
1164 <para>Which version of GHC you need will depend on the
1165 packages you intend to build. GHC itself will normally
1166 build using one of several older versions of itself - check
1167 the announcement or release notes for details.</para>
1173 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1174 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1176 <para><emphasis>You have to have Perl to proceed!</emphasis>
1177 Perl version 5 at least is required. GHC has been known to
1178 tickle bugs in Perl, so if you find that Perl crashes when
1179 running GHC try updating (or downgrading) your Perl
1180 installation. Versions of Perl that we use and are known to
1181 be fairly stable are 5.005 and 5.6.1.</para>
1183 <para>For Win32 platforms, you should use the binary
1184 supplied in the InstallShield (copy it to
1185 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1188 <para>Perl should be put somewhere so that it can be invoked
1189 by the <literal>#!</literal> script-invoking
1190 mechanism. The full pathname may need to be less than 32
1191 characters long on some systems.</para>
1196 <term>GNU C (<command>gcc</command>)</term>
1197 <indexterm><primary>pre-supposed: GCC (GNU C
1198 compiler)</primary></indexterm> <indexterm><primary>GCC (GNU C
1199 compiler), pre-supposed</primary></indexterm>
1201 <para>We recommend using GCC version 2.95.2 on all
1202 platforms. Failing that, version 2.7.2 is stable on most
1203 platforms. Earlier versions of GCC can be assumed not to
1204 work, and versions in between 2.7.2 and 2.95.2 (including
1205 <command>egcs</command>) have varying degrees of stability
1206 depending on the platform.</para>
1208 <para>If your GCC dies with “internal error” on
1209 some GHC source file, please let us know, so we can report
1210 it and get things improved. (Exception: on iX86
1211 boxes—you may need to fiddle with GHC's
1212 <option>-monly-N-regs</option> option; see the User's
1218 <term>GNU Make</term>
1219 <indexterm><primary>make</primary><secondary>GNU</secondary>
1222 <para>The fptools build system makes heavy use of features
1223 specific to GNU <command>make</command>, so you must have
1224 this installed in order to build any of the fptools
1231 <indexterm><primary>Happy</primary></indexterm>
1233 <para>Happy is a parser generator tool for Haskell, and is
1234 used to generate GHC's parsers. Happy is written in
1235 Haskell, and is a project in the CVS repository
1236 (<literal>fptools/happy</literal>). It can be built from
1237 source, but bear in mind that you'll need GHC installed in
1238 order to build it. To avoid the chicken/egg problem,
1239 install a binary distribtion of either Happy or GHC to get
1240 started. Happy distributions are available from <ulink
1241 url="http://www.haskell.org/happy/">Happy's Web
1242 Page</ulink>.</para>
1247 <term>Autoconf</term>
1248 <indexterm><primary>pre-supposed: Autoconf</primary></indexterm>
1249 <indexterm><primary>Autoconf, pre-supposed</primary></indexterm>
1251 <para>GNU Autoconf is needed if you intend to build from the
1252 CVS sources, it is <emphasis>not</emphasis> needed if you
1253 just intend to build a standard source distribution.</para>
1255 <para>Autoconf builds the <command>configure</command>
1256 script from <filename>configure.in</filename> and
1257 <filename>aclocal.m4</filename>. If you modify either of
1258 these files, you'll need <command>autoconf</command> to
1259 rebuild <filename>configure</filename>.</para>
1264 <term><command>sed</command></term>
1265 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1266 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1268 <para>You need a working <command>sed</command> if you are
1269 going to build from sources. The build-configuration stuff
1270 needs it. GNU sed version 2.0.4 is no good! It has a bug
1271 in it that is tickled by the build-configuration. 2.0.5 is
1272 OK. Others are probably OK too (assuming we don't create too
1273 elaborate configure scripts.)</para>
1278 <para>One <literal>fptools</literal> project is worth a quick note
1279 at this point, because it is useful for all the others:
1280 <literal>glafp-utils</literal> contains several utilities which
1281 aren't particularly Glasgow-ish, but Occasionally Indispensable.
1282 Like <command>lndir</command> for creating symbolic link
1285 <sect2 id="pre-supposed-gph-tools">
1286 <title>Tools for building parallel GHC (GPH)</title>
1290 <term>PVM version 3:</term>
1291 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1292 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1294 <para>PVM is the Parallel Virtual Machine on which
1295 Parallel Haskell programs run. (You only need this if you
1296 plan to run Parallel Haskell. Concurent Haskell, which
1297 runs concurrent threads on a uniprocessor doesn't need
1298 it.) Underneath PVM, you can have (for example) a network
1299 of workstations (slow) or a multiprocessor box
1302 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1303 It is readily available on the net; I think I got it from
1304 <literal>research.att.com</literal>, in
1305 <filename>netlib</filename>.</para>
1307 <para>A PVM installation is slightly quirky, but easy to
1308 do. Just follow the <filename>Readme</filename>
1309 instructions.</para>
1314 <term><command>bash</command>:</term>
1315 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1317 <para>Sadly, the <command>gr2ps</command> script, used to
1318 convert “parallelism profiles” to PostScript,
1319 is written in Bash (GNU's Bourne Again shell). This bug
1320 will be fixed (someday).</para>
1326 <sect2 id="pre-supposed-other-tools">
1327 <title>Other useful tools</title>
1332 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1333 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1335 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1336 to build a couple of utilities in
1337 <literal>glafp-utils</literal>. Depending on your
1338 operating system, the supplied <command>lex</command> may
1339 or may not work; you should get the GNU version.</para>
1344 <para>More tools are required if you want to format the documentation
1345 that comes with GHC and other fptools projects. See <xref
1346 linkend="building-docs">.</para>
1350 <sect1 id="sec-building-from-source">
1351 <title>Building from source</title>
1353 <indexterm><primary>Building from source</primary></indexterm>
1354 <indexterm><primary>Source, building from</primary></indexterm>
1356 <para>You've been rash enough to want to build some of the Glasgow
1357 Functional Programming tools (GHC, Happy, nofib, etc.) from
1358 source. You've slurped the source, from the CVS repository or
1359 from a source distribution, and now you're sitting looking at a
1360 huge mound of bits, wondering what to do next.</para>
1362 <para>Gingerly, you type <command>make</command>. Wrong
1365 <para>This rest of this guide is intended for duffers like me, who
1366 aren't really interested in Makefiles and systems configurations,
1367 but who need a mental model of the interlocking pieces so that
1368 they can make them work, extend them consistently when adding new
1369 software, and lay hands on them gently when they don't
1372 <sect2 id="quick-start">
1373 <title>Quick Start</title>
1375 <para>If you are starting from a source distribution, and just
1376 want a completely standard build, then the following should
1379 <screen>$ ./configure
1384 <para>For GHC, this will do a 2-stage bootstrap build of the
1385 compiler, with profiling libraries, and install the
1388 <para>If you want to do anything at all non-standard, or you
1389 want to do some development, read on...</para>
1392 <sect2 id="sec-source-tree">
1393 <title>Your source tree</title>
1395 <para>The source code is held in your <emphasis>source
1396 tree</emphasis>. The root directory of your source tree
1397 <emphasis>must</emphasis> contain the following directories and
1402 <para><filename>Makefile</filename>: the root
1407 <para><filename>mk/</filename>: the directory that contains
1408 the main Makefile code, shared by all the
1409 <literal>fptools</literal> software.</para>
1413 <para><filename>configure.in</filename>,
1414 <filename>config.sub</filename>,
1415 <filename>config.guess</filename>: these files support the
1416 configuration process.</para>
1420 <para><filename>install-sh</filename>.</para>
1424 <para>All the other directories are individual
1425 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1426 system—for example, the Glasgow Haskell Compiler
1427 (<literal>ghc</literal>), the Happy parser generator
1428 (<literal>happy</literal>), the <literal>nofib</literal>
1429 benchmark suite, and so on. You can have zero or more of these.
1430 Needless to say, some of them are needed to build others.</para>
1432 <para>The important thing to remember is that even if you want
1433 only one project (<literal>happy</literal>, say), you must have
1434 a source tree whose root directory contains
1435 <filename>Makefile</filename>, <filename>mk/</filename>,
1436 <filename>configure.in</filename>, and the project(s) you want
1437 (<filename>happy/</filename> in this case). You cannot get by
1438 with just the <filename>happy/</filename> directory.</para>
1442 <title>Build trees</title>
1443 <indexterm><primary>build trees</primary></indexterm>
1444 <indexterm><primary>link trees, for building</primary></indexterm>
1446 <para>If you just want to build the software once on a single
1447 platform, then your source tree can also be your build tree, and
1448 you can skip the rest of this section.</para>
1450 <para>We often want to build multiple versions of our software
1451 for different architectures, or with different options
1452 (e.g. profiling). It's very desirable to share a single copy of
1453 the source code among all these builds.</para>
1455 <para>So for every source tree we have zero or more
1456 <emphasis>build trees</emphasis>. Each build tree is initially
1457 an exact copy of the source tree, except that each file is a
1458 symbolic link to the source file, rather than being a copy of
1459 the source file. There are “standard” Unix
1460 utilities that make such copies, so standard that they go by
1462 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1463 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1464 are two (If you don't have either, the source distribution
1465 includes sources for the X11
1466 <command>lndir</command>—check out
1467 <filename>fptools/glafp-utils/lndir</filename>). See <Xref
1468 LinkEnd="sec-storysofar"> for a typical invocation.</para>
1470 <para>The build tree does not need to be anywhere near the
1471 source tree in the file system. Indeed, one advantage of
1472 separating the build tree from the source is that the build tree
1473 can be placed in a non-backed-up partition, saving your systems
1474 support people from backing up untold megabytes of
1475 easily-regenerated, and rapidly-changing, gubbins. The golden
1476 rule is that (with a single exception—<XRef
1477 LinkEnd="sec-build-config">) <emphasis>absolutely everything in
1478 the build tree is either a symbolic link to the source tree, or
1479 else is mechanically generated</emphasis>. It should be
1480 perfectly OK for your build tree to vanish overnight; an hour or
1481 two compiling and you're on the road again.</para>
1483 <para>You need to be a bit careful, though, that any new files
1484 you create (if you do any development work) are in the source
1485 tree, not a build tree!</para>
1487 <para>Remember, that the source files in the build tree are
1488 <emphasis>symbolic links</emphasis> to the files in the source
1489 tree. (The build tree soon accumulates lots of built files like
1490 <filename>Foo.o</filename>, as well.) You can
1491 <emphasis>delete</emphasis> a source file from the build tree
1492 without affecting the source tree (though it's an odd thing to
1493 do). On the other hand, if you <emphasis>edit</emphasis> a
1494 source file from the build tree, you'll edit the source-tree
1495 file directly. (You can set up Emacs so that if you edit a
1496 source file from the build tree, Emacs will silently create an
1497 edited copy of the source file in the build tree, leaving the
1498 source file unchanged; but the danger is that you think you've
1499 edited the source file whereas actually all you've done is edit
1500 the build-tree copy. More commonly you do want to edit the
1501 source file.)</para>
1503 <para>Like the source tree, the top level of your build tree
1504 must be (a linked copy of) the root directory of the
1505 <literal>fptools</literal> suite. Inside Makefiles, the root of
1506 your build tree is called
1507 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1508 In the rest of this document path names are relative to
1509 <constant>$(FPTOOLS_TOP)</constant> unless
1510 otherwise stated. For example, the file
1511 <filename>ghc/mk/target.mk</filename> is actually
1512 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</filename>.</para>
1515 <sect2 id="sec-build-config">
1516 <title>Getting the build you want</title>
1518 <para>When you build <literal>fptools</literal> you will be
1519 compiling code on a particular <emphasis>host
1520 platform</emphasis>, to run on a particular <emphasis>target
1521 platform</emphasis> (usually the same as the host
1522 platform)<indexterm><primary>platform</primary></indexterm>.
1523 The difficulty is that there are minor differences between
1524 different platforms; minor, but enough that the code needs to be
1525 a bit different for each. There are some big differences too:
1526 for a different architecture we need to build GHC with a
1527 different native-code generator.</para>
1529 <para>There are also knobs you can turn to control how the
1530 <literal>fptools</literal> software is built. For example, you
1531 might want to build GHC optimised (so that it runs fast) or
1532 unoptimised (so that you can compile it fast after you've
1533 modified it. Or, you might want to compile it with debugging on
1534 (so that extra consistency-checking code gets included) or off.
1537 <para>All of this stuff is called the
1538 <emphasis>configuration</emphasis> of your build. You set the
1539 configuration using a three-step process.</para>
1543 <term>Step 1: get ready for configuration.</term>
1545 <para>NOTE: if you're starting from a source distribution,
1546 rather than CVS sources, you can skip this step.</para>
1548 <para>Change directory to
1549 <constant>$(FPTOOLS_TOP)</constant> and
1551 <command>autoconf</command><indexterm><primary>autoconf</primary></indexterm>
1552 (with no arguments). This GNU program converts
1553 <filename><constant>$(FPTOOLS_TOP)</constant>/configure.in</filename>
1554 to a shell script called
1555 <filename><constant>$(FPTOOLS_TOP)</constant>/configure</filename>.
1558 <para>Some projects, including GHC, have their own
1559 configure script. If there's an
1560 <constant>$(FPTOOLS_TOP)/<project>/configure.in</constant>,
1561 then you need to run <command>autoconf</command> in that
1562 directory too.</para>
1564 <para>Both these steps are completely
1565 platform-independent; they just mean that the
1566 human-written file (<filename>configure.in</filename>) can
1567 be short, although the resulting shell script,
1568 <command>configure</command>, and
1569 <filename>mk/config.h.in</filename>, are long.</para>
1574 <term>Step 2: system configuration.</term>
1576 <para>Runs the newly-created <command>configure</command>
1577 script, thus:</para>
1580 ./configure <optional><parameter>args</parameter></optional>
1583 <para><command>configure</command>'s mission is to scurry
1584 round your computer working out what architecture it has,
1585 what operating system, whether it has the
1586 <Function>vfork</Function> system call, where
1587 <command>yacc</command> is kept, whether
1588 <command>gcc</command> is available, where various obscure
1589 <literal>#include</literal> files are, whether it's a
1590 leap year, and what the systems manager had for lunch. It
1591 communicates these snippets of information in two
1598 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1600 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1601 substituting for things between
1602 “<literal>@</literal>” brackets. So,
1603 “<literal>@HaveGcc@</literal>” will be
1604 replaced by “<literal>YES</literal>” or
1605 “<literal>NO</literal>” depending on what
1606 <command>configure</command> finds.
1607 <filename>mk/config.mk</filename> is included by every
1608 Makefile (directly or indirectly), so the
1609 configuration information is thereby communicated to
1610 all Makefiles.</para>
1614 <para> It translates
1615 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1617 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1618 The latter is <literal>#include</literal>d by
1619 various C programs, which can thereby make use of
1620 configuration information.</para>
1624 <para><command>configure</command> takes some optional
1625 arguments. Use <literal>./configure --help</literal> to
1626 get a list of the available arguments. Here are some of
1627 the ones you might need:</para>
1631 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1632 <indexterm><primary><literal>--with-ghc</literal></primary>
1635 <para>Specifies the path to an installed GHC which
1636 you would like to use. This compiler will be used
1637 for compiling GHC-specific code (eg. GHC itself).
1638 This option <emphasis>cannot</emphasis> be specified
1639 using <filename>build.mk</filename> (see later),
1640 because <command>configure</command> needs to
1641 auto-detect the version of GHC you're using. The
1642 default is to look for a compiler named
1643 <literal>ghc</literal> in your path.</para>
1648 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1649 <indexterm><primary><literal>--with-hc</literal></primary>
1652 <para>Specifies the path to any installed Haskell
1653 compiler. This compiler will be used for compiling
1654 generic Haskell code. The default is to use
1655 <literal>ghc</literal>.</para>
1660 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1661 <indexterm><primary><literal>--with-gcc</literal></primary>
1664 <para>Specifies the path to the installed GCC. This
1665 compiler will be used to compile all C files,
1666 <emphasis>except</emphasis> any generated by the
1667 installed Haskell compiler, which will have its own
1668 idea of which C compiler (if any) to use. The
1669 default is to use <literal>gcc</literal>.</para>
1674 <para><command>configure</command> caches the results of
1675 its run in <filename>config.cache</filename>. Quite often
1676 you don't want that; you're running
1677 <command>configure</command> a second time because
1678 something has changed. In that case, simply delete
1679 <filename>config.cache</filename>.</para>
1684 <term>Step 3: build configuration.</term>
1686 <para>Next, you say how this build of
1687 <literal>fptools</literal> is to differ from the standard
1688 defaults by creating a new file
1689 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1690 <emphasis>in the build tree</emphasis>. This file is the
1691 one and only file you edit in the build tree, precisely
1692 because it says how this build differs from the source.
1693 (Just in case your build tree does die, you might want to
1694 keep a private directory of <filename>build.mk</filename>
1695 files, and use a symbolic link in each build tree to point
1696 to the appropriate one.) So
1697 <filename>mk/build.mk</filename> never exists in the
1698 source tree—you create one in each build tree from
1699 the template. We'll discuss what to put in it
1705 <para>And that's it for configuration. Simple, eh?</para>
1707 <para>What do you put in your build-specific configuration file
1708 <filename>mk/build.mk</filename>? <emphasis>For almost all
1709 purposes all you will do is put make variable definitions that
1710 override those in</emphasis>
1711 <filename>mk/config.mk.in</filename>. The whole point of
1712 <filename>mk/config.mk.in</filename>—and its derived
1713 counterpart <filename>mk/config.mk</filename>—is to define
1714 the build configuration. It is heavily commented, as you will
1715 see if you look at it. So generally, what you do is look at
1716 <filename>mk/config.mk.in</filename>, and add definitions in
1717 <filename>mk/build.mk</filename> that override any of the
1718 <filename>config.mk</filename> definitions that you want to
1719 change. (The override occurs because the main boilerplate file,
1720 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1721 includes <filename>build.mk</filename> after
1722 <filename>config.mk</filename>.)</para>
1724 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1725 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1727 <para>For example, <filename>config.mk.in</filename> contains
1728 the definition:</para>
1731 GhcHcOpts=-O -Rghc-timing
1734 <para>The accompanying comment explains that this is the list of
1735 flags passed to GHC when building GHC itself. For doing
1736 development, it is wise to add <literal>-DDEBUG</literal>, to
1737 enable debugging code. So you would add the following to
1738 <filename>build.mk</filename>:</para>
1740 <para>or, if you prefer,</para>
1743 GhcHcOpts += -DDEBUG
1746 <para>GNU <command>make</command> allows existing definitions to
1747 have new text appended using the “<literal>+=</literal>”
1748 operator, which is quite a convenient feature.)</para>
1750 <para>If you want to remove the <literal>-O</literal> as well (a
1751 good idea when developing, because the turn-around cycle gets a
1752 lot quicker), you can just override
1753 <literal>GhcLibHcOpts</literal> altogether:</para>
1756 GhcHcOpts=-DDEBUG -Rghc-timing
1759 <para>When reading <filename>config.mk.in</filename>, remember
1760 that anything between “@...@” signs is going to be substituted
1761 by <command>configure</command> later. You
1762 <emphasis>can</emphasis> override the resulting definition if
1763 you want, but you need to be a bit surer what you are doing.
1764 For example, there's a line that says:</para>
1770 <para>This defines the Make variables <constant>YACC</constant>
1771 to the pathname for a <command>yacc</command> that
1772 <command>configure</command> finds somewhere. If you have your
1773 own pet <command>yacc</command> you want to use instead, that's
1774 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1780 <para>You do not <emphasis>have</emphasis> to have a
1781 <filename>mk/build.mk</filename> file at all; if you don't,
1782 you'll get all the default settings from
1783 <filename>mk/config.mk.in</filename>.</para>
1785 <para>You can also use <filename>build.mk</filename> to override
1786 anything that <command>configure</command> got wrong. One place
1787 where this happens often is with the definition of
1788 <constant>FPTOOLS_TOP_ABS</constant>: this
1789 variable is supposed to be the canonical path to the top of your
1790 source tree, but if your system uses an automounter then the
1791 correct directory is hard to find automatically. If you find
1792 that <command>configure</command> has got it wrong, just put the
1793 correct definition in <filename>build.mk</filename>.</para>
1797 <sect2 id="sec-storysofar">
1798 <title>The story so far</title>
1800 <para>Let's summarise the steps you need to carry to get
1801 yourself a fully-configured build tree from scratch.</para>
1805 <para> Get your source tree from somewhere (CVS repository
1806 or source distribution). Say you call the root directory
1807 <filename>myfptools</filename> (it does not have to be
1808 called <filename>fptools</filename>). Make sure that you
1809 have the essential files (see <XRef
1810 LinkEnd="sec-source-tree">).</para>
1815 <para>(Optional) Use <command>lndir</command> or
1816 <command>mkshadowdir</command> to create a build tree.</para>
1820 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1823 <para>(N.B. <command>mkshadowdir</command>'s first argument
1824 is taken relative to its second.) You probably want to give
1825 the build tree a name that suggests its main defining
1826 characteristic (in your mind at least), in case you later
1831 <para>Change directory to the build tree. Everything is
1832 going to happen there now.</para>
1835 $ cd /scratch/joe-bloggs/myfptools-sun4
1841 <para>Prepare for system configuration:</para>
1847 <para>(You can skip this step if you are starting from a
1848 source distribution, and you already have
1849 <filename>configure</filename> and
1850 <filename>mk/config.h.in</filename>.)</para>
1852 <para>Some projects, including GHC itself, have their own
1853 configure scripts, so it is necessary to run autoconf again
1854 in the appropriate subdirectories. eg:</para>
1857 $ (cd ghc; autoconf)
1862 <para>Do system configuration:</para>
1868 <para>Don't forget to check whether you need to add any
1869 arguments to <literal>configure</literal>; for example, a
1870 common requirement is to specify which GHC to use with
1871 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1875 <para>Create the file <filename>mk/build.mk</filename>,
1876 adding definitions for your desired configuration
1885 <para>You can make subsequent changes to
1886 <filename>mk/build.mk</filename> as often as you like. You do
1887 not have to run any further configuration programs to make these
1888 changes take effect. In theory you should, however, say
1889 <command>gmake clean</command>, <command>gmake all</command>,
1890 because configuration option changes could affect
1891 anything—but in practice you are likely to know what's
1896 <title>Making things</title>
1898 <para>At this point you have made yourself a fully-configured
1899 build tree, so you are ready to start building real
1902 <para>The first thing you need to know is that <emphasis>you
1903 must use GNU <command>make</command>, usually called
1904 <command>gmake</command>, not standard Unix
1905 <command>make</command></emphasis>. If you use standard Unix
1906 <command>make</command> you will get all sorts of error messages
1907 (but no damage) because the <literal>fptools</literal>
1908 <command>Makefiles</command> use GNU <command>make</command>'s
1909 facilities extensively.</para>
1911 <para>To just build the whole thing, <command>cd</command> to
1912 the top of your <literal>fptools</literal> tree and type
1913 <command>gmake</command>. This will prepare the tree and build
1914 the various projects in the correct order.</para>
1917 <sect2 id="sec-bootstrapping">
1918 <title>Bootstrapping GHC</title>
1920 <para>GHC requires a 2-stage bootstrap in order to provide
1921 full functionality, including GHCi. By a 2-stage bootstrap, we
1922 mean that the compiler is built once using the installed GHC,
1923 and then again using the compiler built in the first stage. You
1924 can also build a stage 3 compiler, but this normally isn't
1925 necessary except to verify that the stage 2 compiler is working
1928 <para>Note that when doing a bootstrap, the stage 1 compiler
1929 must be built, followed by the runtime system and libraries, and
1930 then the stage 2 compiler. The correct ordering is implemented
1931 by the top-level fptools <filename>Makefile</filename>, so if
1932 you want everything to work automatically it's best to start
1933 <command>make</command> from the top of the tree. When building
1934 GHC, the top-level fptools <filename>Makefile</filename> is set
1935 up to do a 2-stage bootstrap by default (when you say
1936 <command>make</command>). Some other targets it supports
1943 <para>Build everything as normal, including the stage 1
1951 <para>Build the stage 2 compiler only.</para>
1958 <para>Build the stage 3 compiler only.</para>
1963 <term>bootstrap</term> <term>bootstrap2</term>
1965 <para>Build stage 1 followed by stage 2.</para>
1970 <term>bootstrap3</term>
1972 <para>Build stages 1, 2 and 3.</para>
1977 <term>install</term>
1979 <para>Install everything, including the compiler built in
1980 stage 2. To override the stage, say <literal>make install
1981 stage=<replaceable>n</replaceable></literal> where
1982 <replaceable>n</replaceable> is the stage to install.</para>
1987 <para>The top-level <filename>Makefile</filename> also arranges
1988 to do the appropriate <literal>make boot</literal> steps (see
1989 below) before actually building anything.</para>
1991 <para>The <literal>stage1</literal>, <literal>stage2</literal>
1992 and <literal>stage3</literal> targets also work in the
1993 <literal>ghc/compiler</literal> directory, but don't forget that
1994 each stage requires its own <literal>make boot</literal> step:
1995 for example, you must do</para>
1997 <screen>$ make boot stage=2</screen>
1999 <para>before <literal>make stage2</literal> in
2000 <literal>ghc/compiler</literal>.</para>
2003 <sect2 id="sec-standard-targets">
2004 <title>Standard Targets</title>
2005 <indexterm><primary>targets, standard makefile</primary></indexterm>
2006 <indexterm><primary>makefile targets</primary></indexterm>
2008 <para>In any directory you should be able to make the following:</para>
2012 <term><literal>boot</literal></term>
2014 <para>does the one-off preparation required to get ready
2015 for the real work. Notably, it does <command>gmake
2016 depend</command> in all directories that contain programs.
2017 It also builds the necessary tools for compilation to
2020 <para>Invoking the <literal>boot</literal> target
2021 explicitly is not normally necessary. From the top-level
2022 <literal>fptools</literal> directory, invoking
2023 <literal>gmake</literal> causes <literal>gmake boot
2024 all</literal> to be invoked in each of the project
2025 subdirectories, in the order specified by
2026 <literal>$(AllTargets)</literal> in
2027 <literal>config.mk</literal>.</para>
2029 <para>If you're working in a subdirectory somewhere and
2030 need to update the dependencies, <literal>gmake
2031 boot</literal> is a good way to do it.</para>
2036 <term><literal>all</literal></term>
2038 <para>makes all the final target(s) for this Makefile.
2039 Depending on which directory you are in a “final
2040 target” may be an executable program, a library
2041 archive, a shell script, or a Postscript file. Typing
2042 <command>gmake</command> alone is generally the same as
2043 typing <command>gmake all</command>.</para>
2048 <term><literal>install</literal></term>
2050 <para>installs the things built by <literal>all</literal>
2051 (except for the documentation). Where does it install
2052 them? That is specified by
2053 <filename>mk/config.mk.in</filename>; you can override it
2054 in <filename>mk/build.mk</filename>, or by running
2055 <command>configure</command> with command-line arguments
2056 like <literal>--bindir=/home/simonpj/bin</literal>; see
2057 <literal>./configure --help</literal> for the full
2063 <term><literal>install-docs</literal></term>
2065 <para>installs the documentation. Otherwise behaves just
2066 like <literal>install</literal>.</para>
2071 <term><literal>uninstall</literal></term>
2073 <para>reverses the effect of
2074 <literal>install</literal>.</para>
2079 <term><literal>clean</literal></term>
2081 <para>Delete all files from the current directory that are
2082 normally created by building the program. Don't delete
2083 the files that record the configuration, or files
2084 generated by <command>gmake boot</command>. Also preserve
2085 files that could be made by building, but normally aren't
2086 because the distribution comes with them.</para>
2091 <term><literal>distclean</literal></term>
2093 <para>Delete all files from the current directory that are
2094 created by configuring or building the program. If you
2095 have unpacked the source and built the program without
2096 creating any other files, <literal>make
2097 distclean</literal> should leave only the files that were
2098 in the distribution.</para>
2103 <term><literal>mostlyclean</literal></term>
2105 <para>Like <literal>clean</literal>, but may refrain from
2106 deleting a few files that people normally don't want to
2112 <term><literal>maintainer-clean</literal></term>
2114 <para>Delete everything from the current directory that
2115 can be reconstructed with this Makefile. This typically
2116 includes everything deleted by
2117 <literal>distclean</literal>, plus more: C source files
2118 produced by Bison, tags tables, Info files, and so
2121 <para>One exception, however: <literal>make
2122 maintainer-clean</literal> should not delete
2123 <filename>configure</filename> even if
2124 <filename>configure</filename> can be remade using a rule
2125 in the <filename>Makefile</filename>. More generally,
2126 <literal>make maintainer-clean</literal> should not delete
2127 anything that needs to exist in order to run
2128 <filename>configure</filename> and then begin to build the
2134 <term><literal>check</literal></term>
2136 <para>run the test suite.</para>
2141 <para>All of these standard targets automatically recurse into
2142 sub-directories. Certain other standard targets do not:</para>
2146 <term><literal>configure</literal></term>
2148 <para>is only available in the root directory
2149 <constant>$(FPTOOLS_TOP)</constant>; it has
2150 been discussed in <XRef
2151 LinkEnd="sec-build-config">.</para>
2156 <term><literal>depend</literal></term>
2158 <para>make a <filename>.depend</filename> file in each
2159 directory that needs it. This <filename>.depend</filename>
2160 file contains mechanically-generated dependency
2161 information; for example, suppose a directory contains a
2162 Haskell source module <filename>Foo.lhs</filename> which
2163 imports another module <literal>Baz</literal>. Then the
2164 generated <filename>.depend</filename> file will contain
2165 the dependency:</para>
2171 <para>which says that the object file
2172 <filename>Foo.o</filename> depends on the interface file
2173 <filename>Baz.hi</filename> generated by compiling module
2174 <literal>Baz</literal>. The <filename>.depend</filename>
2175 file is automatically included by every Makefile.</para>
2180 <term><literal>binary-dist</literal></term>
2182 <para>make a binary distribution. This is the target we
2183 use to build the binary distributions of GHC and
2189 <term><literal>dist</literal></term>
2191 <para>make a source distribution. Note that this target
2192 does “make distclean” as part of its work;
2193 don't use it if you want to keep what you've built.</para>
2198 <para>Most <filename>Makefile</filename>s have targets other
2199 than these. You can discover them by looking in the
2200 <filename>Makefile</filename> itself.</para>
2204 <title>Using a project from the build tree</title>
2206 <para>If you want to build GHC (say) and just use it direct from
2207 the build tree without doing <literal>make install</literal>
2208 first, you can run the in-place driver script:
2209 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2211 <para> Do <emphasis>NOT</emphasis> use
2212 <filename>ghc/compiler/ghc</filename>, or
2213 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2214 scripts intended for installation, and contain hard-wired paths
2215 to the installed libraries, rather than the libraries in the
2218 <para>Happy can similarly be run from the build tree, using
2219 <filename>happy/src/happy-inplace</filename>.</para>
2223 <title>Fast Making</title>
2225 <indexterm><primary>fastmake</primary></indexterm>
2226 <indexterm><primary>dependencies, omitting</primary></indexterm>
2227 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2229 <para>Sometimes the dependencies get in the way: if you've made
2230 a small change to one file, and you're absolutely sure that it
2231 won't affect anything else, but you know that
2232 <command>make</command> is going to rebuild everything anyway,
2233 the following hack may be useful:</para>
2239 <para>This tells the make system to ignore dependencies and just
2240 build what you tell it to. In other words, it's equivalent to
2241 temporarily removing the <filename>.depend</filename> file in
2242 the current directory (where <command>mkdependHS</command> and
2243 friends store their dependency information).</para>
2245 <para>A bit of history: GHC used to come with a
2246 <command>fastmake</command> script that did the above job, but
2247 GNU make provides the features we need to do it without
2248 resorting to a script. Also, we've found that fastmaking is
2249 less useful since the advent of GHC's recompilation checker (see
2250 the User's Guide section on "Separate Compilation").</para>
2254 <sect1 id="sec-makefile-arch">
2255 <title>The <filename>Makefile</filename> architecture</title>
2256 <indexterm><primary>makefile architecture</primary></indexterm>
2258 <para><command>make</command> is great if everything
2259 works—you type <command>gmake install</command> and lo! the
2260 right things get compiled and installed in the right places. Our
2261 goal is to make this happen often, but somehow it often doesn't;
2262 instead some weird error message eventually emerges from the
2263 bowels of a directory you didn't know existed.</para>
2265 <para>The purpose of this section is to give you a road-map to
2266 help you figure out what is going right and what is going
2270 <title>Debugging</title>
2272 <para>Debugging <filename>Makefile</filename>s is something of a
2273 black art, but here's a couple of tricks that we find
2274 particularly useful. The following command allows you to see
2275 the contents of any make variable in the context of the current
2276 <filename>Makefile</filename>:</para>
2278 <screen>$ make show VALUE=HS_SRCS</screen>
2280 <para>where you can replace <literal>HS_SRCS</literal> with the
2281 name of any variable you wish to see the value of.</para>
2283 <para>GNU make has a <option>-d</option> option which generates
2284 a dump of the decision procedure used to arrive at a conclusion
2285 about which files should be recompiled. Sometimes useful for
2286 tracking down problems with superfluous or missing
2287 recompilations.</para>
2291 <title>A small project</title>
2293 <para>To get started, let us look at the
2294 <filename>Makefile</filename> for an imaginary small
2295 <literal>fptools</literal> project, <literal>small</literal>.
2296 Each project in <literal>fptools</literal> has its own directory
2297 in <constant>FPTOOLS_TOP</constant>, so the
2298 <literal>small</literal> project will have its own directory
2299 <constant>FPOOLS_TOP/small/</constant>. Inside the
2300 <filename>small/</filename> directory there will be a
2301 <filename>Makefile</filename>, looking something like
2304 <indexterm><primary>Makefile, minimal</primary></indexterm>
2307 # Makefile for fptools project "small"
2310 include $(TOP)/mk/boilerplate.mk
2312 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2315 include $(TOP)/target.mk
2318 <para>this <filename>Makefile</filename> has three
2323 <para>The first section includes
2326 One of the most important
2327 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2328 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2333 a file of “boilerplate” code from the level
2334 above (which in this case will be
2335 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2336 As its name suggests, <filename>boilerplate.mk</filename>
2337 consists of a large quantity of standard
2338 <filename>Makefile</filename> code. We discuss this
2339 boilerplate in more detail in <XRef LinkEnd="sec-boiler">.
2340 <indexterm><primary>include, directive in
2341 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2342 inclusion</primary></indexterm></para>
2344 <para>Before the <literal>include</literal> statement, you
2345 must define the <command>make</command> variable
2346 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2347 to be the directory containing the <filename>mk</filename>
2348 directory in which the <filename>boilerplate.mk</filename>
2349 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2352 include ../mk/boilerplate.mk # NO NO NO
2356 <para>Why? Because the <filename>boilerplate.mk</filename>
2357 file needs to know where it is, so that it can, in turn,
2358 <literal>include</literal> other files. (Unfortunately,
2359 when an <literal>include</literal>d file does an
2360 <literal>include</literal>, the filename is treated relative
2361 to the directory in which <command>gmake</command> is being
2362 run, not the directory in which the
2363 <literal>include</literal>d sits.) In general,
2364 <emphasis>every file <filename>foo.mk</filename> assumes
2366 <filename><constant>$(TOP)</constant>/mk/foo.mk</filename>
2367 refers to itself.</emphasis> It is up to the
2368 <filename>Makefile</filename> doing the
2369 <literal>include</literal> to ensure this is the case.</para>
2371 <para>Files intended for inclusion in other
2372 <filename>Makefile</filename>s are written to have the
2373 following property: <emphasis>after
2374 <filename>foo.mk</filename> is <literal>include</literal>d,
2375 it leaves <constant>TOP</constant> containing the same value
2376 as it had just before the <literal>include</literal>
2377 statement</emphasis>. In our example, this invariant
2378 guarantees that the <literal>include</literal> for
2379 <filename>target.mk</filename> will look in the same
2380 directory as that for <filename>boilerplate.mk</filename>.</para>
2384 <para> The second section defines the following standard
2385 <command>make</command> variables:
2386 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2387 (the source files from which is to be built), and
2388 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2389 (the executable binary to be built). We will discuss in
2390 more detail what the “standard variables” are,
2391 and how they affect what happens, in <XRef
2392 LinkEnd="sec-targets">.</para>
2394 <para>The definition for <constant>SRCS</constant> uses the
2395 useful GNU <command>make</command> construct
2396 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2397 which expands to a list of all the files matching the
2398 pattern <literal>pat</literal> in the current directory. In
2399 this example, <constant>SRCS</constant> is set to the list
2400 of all the <filename>.lhs</filename> and
2401 <filename>.c</filename> files in the directory. (Let's
2402 suppose there is one of each, <filename>Foo.lhs</filename>
2403 and <filename>Baz.c</filename>.)</para>
2407 <para>The last section includes a second file of standard
2409 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2410 It contains the rules that tell <command>gmake</command> how
2411 to make the standard targets (<Xref
2412 LinkEnd="sec-standard-targets">). Why, you ask, can't this
2413 standard code be part of
2414 <filename>boilerplate.mk</filename>? Good question. We
2415 discuss the reason later, in <Xref
2416 LinkEnd="sec-boiler-arch">.</para>
2418 <para>You do not <emphasis>have</emphasis> to
2419 <literal>include</literal> the
2420 <filename>target.mk</filename> file. Instead, you can write
2421 rules of your own for all the standard targets. Usually,
2422 though, you will find quite a big payoff from using the
2423 canned rules in <filename>target.mk</filename>; the price
2424 tag is that you have to understand what canned rules get
2425 enabled, and what they do (<Xref
2426 LinkEnd="sec-targets">).</para>
2430 <para>In our example <filename>Makefile</filename>, most of the
2431 work is done by the two <literal>include</literal>d files. When
2432 you say <command>gmake all</command>, the following things
2437 <para><command>gmake</command> figures out that the object
2438 files are <filename>Foo.o</filename> and
2439 <filename>Baz.o</filename>.</para>
2443 <para>It uses a boilerplate pattern rule to compile
2444 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2445 using a Haskell compiler. (Which one? That is set in the
2446 build configuration.)</para>
2450 <para>It uses another standard pattern rule to compile
2451 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2452 using a C compiler. (Ditto.)</para>
2456 <para>It links the resulting <filename>.o</filename> files
2457 together to make <literal>small</literal>, using the Haskell
2458 compiler to do the link step. (Why not use
2459 <command>ld</command>? Because the Haskell compiler knows
2460 what standard libraries to link in. How did
2461 <command>gmake</command> know to use the Haskell compiler to
2462 do the link, rather than the C compiler? Because we set the
2463 variable <constant>HS_PROG</constant> rather than
2464 <constant>C_PROG</constant>.)</para>
2468 <para>All <filename>Makefile</filename>s should follow the above
2469 three-section format.</para>
2473 <title>A larger project</title>
2475 <para>Larger projects are usually structured into a number of
2476 sub-directories, each of which has its own
2477 <filename>Makefile</filename>. (In very large projects, this
2478 sub-structure might be iterated recursively, though that is
2479 rare.) To give you the idea, here's part of the directory
2480 structure for the (rather large) GHC project:</para>
2490 ...source files for documentation...
2493 ...source files for driver...
2496 parser/...source files for parser...
2497 renamer/...source files for renamer...
2501 <para>The sub-directories <filename>docs</filename>,
2502 <filename>driver</filename>, <filename>compiler</filename>, and
2503 so on, each contains a sub-component of GHC, and each has its
2504 own <filename>Makefile</filename>. There must also be a
2505 <filename>Makefile</filename> in
2506 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>.
2507 It does most of its work by recursively invoking
2508 <command>gmake</command> on the <filename>Makefile</filename>s
2509 in the sub-directories. We say that
2510 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2511 <filename>Makefile</filename></emphasis>, because it does little
2512 except organise its children, while the
2513 <filename>Makefile</filename>s in the sub-directories are all
2514 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2515 principle the sub-directories might themselves contain a
2516 non-leaf <filename>Makefile</filename> and several
2517 sub-sub-directories, but that does not happen in GHC.)</para>
2519 <para>The <filename>Makefile</filename> in
2520 <filename>ghc/compiler</filename> is considered a leaf
2521 <filename>Makefile</filename> even though the
2522 <filename>ghc/compiler</filename> has sub-directories, because
2523 these sub-directories do not themselves have
2524 <filename>Makefile</filename>s in them. They are just used to
2525 structure the collection of modules that make up GHC, but all
2526 are managed by the single <filename>Makefile</filename> in
2527 <filename>ghc/compiler</filename>.</para>
2529 <para>You will notice that <filename>ghc/</filename> also
2530 contains a directory <filename>ghc/mk/</filename>. It contains
2531 GHC-specific <filename>Makefile</filename> boilerplate code.
2532 More precisely:</para>
2536 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2537 at the top of <filename>ghc/Makefile</filename>, and of all
2538 the leaf <filename>Makefile</filename>s in the
2539 sub-directories. It in turn <literal>include</literal>s the
2540 main boilerplate file
2541 <filename>mk/boilerplate.mk</filename>.</para>
2545 <para><filename>ghc/mk/target.mk</filename> is
2546 <literal>include</literal>d at the bottom of
2547 <filename>ghc/Makefile</filename>, and of all the leaf
2548 <filename>Makefile</filename>s in the sub-directories. It
2549 in turn <literal>include</literal>s the file
2550 <filename>mk/target.mk</filename>.</para>
2554 <para>So these two files are the place to look for GHC-wide
2555 customisation of the standard boilerplate.</para>
2558 <sect2 id="sec-boiler-arch">
2559 <title>Boilerplate architecture</title>
2560 <indexterm><primary>boilerplate architecture</primary></indexterm>
2562 <para>Every <filename>Makefile</filename> includes a
2563 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2564 file at the top, and
2565 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2566 file at the bottom. In this section we discuss what is in these
2567 files, and why there have to be two of them. In general:</para>
2571 <para><filename>boilerplate.mk</filename> consists of:</para>
2575 <para><emphasis>Definitions of millions of
2576 <command>make</command> variables</emphasis> that
2577 collectively specify the build configuration. Examples:
2578 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2579 the options to feed to the Haskell compiler;
2580 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2581 the sub-directories to enable within the
2582 <literal>nofib</literal> project;
2583 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2584 the name of the Haskell compiler to use when compiling
2585 GHC in the <literal>ghc</literal> project.</para>
2589 <para><emphasis>Standard pattern rules</emphasis> that
2590 tell <command>gmake</command> how to construct one file
2591 from another.</para>
2595 <para><filename>boilerplate.mk</filename> needs to be
2596 <literal>include</literal>d at the <emphasis>top</emphasis>
2597 of each <filename>Makefile</filename>, so that the user can
2598 replace the boilerplate definitions or pattern rules by
2599 simply giving a new definition or pattern rule in the
2600 <filename>Makefile</filename>. <command>gmake</command>
2601 simply takes the last definition as the definitive one.</para>
2603 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2604 definitions, it is also quite common to
2605 <emphasis>augment</emphasis> them. For example, a
2606 <filename>Makefile</filename> might say:</para>
2612 <para>thereby adding “<option>-O</option>” to
2614 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2618 <para><filename>target.mk</filename> contains
2619 <command>make</command> rules for the standard targets
2620 described in <Xref LinkEnd="sec-standard-targets">. These
2621 rules are selectively included, depending on the setting of
2622 certain <command>make</command> variables. These variables
2623 are usually set in the middle section of the
2624 <filename>Makefile</filename> between the two
2625 <literal>include</literal>s.</para>
2627 <para><filename>target.mk</filename> must be included at the
2628 end (rather than being part of
2629 <filename>boilerplate.mk</filename>) for several tiresome
2635 <para><command>gmake</command> commits target and
2636 dependency lists earlier than it should. For example,
2637 <FIlename>target.mk</FIlename> has a rule that looks
2641 $(HS_PROG) : $(OBJS)
2642 $(HC) $(LD_OPTS) $< -o $@
2645 <para>If this rule was in
2646 <filename>boilerplate.mk</filename> then
2647 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2649 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2650 would not have their final values at the moment
2651 <command>gmake</command> encountered the rule. Alas,
2652 <command>gmake</command> takes a snapshot of their
2653 current values, and wires that snapshot into the rule.
2654 (In contrast, the commands executed when the rule
2655 “fires” are only substituted at the moment
2656 of firing.) So, the rule must follow the definitions
2657 given in the <filename>Makefile</filename> itself.</para>
2661 <para>Unlike pattern rules, ordinary rules cannot be
2662 overriden or replaced by subsequent rules for the same
2663 target (at least, not without an error message).
2664 Including ordinary rules in
2665 <filename>boilerplate.mk</filename> would prevent the
2666 user from writing rules for specific targets in specific
2671 <para>There are a couple of other reasons I've
2672 forgotten, but it doesn't matter too much.</para>
2679 <sect2 id="sec-boiler">
2680 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2681 <indexterm><primary>boilerplate.mk</primary></indexterm>
2683 <para>If you look at
2684 <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename>
2685 you will find that it consists of the following sections, each
2686 held in a separate file:</para>
2690 <term><filename>config.mk</filename></term>
2691 <indexterm><primary>config.mk</primary></indexterm>
2693 <para>is the build configuration file we discussed at
2694 length in <Xref LinkEnd="sec-build-config">.</para>
2699 <term><filename>paths.mk</filename></term>
2700 <indexterm><primary>paths.mk</primary></indexterm>
2702 <para>defines <command>make</command> variables for
2703 pathnames and file lists. This file contains code for
2704 automatically compiling lists of source files and deriving
2705 lists of object files from those. The results can be
2706 overriden in the <filename>Makefile</filename>, but in
2707 most cases the automatic setup should do the right
2710 <para>The following variables may be set in the
2711 <filename>Makefile</filename> to affect how the automatic
2712 source file search is done:</para>
2716 <term><literal>ALL_DIRS</literal></term>
2717 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2720 <para>Set to a list of directories to search in
2721 addition to the current directory for source
2727 <term><literal>EXCLUDE_SRCS</literal></term>
2728 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2731 <para>Set to a list of source files (relative to the
2732 current directory) to omit from the automatic
2733 search. The source searching machinery is clever
2734 enough to know that if you exclude a source file
2735 from which other sources are derived, then the
2736 derived sources should also be excluded. For
2737 example, if you set <literal>EXCLUDED_SRCS</literal>
2738 to include <filename>Foo.y</filename>, then
2739 <filename>Foo.hs</filename> will also be
2745 <term><literal>EXTRA_SRCS</literal></term>
2746 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2749 <para>Set to a list of extra source files (perhaps
2750 in directories not listed in
2751 <literal>ALL_DIRS</literal>) that should be
2757 <para>The results of the automatic source file search are
2758 placed in the following make variables:</para>
2762 <term><literal>SRCS</literal></term>
2763 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2765 <para>All source files found, sorted and without
2766 duplicates, including those which might not exist
2767 yet but will be derived from other existing sources.
2768 <literal>SRCS</literal> <emphasis>can</emphasis> be
2769 overriden if necessary, in which case the variables
2770 below will follow suit.</para>
2775 <term><literal>HS_SRCS</literal></term>
2776 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2778 <para>all Haskell source files in the current
2779 directory, including those derived from other source
2780 files (eg. Happy sources also give rise to Haskell
2786 <term><literal>HS_OBJS</literal></term>
2787 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2789 <para>Object files derived from
2790 <literal>HS_SRCS</literal>.</para>
2795 <term><literal>HS_IFACES</literal></term>
2796 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2798 <para>Interface files (<literal>.hi</literal> files)
2799 derived from <literal>HS_SRCS</literal>.</para>
2804 <term><literal>C_SRCS</literal></term>
2805 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2807 <para>All C source files found.</para>
2812 <term><literal>C_OBJS</literal></term>
2813 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2815 <para>Object files derived from
2816 <literal>C_SRCS</literal>.</para>
2821 <term><literal>SCRIPT_SRCS</literal></term>
2822 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2824 <para>All script source files found
2825 (<literal>.lprl</literal> files).</para>
2830 <term><literal>SCRIPT_OBJS</literal></term>
2831 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2833 <para><quote>object</quote> files derived from
2834 <literal>SCRIPT_SRCS</literal>
2835 (<literal>.prl</literal> files).</para>
2840 <term><literal>HSC_SRCS</literal></term>
2841 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2843 <para>All <literal>hsc2hs</literal> source files
2844 (<literal>.hsc</literal> files).</para>
2849 <term><literal>HAPPY_SRCS</literal></term>
2850 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2852 <para>All <literal>happy</literal> source files
2853 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2858 <term><literal>OBJS</literal></term>
2859 <indexterm><primary>OBJS</primary></indexterm>
2861 <para>the concatenation of
2862 <literal>$(HS_OBJS)</literal>,
2863 <literal>$(C_OBJS)</literal>, and
2864 <literal>$(SCRIPT_OBJS)</literal>.</para>
2869 <para>Any or all of these definitions can easily be
2870 overriden by giving new definitions in your
2871 <filename>Makefile</filename>.</para>
2873 <para>What, exactly, does <filename>paths.mk</filename>
2874 consider a <quote>source file</quote> to be? It's based
2875 on the file's suffix (e.g. <filename>.hs</filename>,
2876 <filename>.lhs</filename>, <filename>.c</filename>,
2877 <filename>.hy</filename>, etc), but this is the kind of
2878 detail that changes, so rather than enumerate the source
2879 suffices here the best thing to do is to look in
2880 <filename>paths.mk</filename>.</para>
2885 <term><filename>opts.mk</filename></term>
2886 <indexterm><primary>opts.mk</primary></indexterm>
2888 <para>defines <command>make</command> variables for option
2889 strings to pass to each program. For example, it defines
2890 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2891 the option strings to pass to the Haskell compiler. See
2892 <Xref LinkEnd="sec-suffix">.</para>
2897 <term><filename>suffix.mk</filename></term>
2898 <indexterm><primary>suffix.mk</primary></indexterm>
2900 <para>defines standard pattern rules—see <Xref
2901 LinkEnd="sec-suffix">.</para>
2906 <para>Any of the variables and pattern rules defined by the
2907 boilerplate file can easily be overridden in any particular
2908 <filename>Makefile</filename>, because the boilerplate
2909 <literal>include</literal> comes first. Definitions after this
2910 <literal>include</literal> directive simply override the default
2911 ones in <filename>boilerplate.mk</filename>.</para>
2914 <sect2 id="sec-suffix">
2915 <title>Pattern rules and options</title>
2916 <indexterm><primary>Pattern rules</primary></indexterm>
2919 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2920 defines standard <emphasis>pattern rules</emphasis> that say how
2921 to build one kind of file from another, for example, how to
2922 build a <filename>.o</filename> file from a
2923 <filename>.c</filename> file. (GNU <command>make</command>'s
2924 <emphasis>pattern rules</emphasis> are more powerful and easier
2925 to use than Unix <command>make</command>'s <emphasis>suffix
2926 rules</emphasis>.)</para>
2928 <para>Almost all the rules look something like this:</para>
2933 $(CC) $(CC_OPTS) -c $< -o $@
2936 <para>Here's how to understand the rule. It says that
2937 <emphasis>something</emphasis><filename>.o</filename> (say
2938 <filename>Foo.o</filename>) can be built from
2939 <emphasis>something</emphasis><filename>.c</filename>
2940 (<filename>Foo.c</filename>), by invoking the C compiler (path
2941 name held in <constant>$(CC)</constant>), passing to it
2942 the options <constant>$(CC_OPTS)</constant> and
2943 the rule's dependent file of the rule
2944 <literal>$<</literal> (<filename>Foo.c</filename> in
2945 this case), and putting the result in the rule's target
2946 <literal>$@</literal> (<filename>Foo.o</filename> in this
2949 <para>Every program is held in a <command>make</command>
2950 variable defined in <filename>mk/config.mk</filename>—look
2951 in <filename>mk/config.mk</filename> for the complete list. One
2952 important one is the Haskell compiler, which is called
2953 <constant>$(HC)</constant>.</para>
2955 <para>Every program's options are are held in a
2956 <command>make</command> variables called
2957 <constant><prog>_OPTS</constant>. the
2958 <constant><prog>_OPTS</constant> variables are
2959 defined in <filename>mk/opts.mk</filename>. Almost all of them
2960 are defined like this:</para>
2963 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
2966 <para>The four variables from which
2967 <constant>CC_OPTS</constant> is built have the following
2972 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
2974 <para>options passed to all C compilations.</para>
2979 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
2981 <para>options passed to C compilations for way
2982 <literal><way></literal>. For example,
2983 <constant>WAY_mp_CC_OPTS</constant>
2984 gives options to pass to the C compiler when compiling way
2985 <literal>mp</literal>. The variable
2986 <constant>WAY_CC_OPTS</constant> holds
2987 options to pass to the C compiler when compiling the
2988 standard way. (<Xref LinkEnd="sec-ways"> dicusses
2989 multi-way compilation.)</para>
2994 <term><constant><module>_CC_OPTS</constant>:</term>
2996 <para>options to pass to the C compiler that are specific
2997 to module <literal><module></literal>. For example,
2998 <constant>SMap_CC_OPTS</constant> gives the
2999 specific options to pass to the C compiler when compiling
3000 <filename>SMap.c</filename>.</para>
3005 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3007 <para>extra options to pass to all C compilations. This
3008 is intended for command line use, thus:</para>
3011 gmake libHS.a EXTRA_CC_OPTS="-v"
3018 <sect2 id="sec-targets">
3019 <title>The main <filename>mk/target.mk</filename> file</title>
3020 <indexterm><primary>target.mk</primary></indexterm>
3022 <para><filename>target.mk</filename> contains canned rules for
3023 all the standard targets described in <Xref
3024 LinkEnd="sec-standard-targets">. It is complicated by the fact
3025 that you don't want all of these rules to be active in every
3026 <filename>Makefile</filename>. Rather than have a plethora of
3027 tiny files which you can include selectively, there is a single
3028 file, <filename>target.mk</filename>, which selectively includes
3029 rules based on whether you have defined certain variables in
3030 your <filename>Makefile</filename>. This section explains what
3031 rules you get, what variables control them, and what the rules
3032 do. Hopefully, you will also get enough of an idea of what is
3033 supposed to happen that you can read and understand any weird
3034 special cases yourself.</para>
3038 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3040 <para>If <constant>HS_PROG</constant> is defined,
3041 you get rules with the following targets:</para>
3045 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3047 <para>itself. This rule links
3048 <constant>$(OBJS)</constant> with the Haskell
3049 runtime system to get an executable called
3050 <constant>$(HS_PROG)</constant>.</para>
3055 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3058 <constant>$(HS_PROG)</constant> in
3059 <constant>$(bindir)</constant>.</para>
3068 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3070 <para>is similar to <constant>HS_PROG</constant>,
3071 except that the link step links
3072 <constant>$(C_OBJS)</constant> with the C
3073 runtime system.</para>
3078 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3080 <para>is similar to <constant>HS_PROG</constant>,
3081 except that it links
3082 <constant>$(LIB_OBJS)</constant> to make the
3083 library archive <constant>$(LIBRARY)</constant>,
3084 and <literal>install</literal> installs it in
3085 <constant>$(libdir)</constant>.</para>
3090 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3092 <para>…</para>
3097 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3099 <para>…</para>
3104 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3106 <para>If <constant>HS_SRCS</constant> is defined
3107 and non-empty, a rule for the target
3108 <literal>depend</literal> is included, which generates
3109 dependency information for Haskell programs. Similarly
3110 for <constant>C_SRCS</constant>.</para>
3115 <para>All of these rules are “double-colon” rules,
3119 install :: $(HS_PROG)
3120 ...how to install it...
3123 <para>GNU <command>make</command> treats double-colon rules as
3124 separate entities. If there are several double-colon rules for
3125 the same target it takes each in turn and fires it if its
3126 dependencies say to do so. This means that you can, for
3127 example, define both <constant>HS_PROG</constant> and
3128 <constant>LIBRARY</constant>, which will generate two rules for
3129 <literal>install</literal>. When you type <command>gmake
3130 install</command> both rules will be fired, and both the program
3131 and the library will be installed, just as you wanted.</para>
3134 <sect2 id="sec-subdirs">
3135 <title>Recursion</title>
3136 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3137 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3139 <para>In leaf <filename>Makefile</filename>s the variable
3140 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3141 is undefined. In non-leaf <filename>Makefile</filename>s,
3142 <constant>SUBDIRS</constant> is set to the list of
3143 sub-directories that contain subordinate
3144 <filename>Makefile</filename>s. <emphasis>It is up to you to
3145 set <constant>SUBDIRS</constant> in the
3146 <filename>Makefile</filename>.</emphasis> There is no automation
3147 here—<constant>SUBDIRS</constant> is too important to
3150 <para>When <constant>SUBDIRS</constant> is defined,
3151 <filename>target.mk</filename> includes a rather neat rule for
3152 the standard targets (<Xref LinkEnd="sec-standard-targets"> that
3153 simply invokes <command>make</command> recursively in each of
3154 the sub-directories.</para>
3156 <para><emphasis>These recursive invocations are guaranteed to
3157 occur in the order in which the list of directories is specified
3158 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3159 be important. For example, when you say <command>gmake
3160 boot</command> it can be important that the recursive invocation
3161 of <command>make boot</command> is done in one sub-directory
3162 (the include files, say) before another (the source files).
3163 Generally, put the most independent sub-directory first, and the
3164 most dependent last.</para>
3167 <sect2 id="sec-ways">
3168 <title>Way management</title>
3169 <indexterm><primary>way management</primary></indexterm>
3171 <para>We sometimes want to build essentially the same system in
3172 several different “ways”. For example, we want to build GHC's
3173 <literal>Prelude</literal> libraries with and without profiling,
3174 so that there is an appropriately-built library archive to link
3175 with when the user compiles his program. It would be possible
3176 to have a completely separate build tree for each such “way”,
3177 but it would be horribly bureaucratic, especially since often
3178 only parts of the build tree need to be constructed in multiple
3182 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3183 contains some clever magic to allow you to build several
3184 versions of a system; and to control locally how many versions
3185 are built and how they differ. This section explains the
3188 <para>The files for a particular way are distinguished by
3189 munging the suffix. The <quote>normal way</quote> is always
3190 built, and its files have the standard suffices
3191 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3192 In addition, you can build one or more extra ways, each
3193 distinguished by a <emphasis>way tag</emphasis>. The object
3194 files and interface files for one of these extra ways are
3195 distinguished by their suffix. For example, way
3196 <literal>mp</literal> has files
3197 <filename>.mp_o</filename> and
3198 <filename>.mp_hi</filename>. Library archives have their
3199 way tag the other side of the dot, for boring reasons; thus,
3200 <filename>libHS_mp.a</filename>.</para>
3202 <para>A <command>make</command> variable called
3203 <constant>way</constant> holds the current way tag.
3204 <emphasis><constant>way</constant> is only ever set on the
3205 command line of <command>gmake</command></emphasis> (usually in
3206 a recursive invocation of <command>gmake</command> by the
3207 system). It is never set inside a
3208 <filename>Makefile</filename>. So it is a global constant for
3209 any one invocation of <command>gmake</command>. Two other
3210 <command>make</command> variables,
3211 <constant>way_</constant> and
3212 <constant>_way</constant> are immediately derived from
3213 <constant>$(way)</constant> and never altered. If
3214 <constant>way</constant> is not set, then neither are
3215 <constant>way_</constant> and
3216 <constant>_way</constant>, and the invocation of
3217 <command>make</command> will build the <quote>normal
3218 way</quote>. If <constant>way</constant> is set, then the other
3219 two variables are set in sympathy. For example, if
3220 <constant>$(way)</constant> is “<literal>mp</literal>”,
3221 then <constant>way_</constant> is set to
3222 “<literal>mp_</literal>” and
3223 <constant>_way</constant> is set to
3224 “<literal>_mp</literal>”. These three variables are
3225 then used when constructing file names.</para>
3227 <para>So how does <command>make</command> ever get recursively
3228 invoked with <constant>way</constant> set? There are two ways
3229 in which this happens:</para>
3233 <para>For some (but not all) of the standard targets, when
3234 in a leaf sub-directory, <command>make</command> is
3235 recursively invoked for each way tag in
3236 <constant>$(WAYS)</constant>. You set
3237 <constant>WAYS</constant> in the
3238 <filename>Makefile</filename> to the list of way tags you
3239 want these targets built for. The mechanism here is very
3240 much like the recursive invocation of
3241 <command>make</command> in sub-directories (<Xref
3242 LinkEnd="sec-subdirs">). It is up to you to set
3243 <constant>WAYS</constant> in your
3244 <filename>Makefile</filename>; this is how you control what
3245 ways will get built.</para>
3249 <para>For a useful collection of targets (such as
3250 <filename>libHS_mp.a</filename>,
3251 <filename>Foo.mp_o</filename>) there is a rule which
3252 recursively invokes <command>make</command> to make the
3253 specified target, setting the <constant>way</constant>
3254 variable. So if you say <command>gmake
3255 Foo.mp_o</command> you should see a recursive
3256 invocation <command>gmake Foo.mp_o way=mp</command>,
3257 and <emphasis>in this recursive invocation the pattern rule
3258 for compiling a Haskell file into a <filename>.o</filename>
3259 file will match</emphasis>. The key pattern rules (in
3260 <filename>suffix.mk</filename>) look like this:
3264 $(HC) $(HC_OPTS) $< -o $@
3271 <para>You can invoke <command>make</command> with a
3272 particular <literal>way</literal> setting yourself, in order
3273 to build files related to a particular
3274 <literal>way</literal> in the current directory. eg.
3280 will build files for the profiling way only in the current
3287 <title>When the canned rule isn't right</title>
3289 <para>Sometimes the canned rule just doesn't do the right thing.
3290 For example, in the <literal>nofib</literal> suite we want the
3291 link step to print out timing information. The thing to do here
3292 is <emphasis>not</emphasis> to define
3293 <constant>HS_PROG</constant> or
3294 <constant>C_PROG</constant>, and instead define a special
3295 purpose rule in your own <filename>Makefile</filename>. By
3296 using different variable names you will avoid the canned rules
3297 being included, and conflicting with yours.</para>
3301 <sect1 id="building-docs">
3302 <title>Building the documentation</title>
3304 <sect2 id="pre-supposed-doc-tools">
3305 <title>Tools for building the Documentation</title>
3307 <para>The following additional tools are required if you want to
3308 format the documentation that comes with the
3309 <literal>fptools</literal> projects:</para>
3313 <term>DocBook</term>
3314 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3315 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3317 <para>Much of our documentation is written in SGML, using
3318 the DocBook DTD. Instructions on installing and
3319 configuring the DocBook tools are below.</para>
3325 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3326 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3328 <para>A decent TeX distribution is required if you want to
3329 produce printable documentation. We recomment teTeX,
3330 which includes just about everything you need.</para>
3335 <term>Haddock</term>
3336 <indexterm><primary>Haddock</primary>
3339 <para>Haddock is a Haskell documentation tool that we use
3340 for automatically generating documentation from the
3341 library source code. It is an <literal>fptools</literal>
3342 project in itself. To build documentation for the
3343 libraries (<literal>fptools/libraries</literal>) you
3344 should check out and build Haddock in
3345 <literal>fptools/haddock</literal>. Haddock requires GHC
3353 <title>Installing the DocBook tools</title>
3356 <title>Installing the DocBook tools on Linux</title>
3358 <para>If you're on a recent RedHat system (7.0+), you probably
3359 have working DocBook tools already installed. The configure
3360 script should detect your setup and you're away.</para>
3362 <para>If you don't have DocBook tools installed, and you are
3363 using a system that can handle RedHat RPM packages, you can
3364 probably use the <ULink
3365 URL="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3366 DocBook tools</ULink>, which is the most shrink-wrapped SGML
3367 suite that we could find. You need all the RPMs except for
3368 psgml (i.e. <Filename>docbook</Filename>,
3369 <Filename>jade</Filename>, <Filename>jadetex</Filename>,
3370 <Filename>sgmlcommon</Filename> and
3371 <Filename>stylesheets</Filename>). Note that most of these
3372 RPMs are architecture neutral, so are likely to be found in a
3373 <Filename>noarch</Filename> directory. The SuSE RPMs also
3374 work; the RedHat ones <Emphasis>don't</Emphasis> in RedHat 6.2
3375 (7.0 and later should be OK), but they are easy to fix: just
3377 <Filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</Filename>
3378 to <Filename>/usr/lib/sgml/lib/dblib.dsl</Filename>. </para>
3382 <title>Installing DocBook on FreeBSD</title>
3384 <para>On FreeBSD systems, the easiest way to get DocBook up
3385 and running is to install it from the ports tree or a
3386 pre-compiled package (packages are available from your local
3387 FreeBSD mirror site).</para>
3389 <para>To use the ports tree, do this:
3391 $ cd /usr/ports/textproc/docproj
3394 This installs the FreeBSD documentation project tools, which
3395 includes everything needed to format the GHC
3396 documentation.</para>
3400 <title>Installing from binaries on Windows</title>
3402 <Para>It's a good idea to use Norman Walsh's <ULink
3403 URL="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3404 notes</ULink> as a guide. You should get version 3.1 of
3405 DocBook, and note that his file <Filename>test.sgm</Filename>
3406 won't work, as it needs version 3.0. You should unpack Jade
3407 into <Filename>\Jade</Filename>, along with the entities,
3408 DocBook into <Filename>\docbook</Filename>, and the DocBook
3409 stylesheets into <Filename>\docbook\stylesheets</Filename> (so
3410 they actually end up in
3411 <Filename>\docbook\stylesheets\docbook</Filename>).</para>
3416 <title>Installing the DocBook tools from source</title>
3421 <para>Install <ULink
3422 URL="http://openjade.sourceforge.net/">OpenJade</ULink>
3423 (Windows binaries are available as well as sources). If you
3424 want DVI, PS, or PDF then install JadeTeX from the
3425 <Filename>dsssl</Filename> subdirectory. (If you get the
3429 ! LaTeX Error: Unknown option implicit=false' for package hyperref'.
3432 your version of <Command>hyperref</Command> is out of date;
3433 download it from CTAN
3434 (<Filename>macros/latex/contrib/supported/hyperref</Filename>),
3435 and make it, ensuring that you have first removed or renamed
3436 your old copy. If you start getting file not found errors
3437 when making the test for <Command>hyperref</Command>, you
3438 can abort at that point and proceed straight to
3439 <Command>make install</Command>, or enter them as
3440 <Filename>../</Filename><Emphasis>filename</Emphasis>.)</para>
3442 <para>Make links from <Filename>virtex</Filename> to
3443 <Filename>jadetex</Filename> and
3444 <Filename>pdfvirtex</Filename> to
3445 <Filename>pdfjadetex</Filename> (otherwise DVI, PostScript
3446 and PDF output will not work). Copy
3447 <Filename>dsssl/*.{dtd,dsl}</Filename> and
3448 <Filename>catalog</Filename> to
3449 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3453 <title>DocBook and the DocBook stylesheets</title>
3455 <para>Get a Zip of <ULink
3456 URL="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ULink>
3457 and install the contents in
3458 <Filename>/usr/[local/]/lib/sgml</Filename>.</para>
3460 <para>Get the <ULink
3461 URL="http://nwalsh.com/docbook/dsssl/">DocBook
3462 stylesheets</ULink> and install in
3463 <Filename>/usr/[local/]lib/sgml/stylesheets</Filename>
3464 (thereby creating a subdirectory docbook). For indexing,
3465 copy or link <Filename>collateindex.pl</Filename> from the
3466 DocBook stylesheets archive in <Filename>bin</Filename> into
3467 a directory on your <Constant>PATH</Constant>.</para>
3469 <para>Download the <ULink
3470 URL="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3471 entities</ULink> into
3472 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3478 <title>Configuring the DocBook tools</title>
3480 <Para>Once the DocBook tools are installed, the configure script
3481 will detect them and set up the build system accordingly. If you
3482 have a system that isn't supported, let us know, and we'll try
3487 <title>Remaining problems</title>
3489 <para>If you install from source, you'll get a pile of warnings
3492 <Screen>DTDDECL catalog entries are not supported</Screen>
3494 every time you build anything. These can safely be ignored, but
3495 if you find them tedious you can get rid of them by removing all
3496 the <Constant>DTDDECL</Constant> entries from
3497 <Filename>docbook.cat</Filename>.</para>
3501 <title>Building the documentation</title>
3503 <para>To build documentation in a certain format, you can
3504 say, for example,</para>
3510 <para>to build HTML documentation below the current directory.
3511 The available formats are: <literal>dvi</literal>,
3512 <literal>ps</literal>, <literal>pdf</literal>,
3513 <literal>html</literal>, and <literal>rtf</literal>. Note that
3514 not all documentation can be built in all of these formats: HTML
3515 documentation is generally supported everywhere, and DocBook
3516 documentation might support the other formats (depending on what
3517 other tools you have installed).</para>
3519 <para>All of these targets are recursive; that is, saying
3520 <literal>make html</literal> will make HTML docs for all the
3521 documents recursively below the current directory.</para>
3523 <para>Because there are many different formats that the DocBook
3524 documentation can be generated in, you have to select which ones
3525 you want by setting the <literal>SGMLDocWays</literal> variable
3526 to a list of them. For example, in
3527 <filename>build.mk</filename> you might have a line:</para>
3530 SGMLDocWays = html ps
3533 <para>This will cause the documentation to be built in the requested
3534 formats as part of the main build (the default is not to build
3535 any documentation at all).</para>
3539 <title>Installing the documentation</title>
3541 <para>To install the documentation, use:</para>
3547 <para>This will install the documentation into
3548 <literal>$(datadir)</literal> (which defaults to
3549 <literal>$(prefix)/share</literal>). The exception is HTML
3550 documentation, which goes into
3551 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3553 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3554 to a list of formats, the <literal>install-docs</literal> target
3555 won't do anything for SGML documentation.</para>
3561 <sect1 id="sec-porting-ghc">
3562 <title>Porting GHC</title>
3564 <para>This section describes how to port GHC to a currenly
3565 unsupported platform. There are two distinct
3566 possibilities:</para>
3570 <para>The hardware architecture for your system is already
3571 supported by GHC, but you're running an OS that isn't
3572 supported (or perhaps has been supported in the past, but
3573 currently isn't). This is the easiest type of porting job,
3574 but it still requires some careful bootstrapping. Proceed to
3575 <xref linkend="sec-booting-from-hc">.</para>
3579 <para>Your system's hardware architecture isn't supported by
3580 GHC. This will be a more difficult port (though by comparison
3581 perhaps not as difficult as porting gcc). Proceed to <xref
3582 linkend="unregisterised-porting">.</para>
3586 <sect2 id="sec-booting-from-hc">
3587 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3589 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3590 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3591 <indexterm><primary>porting GHC</primary></indexterm>
3593 <para>Bootstrapping GHC on a system without GHC already
3594 installed is achieved by taking the intermediate C files (known
3595 as HC files) from a GHC compilation on a supported system to the
3596 target machine, and compiling them using gcc to get a working
3599 <para><emphasis>NOTE: GHC versions 5.xx and later are
3600 significantly harder to bootstrap from C than earlier versions.
3601 We recommend starting from version 4.08.2 if you need to
3602 bootstrap in this way.</emphasis></para>
3604 <para>HC files are architecture-dependent (but not
3605 OS-dependent), so you have to get a set that were generated on
3606 similar hardware. There may be some supplied on the GHC
3607 download page, otherwise you'll have to compile some up
3608 yourself, or start from <emphasis>unregisterised</emphasis> HC
3609 files - see <xref linkend="unregisterised-porting">.</para>
3611 <para>The following steps should result in a working GHC build
3612 with full libraries:</para>
3616 <para>Unpack the HC files on top of a fresh source tree
3617 (make sure the source tree version matches the version of
3618 the HC files <emphasis>exactly</emphasis>!). This will
3619 place matching <filename>.hc</filename> files next to the
3620 corresponding Haskell source (<filename>.hs</filename> or
3621 <filename>.lhs</filename>) in the compiler subdirectory
3622 <filename>ghc/compiler</filename> and in the libraries
3623 (subdirectories of <filename>hslibs</filename> and
3624 <literal>libraries</literal>).</para>
3628 <para>The actual build process is fully automated by the
3629 <filename>hc-build</filename> script located in the
3630 <filename>distrib</filename> directory. If you eventually
3631 want to install GHC into the directory
3632 <replaceable>dir</replaceable>, the following
3633 command will execute the whole build process (it won't
3634 install yet):</para>
3637 foo% distrib/hc-build --prefix=<replaceable>dir</replaceable>
3639 <indexterm><primary>--hc-build</primary></indexterm>
3641 <para>By default, the installation directory is
3642 <filename>/usr/local</filename>. If that is what you want,
3643 you may omit the argument to <filename>hc-build</filename>.
3644 Generally, any option given to <filename>hc-build</filename>
3645 is passed through to the configuration script
3646 <filename>configure</filename>. If
3647 <filename>hc-build</filename> successfully completes the
3648 build process, you can install the resulting system, as
3658 <sect2 id="unregisterised-porting">
3659 <title>Porting GHC to a new architecture</title>
3661 <para>The first step in porting to a new architecture is to get
3662 an <firstterm>unregisterised</firstterm> build working. An
3663 unregisterised build is one that compiles via vanilla C only.
3664 By contrast, a registerised build uses the following
3665 architecture-specific hacks for speed:</para>
3669 <para>Global register variables: certain abstract machine
3670 <quote>registers</quote> are mapped to real machine
3671 registers, depending on how many machine registers are
3673 <filename>ghc/includes/MachRegs.h</filename>).</para>
3677 <para>Assembly-mangling: when compiling via C, we feed the
3678 assembly generated by gcc though a Perl script known as the
3679 <firstterm>mangler</firstterm> (see
3680 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3681 mangler rearranges the assembly to support tail-calls and
3682 various other optimisations.</para>
3686 <para>In an unregisterised build, neither of these hacks are
3687 used — the idea is that the C code generated by the
3688 compiler should compile using gcc only. The lack of these
3689 optimisations costs about a factor of two in performance, but
3690 since unregisterised compilation is usually just a step on the
3691 way to a full registerised port, we don't mind too much.</para>
3694 <title>Building an unregisterised port</title>
3696 <para>The first step is to get some unregisterised HC files.
3697 Either (a) download them from the GHC site (if there are
3698 some available for the right version of GHC), or
3699 (b) build them yourself on any machine with a working
3700 GHC. If at all possible this should be a machine with the
3701 same word size as the target.</para>
3703 <para>There is a script available which should automate the
3704 process of doing the 2-stage bootstrap necessary to get the
3705 unregisterised HC files - it's available in <ulink
3706 url="http://cvs.haskell.org/cgi-bin/cvsweb.cgi/fptools/distrib/cross-port"><filename>fptools/distrib/cross-port</filename></ulink>
3709 <para>Now take these unregisterised HC files to the target
3710 platform and bootstrap a compiler from them as per the
3711 instructions in <xref linkend="sec-booting-from-hc">. In
3712 <filename>build.mk</filename>, you need to tell the build
3713 system that the compiler you're building is
3714 (a) unregisterised itself, and (b) builds
3715 unregisterised binaries. This varies depending on the GHC
3716 version you're bootstraping:</para>
3719 # build.mk for GHC 4.08.x
3720 GhcWithRegisterised=NO
3724 # build.mk for GHC 5.xx and 6.x
3725 GhcUnregisterised=YES
3728 <para>Versions 5.xx and 6.x only: use the option
3729 <option>--enable-hc-boot-unregisterised</option> instead of
3730 <option>--enable-hc-boot</option> when running
3731 <filename>./configure</filename>.</para>
3733 <para>The build may not go through cleanly. We've tried to
3734 stick to writing portable code in most parts of the compiler,
3735 so it should compile on any POSIXish system with gcc, but in
3736 our experience most systems differ from the standards in one
3737 way or another. Deal with any problems as they arise - if you
3738 get stuck, ask the experts on
3739 <email>glasgow-haskell-users@haskell.org</email>.</para>
3741 <para>Once you have the unregisterised compiler up and
3742 running, you can use it to start a registerised port. The
3743 following sections describe the various parts of the system
3744 that will need architecture-specific tweaks in order to get a
3745 registerised build going.</para>
3747 <para>Lots of useful information about the innards of GHC is
3748 available in the <ulink
3749 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3750 Commentary</ulink>, which might be helpful if you run into
3751 some code which needs tweaking for your system.</para>
3755 <title>Porting the RTS</title>
3757 <para>The following files need architecture-specific code for a
3758 registerised build:</para>
3762 <term><filename>ghc/includes/MachRegs.h</filename></term>
3763 <indexterm><primary><filename>MachRegs.h</filename></primary>
3766 <para>Defines the STG-register to machine-register
3767 mapping. You need to know your platform's C calling
3768 convention, and which registers are generally available
3769 for mapping to global register variables. There are
3770 plenty of useful comments in this file.</para>
3774 <term><filename>ghc/includes/TailCalls.h</filename></term>
3775 <indexterm><primary><filename>TailCalls.h</filename></primary>
3778 <para>Macros that cooperate with the mangler (see <xref
3779 linkend="sec-mangler">) to make proper tail-calls
3784 <term><filename>ghc/rts/Adjustor.c</filename></term>
3785 <indexterm><primary><filename>Adjustor.c</filename></primary>
3789 <literal>foreign import "wrapper"</literal>
3791 <literal>foreign export dynamic</literal>).
3792 Not essential for getting GHC bootstrapped, so this file
3793 can be deferred until later if necessary.</para>
3797 <term><filename>ghc/rts/StgCRun.c</filename></term>
3798 <indexterm><primary><filename>StgCRun.c</filename></primary>
3801 <para>The little assembly layer between the C world and
3802 the Haskell world. See the comments and code for the
3803 other architectures in this file for pointers.</para>
3807 <term><filename>ghc/rts/MBlock.h</filename></term>
3808 <term><filename>ghc/rts/MBlock.c</filename></term>
3809 <indexterm><primary><filename>MBlock.h</filename></primary>
3811 <indexterm><primary><filename>MBlock.c</filename></primary>
3814 <para>These files are really OS-specific rather than
3815 architecture-specific. In <filename>MBlock.h</filename>
3816 is specified the absolute location at which the RTS
3817 should try to allocate memory on your platform (try to
3818 find an area which doesn't conflict with code or dynamic
3819 libraries). In <filename>Mblock.c</filename> you might
3820 need to tweak the call to <literal>mmap()</literal> for
3827 <sect3 id="sec-mangler">
3828 <title>The mangler</title>
3830 <para>The mangler is an evil Perl-script that rearranges the
3831 assembly code output from gcc to do two main things:</para>
3835 <para>Remove function prologues and epilogues, and all
3836 movement of the C stack pointer. This is to support
3837 tail-calls: every code block in Haskell code ends in an
3838 explicit jump, so we don't want the C-stack overflowing
3839 while we're jumping around between code blocks.</para>
3842 <para>Move the <firstterm>info table</firstterm> for a
3843 closure next to the entry code for that closure. In
3844 unregisterised code, info tables contain a pointer to the
3845 entry code, but in registerised compilation we arrange
3846 that the info table is shoved right up against the entry
3847 code, and addressed backwards from the entry code pointer
3848 (this saves a word in the info table and an extra
3849 indirection when jumping to the closure entry
3854 <para>The mangler is abstracted to a certain extent over some
3855 architecture-specific things such as the particular assembler
3856 directives used to herald symbols. Take a look at the
3857 definitions for other architectures and use these as a
3858 starting point.</para>
3862 <title>The native code generator</title>
3864 <para>The native code generator isn't essential to getting a
3865 registerised build going, but it's a desirable thing to have
3866 because it can cut compilation times in half. The native code
3867 generator is described in some detail in the <ulink
3868 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3869 commentary</ulink>.</para>
3875 <para>To support GHCi, you need to port the dynamic linker
3876 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
3877 currently supports the ELF and PEi386 object file formats - if
3878 your platform uses one of these then you probably don't have
3879 to do anything except fiddle with the
3880 <literal>#ifdef</literal>s at the top of
3881 <filename>Linker.c</filename> to tell it about your OS.</para>
3883 <para>If your system uses a different object file format, then
3884 you have to write a linker — good luck!</para>
3890 <sect1 id="sec-build-pitfalls">
3891 <title>Known pitfalls in building Glasgow Haskell
3893 <indexterm><primary>problems, building</primary></indexterm>
3894 <indexterm><primary>pitfalls, in building</primary></indexterm>
3895 <indexterm><primary>building pitfalls</primary></indexterm></title>
3898 WARNINGS about pitfalls and known “problems”:
3907 One difficulty that comes up from time to time is running out of space
3908 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
3909 compensate for the vagaries of different sysadmin approaches to temp
3911 <indexterm><primary>tmp, running out of space in</primary></indexterm>
3913 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
3914 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
3917 The best way around it is to say
3920 export TMPDIR=<dir>
3923 in your <filename>build.mk</filename> file.
3924 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
3933 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
3934 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
3942 When compiling via C, you'll sometimes get “warning: assignment from
3943 incompatible pointer type” out of GCC. Harmless.
3950 Similarly, <command>ar</command>chiving warning messages like the following are not
3954 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
3955 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
3965 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
3966 get an “Out of heap space” error message. These can vary with the
3967 vagaries of different systems, it seems. The solution is simple:
3974 If you're compiling with GHC 4.00 or later, then the
3975 <emphasis>maximum</emphasis> heap size must have been reached. This
3976 is somewhat unlikely, since the maximum is set to 64M by default.
3977 Anyway, you can raise it with the
3978 <option>-optCrts-M<size></option> flag (add this flag to
3979 <constant><module>_HC_OPTS</constant>
3980 <command>make</command> variable in the appropriate
3981 <filename>Makefile</filename>).
3988 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
3997 and try again: <command>gmake</command>. (see <Xref LinkEnd="sec-suffix"> for information about
3998 <constant><module>_HC_OPTS</constant>.)
4000 Alternatively, just cut to the chase:
4004 % make EXTRA_HC_OPTS=-optCrts-M128M
4013 If you try to compile some Haskell, and you get errors from GCC about
4014 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4015 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4017 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4018 this bug also suggests that you have an old GCC.
4026 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4030 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4031 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4033 ? # or, on some machines: ar s $i
4038 We'd be interested to know if this is still necessary.
4046 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4047 a bit from one Unix to another. One particular gotcha is macro calls
4052 SLIT("Hello, world")
4056 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4057 arguments, so you get
4061 :731: macro `SLIT' used with too many (2) args
4065 Alas, <command>cpp</command> doesn't tell you the offending file!
4067 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4078 <Sect1 id="winbuild"><Title>Notes for building under Windows</Title>
4081 This section summarises how to get the utilities you need on your
4082 Win95/98/NT/2000 machine to use CVS and build GHC. Similar notes for
4083 installing and running GHC may be found in the user guide. In general,
4084 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4085 You should read the GHC installation guide sections on Windows (in the user
4086 guide) before continuing to read these notes.
4090 <sect2 id="cygwin-and-mingw"><Title>Cygwin and MinGW</Title>
4092 <para> The Windows situation for building GHC is rather confusing. This section
4093 tries to clarify, and to establish terminology.</para>
4095 <sect3 id="ghc-mingw"><title>GHC-mingw</title>
4097 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4098 is a collection of header
4099 files and import libraries that allow one to use <command>gcc</command> and produce
4100 native Win32 programs that do not rely on any third-party DLLs. The
4101 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4102 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4105 <para>The GHC that we distribute includes, inside the distribution itself, the MinGW <command>gcc</command>,
4106 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4107 GHC compiles Haskell to C (or to
4108 assembly code), and then invokes these MinGW tools to generate an executable binary.
4109 The resulting binaries can run on any Win32 system.
4111 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4113 <para> The down-side of GHC-mingw is that the MinGW libraries do not support anything like the full
4114 Posix interface. So programs compiled with GHC-mingw cannot import the (Haskell) Posix
4115 library; they have to do
4116 their input output using standard Haskell I/O libraries, or native Win32 bindings.
4120 <sect3 id="ghc-cygwin"><title>GHC-cygwin</title>
4122 <para>There <emphasis>is</emphasis> a way to get the full Posix interface, which is to use Cygwin.
4123 <ulink url="http://www.cygwin.com">Cygwin</ulink> is a complete Unix simulation that runs on Win32.
4124 Cygwin comes with a shell, and all the usual Unix commands: <command>mv</command>, <command>rm</command>,
4125 <command>ls</command>, plus of course <command>gcc</command>, <command>ld</command> and so on.
4126 A C program compiled with the Cygwin <command>gcc</command> certainly can use all of Posix.
4128 <para>So why doesn't GHC use the Cygwin <command>gcc</command> and libraries? Because
4129 Cygwin comes with a DLL <emphasis>that must be linked with every runnable Cygwin-compiled program</emphasis>.
4130 A program compiled by the Cygwin tools cannot run at all unless Cygwin is installed.
4131 If GHC targeted Cygwin, users would have to install Cygwin just to run the Haskell programs
4132 that GHC compiled; and the Cygwin DLL would have to be in the DLL load path.
4133 Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4134 does not change, but the implementation certainly does. Even the interfaces to functions
4135 it exports seem to change occasionally. So programs compiled by GHC might only run with
4136 particular versions of Cygwin. All of this seems very undesirable.
4139 Nevertheless, it is certainly possible to build a version of GHC that targets Cygwin;
4140 we will call that <emphasis>GHC-cygwin</emphasis>. The up-side of GHC-cygwin is
4141 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4145 <sect3><title>HOST_OS vs TARGET_OS</title>
4148 In the source code you'll find various ifdefs looking like:
4150 #ifdef mingw32_HOST_OS
4156 #ifdef mingw32_TARGET_OS
4160 These macros are set by the configure script (via the file config.h).
4161 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4164 The "host" system is the one on which GHC itself will be run.
4167 The "target" system is the one for which the program compiled by GHC will be run.
4170 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4171 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4176 <sect3><title>Summary</title>
4178 <para>Notice that "GHC-mingw" means "GHC that <emphasis>targets</emphasis> MinGW". It says nothing about
4179 how that GHC was <emphasis>built</emphasis>. It is entirely possible to have a GHC-mingw that was built
4180 by compiling GHC's Haskell sources with a GHC-cygwin, or vice versa.</para>
4182 <para>We distribute only a GHC-mingw built by a GHC-mingw; supporting
4183 GHC-cygwin too is beyond our resources. The GHC we distribute
4184 therefore does not require Cygwin to run, nor do the programs it
4185 compiles require Cygwin.</para>
4187 <para>The instructions that follow describe how to build GHC-mingw. It is
4188 possible to build GHC-cygwin, but it's not a supported route, and the build system might
4191 <para>In your build tree, you build a compiler called <Command>ghc-inplace</Command>. It
4192 uses the <Command>gcc</Command> that you specify using the
4193 <option>--with-gcc</option> flag when you run
4194 <Command>configure</Command> (see below).
4195 The makefiles are careful to use <Command>ghc-inplace</Command> (not <Command>gcc</Command>)
4196 to compile any C files, so that it will in turn invoke the right <Command>gcc</Command> rather that
4197 whatever one happens to be in your path. However, the makefiles do use whatever <Command>ld</Command>
4198 and <Command>ar</Command> happen to be in your path. This is a bit naughty, but (a) they are only
4199 used to glom together .o files into a bigger .o file, or a .a file,
4200 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4201 Cygwin and Mingw use the same .o file format. So its ok.
4206 <Sect2><Title>Installing and configuring Cygwin</Title>
4208 <para>You don't need Cygwin to <emphasis>use</emphasis> GHC,
4209 but you do need it to <emphasis>build</emphasis> GHC.</para>
4211 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4212 The installation process is straightforward; we install it in <Filename>c:/cygwin</Filename>.
4213 During the installation dialogue, make sure that you select:
4214 <command>cvs</command>, <command>openssh</command>,
4215 <command>autoconf</command>,
4216 <command>binutils</command> (includes ld and (I think) ar),
4217 <command>gcc</command>,
4218 <command>flex</command>,
4219 <command>make</command>.
4222 <para> Now set the following user environment variables:
4225 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4226 <constant>PATH</constant></para></listitem>
4230 Set <constant>MAKE_MODE</constant> to <Literal>UNIX</Literal>. If you
4231 don't do this you get very weird messages when you type
4232 <Command>make</Command>, such as:
4234 /c: /c: No such file or directory
4239 <listitem><para> Set <constant>SHELL</constant> to
4240 <Filename>c:/cygwin/bin/sh</Filename>. When you invoke a shell in Emacs, this
4241 <constant>SHELL</constant> is what you get.
4244 <listitem><para> Set <constant>HOME</constant> to point to your
4245 home directory. This is where, for example,
4246 <command>bash</command> will look for your <filename>.bashrc</filename>
4247 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4253 There are a few other things to do:
4257 By default, cygwin provides the command shell <filename>ash</filename>
4258 as <filename>sh.exe</filename>. We have often seen build-system problems that
4259 turn out to be due to bugs in <filename>ash</filename>
4261 and length of command lines). On the other hand <filename>bash</filename> seems
4263 So, in <filename>cygwin/bin</filename>
4264 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4265 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4266 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4267 you can't rename a running program!
4273 Some script files used in the make system start with "<Command>#!/bin/perl</Command>",
4274 (and similarly for <Command>sh</Command>). Notice the hardwired path!
4275 So you need to ensure that your <Filename>/bin</Filename> directory has the following
4278 <listitem> <para><Command>sh</Command></para></listitem>
4279 <listitem> <para><Command>perl</Command></para></listitem>
4280 <listitem> <para><Command>cat</Command></para></listitem>
4282 All these come in Cygwin's <Filename>bin</Filename> directory, which you probably have
4283 installed as <Filename>c:/cygwin/bin</Filename>. By default Cygwin mounts "<Filename>/</Filename>" as
4284 <Filename>c:/cygwin</Filename>, so if you just take the defaults it'll all work ok.
4285 (You can discover where your Cygwin
4286 root directory <Filename>/</Filename> is by typing <Command>mount</Command>.)
4287 Provided <Filename>/bin</Filename> points to the Cygwin <Filename>bin</Filename>
4288 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4289 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4295 <para>Finally, here are some things to be aware of when using Cygwin:
4297 <listitem> <para>Cygwin doesn't deal well with filenames that include
4298 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4302 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4303 magical text in it. So other programs that don't use Cygwin's
4304 I/O libraries won't recognise such files as symlinks.
4305 In particular, programs compiled by GHC are meant to be runnable
4306 without having Cygwin, so they don't use the Cygwin library, so
4307 they don't recognise symlinks.
4311 Win32 has a <command>find</command> command which is not the same as Cygwin's find.
4312 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4313 before the Cygwin one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4314 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4315 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4322 <Sect2><Title>Other things you need to install</Title>
4324 <para>You have to install the following other things to build GHC:
4328 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4329 This is what you will use to compile GHC. Add it in your
4330 <constant>PATH</constant>: the installer tells you the path element
4331 you need to add upon completion.
4337 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4338 Happy is a parser generator used to compile the Haskell grammar. Add it in your
4339 <constant>PATH</constant>.
4345 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4346 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw">).
4347 Just pick up a mingw bundle at
4348 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4349 We install it in <filename>c:/mingw</filename>.
4351 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4352 They are only going to get used by explicit access (via the --with-gcc flag you
4353 give to <Command>configure</Command> later). If you do add them to your path
4354 you are likely to get into a mess because their names overlap with Cygwin binaries.
4360 <para>We use <command>emacs</command> a lot, so we install that too.
4361 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4362 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4363 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4364 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4365 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4366 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4372 <para> Finally, check out a copy of GHC sources from
4373 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
4380 <Sect2><Title>Building GHC</Title>
4383 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source">);
4384 the bullets below only tell
4385 you about Windows-specific wrinkles.</para>
4389 Run <Command>autoconf</Command> both in <filename>fptools</filename>
4390 and in <filename>fptools/ghc</filename>. If you omit the latter step you'll
4391 get an error when you run <filename>./configure</filename>:
4394 creating mk/config.h
4395 mk/config.h is unchanged
4397 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4398 ./configure: ./configure: No such file or directory
4399 configure: error: ./configure failed for ghc
4404 <listitem> <para><command>autoconf</command> seems to create the file <filename>configure</filename>
4405 read-only. So if you need to run autoconf again (which I sometimes do for safety's sake),
4408 /usr/bin/autoconf: cannot create configure: permission denied
4410 Solution: delete <filename>configure</filename> first.
4415 You either need to add <filename>ghc</filename> to your
4416 <constant>PATH</constant> before you invoke
4417 <Command>configure</Command>, or use the <Command>configure</Command>
4418 option <option>--with-ghc=c:/ghc/ghc-some-version/bin/ghc</option>.
4423 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4424 This file occasionally remembers out-of-date configuration information, which
4425 can be really confusing.
4431 After <command>autoconf</command> run <command>./configure</command> in
4432 <filename>fptools/</filename> thus:
4435 ./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc
4437 This is the point at which you specify that you are building GHC-mingw
4438 (see <xref linkend="ghc-mingw">). </para>
4440 <para> Both these options are important! It's possible to get into
4441 trouble using the wrong C compiler!</para>
4443 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4444 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4445 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4446 understand a Cygwin path. For example, if you
4447 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4448 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4449 time it tries to invoke it. Worse, the failure comes with
4450 no error message whatsoever. GHC simply fails silently when first invoked,
4451 typically leaving you with this:
4453 make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4454 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4455 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4456 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4457 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4458 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4459 -optc-fomit-frame-pointer -O2 -static
4460 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4461 make[2]: *** [Adjustor.o] Error 1
4462 make[1]: *** [all] Error 1
4463 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4464 make: *** [all] Error 1
4470 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin">)
4471 you'll have to do something more like:
4473 ./configure --with-gcc=...the Cygwin gcc...
4478 <listitem><para> You almost certainly want to set
4482 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config">).
4483 This tells the build system not to split each library into a myriad of little object files, one
4484 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4485 it dramatically increases the time taken to build the libraries in the first place.
4489 <listitem><para> Do not attempt to build the documentation.
4490 It needs all kinds of wierd Jade stuff that we haven't worked out for
4491 Win32.</para></listitem>