1 <!DOCTYPE Article PUBLIC "-//OASIS//DTD DocBook V3.1//EN">
3 <Article id="building-guide">
7 <Title>Building the Glasgow Functional Programming Tools Suite</Title>
8 <Author><OtherName>The GHC Team</OtherName></Author>
9 <Address><Email>glasgow-haskell-{users,bugs}@haskell.org</Email></Address>
10 <PubDate>November 2001</PubDate>
13 <para>The Glasgow fptools suite is a collection of Functional
14 Programming related tools, including the Glasgow Haskell
15 Compiler (GHC). The source code for the whole suite is kept in
16 a single CVS repository and shares a common build and
17 installation system.</para>
19 <para>This guide is intended for people who want to build or
20 modify programs from the Glasgow <Literal>fptools</Literal>
21 suite (as distinct from those who merely want to
22 <Emphasis>run</Emphasis> them). Installation instructions are
23 now provided in the user guide.</para>
25 <para>The bulk of this guide applies to building on Unix
26 systems; see <XRef LinkEnd="winbuild"> for Windows notes.</para>
32 <sect1 id="sec-getting">
33 <title>Getting the sources</title>
35 <para>You can get your hands on the <literal>fptools</literal>
41 <term><indexterm><primary>Source
42 distributions</primary></indexterm>Source distributions</term>
44 <para>You have a supported platform, but (a) you like
45 the warm fuzzy feeling of compiling things yourself;
46 (b) you want to build something ``extra”—e.g., a
47 set of libraries with strictness-analysis turned off; or
48 (c) you want to hack on GHC yourself.</para>
50 <para>A source distribution contains complete sources for
51 one or more projects in the <literal>fptools</literal>
52 suite. Not only that, but the more awkward
53 machine-independent steps are done for you. For example, if
55 <command>happy</command><indexterm><primary>happy</primary></indexterm>
56 you'll find it convenient that the source distribution
57 contains the result of running <command>happy</command> on
58 the parser specifications. If you don't want to alter the
59 parser then this saves you having to find and install
60 <command>happy</command>. You will still need a working
61 version of GHC (preferably version 4.08+) on your machine in
62 order to compile (most of) the sources, however.</para>
67 <term>The CVS repository.</term>
68 <indexterm><primary>CVS repository</primary>
71 <para>We make releases infrequently. If you want more
72 up-to-the minute (but less tested) source code then you need
73 to get access to our CVS repository.</para>
75 <para>All the <literal>fptools</literal> source code is held
76 in a CVS repository. CVS is a pretty good source-code
77 control system, and best of all it works over the
80 <para>The repository holds source code only. It holds no
81 mechanically generated files at all. So if you check out a
82 source tree from CVS you will need to install every utility
83 so that you can build all the derived files from
86 <para>More information about our CVS repository can be found
87 in <xref linkend="sec-cvs">.</para>
92 <para>If you are going to do any building from sources (either
93 from a source distribution or the CVS repository) then you need to
94 read all of this manual in detail.</para>
98 <title>Using the CVS repository</title>
100 <para>We use <ulink url="http://www.cvshome.org/">CVS</ulink> (Concurrent Version System) to keep track of our
101 sources for various software projects. CVS lets several people
102 work on the same software at the same time, allowing changes to be
103 checked in incrementally. </para>
105 <para>This section is a set of guidelines for how to use our CVS
106 repository, and will probably evolve in time. The main thing to
107 remember is that most mistakes can be undone, but if there's
108 anything you're not sure about feel free to bug the local CVS
109 meister (namely Jeff Lewis
110 <email>jlewis@galconn.com</email>). </para>
112 <sect2 id="cvs-access">
113 <title>Getting access to the CVS Repository</title>
115 <para>You can access the repository in one of two ways:
116 read-only (<xref linkend="cvs-read-only">), or read-write (<xref
117 linkend="cvs-read-write">).</para>
119 <sect3 id="cvs-read-only">
120 <title>Remote Read-only CVS Access</title>
122 <para>Read-only access is available to anyone - there's no
123 need to ask us first. With read-only CVS access you can do
124 anything except commit changes to the repository. You can
125 make changes to your local tree, and still use CVS's merge
126 facility to keep your tree up to date, and you can generate
127 patches using 'cvs diff' in order to send to us for
130 <para>To get read-only access to the repository:</para>
134 <para>Make sure that <application>cvs</application> is
135 installed on your machine.</para>
138 <para>Set your <literal>$CVSROOT</literal> environment variable to
139 <literal>:pserver:anoncvs@glass.cse.ogi.edu:/cvs</literal></para>
142 <para>Run the command</para>
146 <para>The password is simply <literal>cvs</literal>. This
147 sets up a file in your home directory called
148 <literal>.cvspass</literal>, which squirrels away the
149 dummy password, so you only need to do this step once.</para>
153 <para>Now go to <xref linkend="cvs-first">.</para>
158 <sect3 id="cvs-read-write">
159 <title>Remote Read-Write CVS Access</title>
161 <para>We generally supply read-write access to folk doing
162 serious development on some part of the source tree, when
163 going through us would be a pain. If you're developing some
164 feature, or think you have the time and inclination to fix
165 bugs in our sources, feel free to ask for read-write
166 access. There is a certain amount of responsibility that goes
167 with commit privileges; we are more likely to grant you access
168 if you've demonstrated your competence by sending us patches
169 via mail in the past.</para>
171 <para>To get remote read-write CVS access, you need to do the
172 following steps.</para>
176 <para>Make sure that <literal>cvs</literal> and
177 <literal>ssh</literal> are both installed on your
182 <para>Generate a DSA private-key/public-key pair, thus:</para>
186 <para>(<literal>ssh-keygen</literal> comes with
187 <literal>ssh</literal>.) Running <literal>ssh-keygen
188 -d</literal> creates the private and public keys in
189 <literal>$HOME/.ssh/id_dsa</literal> and
190 <literal>$HOME/.ssh/id_dsa.pub</literal> respectively
191 (assuming you accept the standard defaults).</para>
193 <para><literal>ssh-keygen -d</literal> will only work if
194 you have Version 2 <literal>ssh</literal> installed; it
195 will fail harmlessly otherwise. If you only have Version
196 1 you can instead generate an RSA key pair using plain</para>
201 <para>Doing so creates the private and public RSA keys in
202 <literal>$HOME/.ssh/identity</literal> and
203 <literal>$HOME/.ssh/identity.pub</literal>
206 <para>[Deprecated.] Incidentally, you can force a Version
207 2 <literal>ssh</literal> to use the Version 1 protocol by
208 creating <literal>$HOME/config</literal> with the
209 following in it:</para>
217 <para>In both cases, <literal>ssh-keygen</literal> will
218 ask for a <firstterm>passphrase</firstterm>. The
219 passphrase is a password that protects your private key.
220 In response to the 'Enter passphrase' question, you can
224 <para>[Recommended.] Enter a passphrase, which you
225 will quote each time you use CVS.
226 <literal>ssh-agent</literal> makes this entirely
230 <para>[Deprecated.] Just hit return (i.e. use an empty
231 passphrase); then you won't need to quote the
232 passphrase when using CVS. The downside is that
233 anyone who can see into your <literal>.ssh</literal>
234 directory, and thereby get your private key, can mess
235 up the repository. So you must keep the
236 <literal>.ssh</literal> directory with draconian
237 no-access permissions.</para>
243 [Windows users.] The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
244 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
245 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
246 and run it as follows:
248 c:\tmp> set CYGWIN32=tty
249 c:\tmp> c:/user/local/bin/ssh-keygen1
252 <para>[Windows users.] To protect your
253 <literal>.ssh</literal> from access by anyone else,
254 right-click your <literal>.ssh</literal> directory, and
255 select <literal>Properties</literal>. If you are not on
256 the access control list, add yourself, and give yourself
257 full permissions (the second panel). Remove everyone else
258 from the access control list. Don't leave them there but
259 deny them access, because 'they' may be a list that
261 <para>[March 2003] In fact <command>ssh</command> 3.6.1 now seems to <emphasis>require</emphasis>
262 you to have Unix permissions 600 (read/write for owner only)
263 on the <literal>.ssh/identity</literal> file, else it
264 bombs out. For your local C drive, it seems that <literal>chmod 600 identity</literal> works,
265 but on Windows NT/XP, it doesn't work on a network drive (exact dteails obscure).
266 The solution seems to be to set the CYGWIN environment
267 variable to "<literal>ntsec neta</literal>". The CYGWIN environment variable is discussed
268 in <ulink url="http://cygwin.com/cygwin-ug-net/using-cygwinenv.html">the Cygwin User's Guide</ulink>,
269 and there are more details in <ulink url="http://cygwin.com/faq/faq_4.html#SEC44">the Cygwin FAQ</ulink>.
274 <para>Send a message to to the CVS repository
275 administrator (currently Jeff Lewis
276 <email>jeff@galconn.com</email>), containing:</para>
279 <para>Your desired user-name.</para>
282 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
283 <literal>.ssh/identity.pub</literal>).</para>
286 <para>He will set up your account.</para>
290 <para>Set the following environment variables:</para>
294 <constant>$HOME</constant>: points to your home directory. This is where CVS
295 will look for its <filename>.cvsrc</filename> file.
301 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
303 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
304 <literal>ssh</literal> assumes that your CVS client
305 understands how to execute shell script
306 ("#!"s,really), which is what
307 <literal>ssh</literal> is. This may not be the case on
308 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
309 <literal>ssh1</literal>.</para>
313 <para><literal>$CVSROOT</literal> to
314 <literal>:ext:</literal><replaceable>your-username</replaceable>
315 <literal>@cvs.haskell.org:/home/cvs/root</literal>
316 where <replaceable>your-username</replaceable> is your user name on
317 <literal>cvs.haskell.org</literal>.
319 <para>The <literal>CVSROOT</literal> environment variable will
320 be recorded in the checked-out tree, so you don't need to set
321 this every time. </para>
327 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
328 if you want to use an Emacs buffer for typing in those long commit messages.
334 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
335 set this to point to <filename>bash.exe</filename>.
346 Put the following in <filename>$HOME/.cvsrc</filename>:
357 These are the default options for the specified CVS commands,
358 and represent better defaults than the usual ones. (Feel
359 free to change them.)
363 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
364 the 8.3 DOS filesystem, but that restriction should have
365 been lifted by now (i.e., you're using VFAT or later filesystems.) If
366 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
374 <para>[Experts.] Once your account is set up, you can get
375 access from other machines without bothering Jeff, thus:</para>
378 <para>Generate a public/private key pair on the new
382 <para>Use ssh to log in to
383 <literal>cvs.haskell.org</literal>, from your old
387 <para>Add the public key for the new machine to the file
388 <literal>$HOME/ssh/authorized_keys</literal> on
389 <literal>cvs.haskell.org</literal>.
390 (<literal>authorized_keys2</literal>, I think, for Version
394 <para>Make sure that the new version of
395 <literal>authorized_keys</literal> still has 600 file
404 <sect2 id="cvs-first">
405 <title>Checking Out a Source Tree</title>
409 <para>Make sure you set your <literal>CVSROOT</literal>
410 environment variable according to either of the remote
411 methods above. The Approved Way to check out a source tree
412 is as follows:</para>
415 $ cvs checkout fpconfig
418 <para>At this point you have a new directory called
419 <literal>fptools</literal> which contains the basic stuff
420 for the fptools suite, including the configuration files and
421 some other junk. </para>
423 <para>[Windows users.] The following messages appear to be harmless:
425 setsockopt IPTOS_LOWDELAY: Invalid argument
426 setsockopt IPTOS_THROUGHPUT: Invalid argument
431 <para>You can call the fptools directory whatever you like,
432 CVS won't mind: </para>
435 $ mv fptools <replaceable>directory</replaceable>
438 <para> NB: after you've read the CVS manual you might be
439 tempted to try</para>
441 $ cvs checkout -d <replaceable>directory</replaceable> fpconfig
444 <para>instead of checking out <literal>fpconfig</literal>
445 and then renaming it. But this doesn't work, and will
446 result in checking out the entire repository instead of just
447 the <literal>fpconfig</literal> bit.</para>
449 $ cd <replaceable>directory</replaceable>
450 $ cvs checkout ghc hslibs libraries
453 <para>The second command here checks out the relevant
454 modules you want to work on. For a GHC build, for instance,
455 you need at least the <literal>ghc</literal>,
456 <literal>hslibs</literal> and <literal>libraries</literal>
457 modules (for a full list of the projects available, see
458 <xref linkend="projects">).</para>
460 <para>Remember that if you do not have
461 <literal>happy</literal> installed, you need to check it out
467 <sect2 id="cvs-committing">
468 <title>Committing Changes</title>
470 <para>This is only if you have read-write access to the
471 repository. For anoncvs users, CVS will issue a "read-only
472 repository" error if you try to commit changes.</para>
476 <para>Build the software, if necessary. Unless you're just
477 working on documentation, you'll probably want to build the
478 software in order to test any changes you make.</para>
482 <para>Make changes. Preferably small ones first.</para>
486 <para>Test them. You can see exactly what changes you've
487 made by using the <literal>cvs diff</literal> command:</para>
491 <para>lists all the changes (using the
492 <literal>diff</literal> command) in and below the current
493 directory. In emacs, <literal>C-c C-v =</literal> runs
494 <literal>cvs diff</literal> on the current buffer and shows
495 you the results.</para>
499 <para>If you changed something in the
500 <literal>fptools/libraries</literal> subdirectories, also run
501 <literal>make html</literal> to check if the documentation can
502 be generated successfully, too.</para>
506 <para>Before checking in a change, you need to update your
513 <para>This pulls in any changes that other people have made,
514 and merges them with yours. If there are any conflicts, CVS
515 will tell you, and you'll have to resolve them before you
516 can check your changes in. The documentation describes what
517 to do in the event of a conflict.</para>
519 <para>It's not always necessary to do a full cvs update
520 before checking in a change, since CVS will always tell you
521 if you try to check in a file that someone else has changed.
522 However, you should still update at regular intervals to
523 avoid making changes that don't work in conjuction with
524 changes that someone else made. Keeping an eye on what goes
525 by on the mailing list can help here.</para>
529 <para>When you're happy that your change isn't going to
530 break anything, check it in. For a one-file change:</para>
533 $ cvs commit <replaceable>filename</replaceable>
536 <para>CVS will then pop up an editor for you to enter a
537 "commit message", this is just a short description
538 of what your change does, and will be kept in the history of
541 <para>If you're using emacs, simply load up the file into a
542 buffer and type <literal>C-x C-q</literal>, and emacs will
543 prompt for a commit message and then check in the file for
546 <para>For a multiple-file change, things are a bit
547 trickier. There are several ways to do this, but this is the
548 way I find easiest. First type the commit message into a
549 temporary file. Then either</para>
552 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
555 <para>or, if nothing else has changed in this part of the
559 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
562 <para>where <replaceable>directory</replaceable> is a common
563 parent directory for all your changes, and
564 <replaceable>commit-message</replaceable> is the name of the
565 file containing the commit message.</para>
567 <para>Shortly afterwards, you'll get some mail from the
568 relevant mailing list saying which files changed, and giving
569 the commit message. For a multiple-file change, you should
570 still get only <emphasis>one</emphasis> message.</para>
575 <sect2 id="cvs-update">
576 <title>Updating Your Source Tree</title>
578 <para>It can be tempting to cvs update just part of a source
579 tree to bring in some changes that someone else has made, or
580 before committing your own changes. This is NOT RECOMMENDED!
581 Quite often changes in one part of the tree are dependent on
582 changes in another part of the tree (the
583 <literal>mk/*.mk</literal> files are a good example where
584 problems crop up quite often). Having an inconsistent tree is a
585 major cause of headaches. </para>
587 <para>So, to avoid a lot of hassle, follow this recipe for
588 updating your tree:</para>
592 $ cvs update -P 2>&1 | tee log</screen>
594 <para>Look at the log file, and fix any conflicts (denoted by a
595 <quote>C</quote> in the first column). New directories may have
596 appeared in the repository; CVS doesn't check these out by
597 default, so to get new directories you have to explicitly do
599 $ cvs update -d</screen>
600 in each project subdirectory. Don't do this at the top level,
601 because then <emphasis>all</emphasis> the projects will be
604 <para>If you're using multiple build trees, then for every build
605 tree you have pointing at this source tree, you need to update
606 the links in case any new files have appeared: </para>
609 $ cd <replaceable>build-tree</replaceable>
610 $ lndir <replaceable>source-tree</replaceable>
613 <para>Some files might have been removed, so you need to remove
614 the links pointing to these non-existent files:</para>
617 $ find . -xtype l -exec rm '{}' \;
620 <para>To be <emphasis>really</emphasis> safe, you should do
623 <screen>$ gmake all</screen>
625 <para>from the top-level, to update the dependencies and build
626 any changed files. </para>
629 <sect2 id="cvs-tags">
630 <title>GHC Tag Policy</title>
632 <para>If you want to check out a particular version of GHC,
633 you'll need to know how we tag versions in the repository. The
634 policy (as of 4.04) is:</para>
638 <para>The tree is branched before every major release. The
639 branch tag is <literal>ghc-x-xx-branch</literal>, where
640 <literal>x-xx</literal> is the version number of the release
641 with the <literal>'.'</literal> replaced by a
642 <literal>'-'</literal>. For example, the 4.04 release lives
643 on <literal>ghc-4-04-branch</literal>.</para>
647 <para>The release itself is tagged with
648 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
649 called <literal>ghc-4-06</literal>.</para>
653 <para>We didn't always follow these guidelines, so to see
654 what tags there are for previous versions, do <literal>cvs
655 log</literal> on a file that's been around for a while (like
656 <literal>fptools/ghc/README</literal>).</para>
660 <para>So, to check out a fresh GHC 4.06 tree you would
664 $ cvs co -r ghc-4-06 fpconfig
666 $ cvs co -r ghc-4-06 ghc hslibs
670 <sect2 id="cvs-hints">
671 <title>General Hints</title>
675 <para>As a general rule: commit changes in small units,
676 preferably addressing one issue or implementing a single
677 feature. Provide a descriptive log message so that the
678 repository records exactly which changes were required to
679 implement a given feature/fix a bug. I've found this
680 <emphasis>very</emphasis> useful in the past for finding out
681 when a particular bug was introduced: you can just wind back
682 the CVS tree until the bug disappears.</para>
686 <para>Keep the sources at least *buildable* at any given
687 time. No doubt bugs will creep in, but it's quite easy to
688 ensure that any change made at least leaves the tree in a
689 buildable state. We do nightly builds of GHC to keep an eye
690 on what things work/don't work each day and how we're doing
691 in relation to previous verions. This idea is truely wrecked
692 if the compiler won't build in the first place!</para>
696 <para>To check out extra bits into an already-checked-out
697 tree, use the following procedure. Suppose you have a
698 checked-out fptools tree containing just ghc, and you want
699 to add nofib to it:</para>
710 $ cvs update -d nofib
713 <para>(the -d flag tells update to create a new
714 directory). If you just want part of the nofib suite, you
719 $ cvs checkout nofib/spectral
722 <para>This works because <literal>nofib</literal> is a
723 module in its own right, and spectral is a subdirectory of
724 the nofib module. The path argument to checkout must always
725 start with a module name. There's no equivalent form of this
726 command using <literal>update</literal>.</para>
732 <sect1 id="projects">
733 <title>What projects are there?</title>
735 <para>The <literal>fptools</literal> suite consists of several
736 <firstterm>projects</firstterm>, most of which can be downloaded,
737 built and installed individually. Each project corresponds to a
738 subdirectory in the source tree, and if checking out from CVS then
739 each project can be checked out individually by sitting in the top
740 level of your source tree and typing <command>cvs checkout
741 <replaceable>project</replaceable></command>.</para>
743 <para>Here is a list of the projects currently available:</para>
747 <term><literal>ghc</literal></term>
748 <indexterm><primary><literal>ghc</literal></primary>
749 <secondary>project</secondary></indexterm>
751 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
752 Haskell Compiler</ulink> (minus libraries). Absolutely
753 required for building GHC.</para>
758 <term><literal>glafp-utils</literal></term>
759 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
761 <para>Utility programs, some of which are used by the
762 build/installation system. Required for pretty much
768 <term><literal>green-card</literal></term>
769 <indexterm><primary><literal>green-card</literal></primary><secondary>project</secondary></indexterm>
772 url="http://www.haskell.org/greencard/">Green Card</ulink>
773 system for generating Haskell foreign function
779 <term><literal>haggis</literal></term>
780 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
783 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
784 Haskell GUI framework.</para>
789 <term><literal>haddock</literal></term>
790 <indexterm><primary><literal>haddock</literal></primary><secondary>project</secondary></indexterm>
793 url="http://www.haskell.org/haddock/">Haddock</ulink>
794 documentation tool.</para>
799 <term><literal>happy</literal></term>
800 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
803 url="http://www.haskell.org/happy/">Happy</ulink> Parser
809 <term><literal>hdirect</literal></term>
810 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
813 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
814 Haskell interoperability tool.</para>
819 <term><literal>hood</literal></term>
820 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
822 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
823 Object Observation Debugger</ulink>.</para>
828 <term><literal>hslibs</literal></term>
829 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
831 <para>Supplemental libraries for GHC
832 (<emphasis>required</emphasis> for building GHC).</para>
837 <term><literal>libraries</literal></term>
838 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
840 <para>Hierarchical Haskell library suite
841 (<emphasis>required</emphasis> for building GHC).</para>
846 <term><literal>mhms</literal></term>
847 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
849 <para>The Modular Haskell Metric System.</para>
854 <term><literal>nofib</literal></term>
855 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
857 <para>The NoFib suite: A collection of Haskell programs used
858 primarily for benchmarking.</para>
863 <term><literal>testsuite</literal></term>
864 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
866 <para>A testing framework, including GHC's regression test
872 <para>So, to build GHC you need at least the
873 <literal>ghc</literal>, <literal>libraries</literal> and
874 <literal>hslibs</literal> projects (a GHC source distribution will
875 already include the bits you need).</para>
878 <sect1 id="sec-build-checks">
879 <title>Things to check before you start</title>
881 <para>Here's a list of things to check before you get
887 <indexterm><primary>Disk space needed</primary></indexterm>
888 <para>Disk space needed: from about 100Mb for a basic GHC
889 build, up to probably 500Mb for a GHC build with everything
890 included (libraries built several different ways,
895 <para>Use an appropriate machine / operating system. <xref
896 linkend="sec-port-info"> lists the supported platforms; if
897 yours isn't amongst these then you can try porting GHC (see
898 <xref linkend="sec-porting-ghc">).</para>
902 <para>Be sure that the “pre-supposed” utilities are
903 installed. <Xref LinkEnd="sec-pre-supposed">
908 <para>If you have any problem when building or installing the
909 Glasgow tools, please check the “known pitfalls” (<Xref
910 LinkEnd="sec-build-pitfalls">). Also check the FAQ for the
911 version you're building, which is part of the User's Guide and
912 available on the <ulink URL="http://www.haskell.org/ghc/" >GHC web
915 <indexterm><primary>bugs</primary><secondary>known</secondary></indexterm>
917 <para>If you feel there is still some shortcoming in our
918 procedure or instructions, please report it.</para>
920 <para>For GHC, please see the <ulink
921 url="http://www.haskell.org/ghc/docs/latest/set/bug-reporting.html">bug-reporting
922 section of the GHC Users' Guide</ulink>, to maximise the
923 usefulness of your report.</para>
925 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
926 <para>If in doubt, please send a message to
927 <email>glasgow-haskell-bugs@haskell.org</email>.
928 <indexterm><primary>bugs</primary><secondary>mailing
929 list</secondary></indexterm></para>
934 <sect1 id="sec-port-info">
935 <title>What machines the Glasgow tools run on</title>
937 <indexterm><primary>ports</primary><secondary>GHC</secondary></indexterm>
938 <indexterm><primary>GHC</primary><secondary>ports</secondary></indexterm>
939 <indexterm><primary>platforms</primary><secondary>supported</secondary></indexterm>
941 <para>The main question is whether or not the Haskell compiler
942 (GHC) runs on your platform.</para>
944 <para>A “platform” is a
945 architecture/manufacturer/operating-system combination, such as
946 <literal>sparc-sun-solaris2</literal>. Other common ones are
947 <literal>alpha-dec-osf2</literal>,
948 <literal>hppa1.1-hp-hpux9</literal>,
949 <literal>i386-unknown-linux</literal>,
950 <literal>i386-unknown-solaris2</literal>,
951 <literal>i386-unknown-freebsd</literal>,
952 <literal>i386-unknown-cygwin32</literal>,
953 <literal>m68k-sun-sunos4</literal>,
954 <literal>mips-sgi-irix5</literal>,
955 <literal>sparc-sun-sunos4</literal>,
956 <literal>sparc-sun-solaris2</literal>,
957 <literal>powerpc-ibm-aix</literal>.</para>
959 <para>Some libraries may only work on a limited number of
960 platforms; for example, a sockets library is of no use unless the
961 operating system supports the underlying BSDisms.</para>
964 <title>What platforms the Haskell compiler (GHC) runs on</title>
966 <indexterm><primary>fully-supported platforms</primary></indexterm>
967 <indexterm><primary>native-code generator</primary></indexterm>
968 <indexterm><primary>registerised ports</primary></indexterm>
969 <indexterm><primary>unregisterised ports</primary></indexterm>
971 <para>The GHC hierarchy of Porting Goodness: (a) Best is a
972 native-code generator; (b) next best is a
973 “registerised” port; (c) the bare minimum is an
974 “unregisterised” port.
975 (“Unregisterised” is so terrible that we won't say
976 more about it).</para>
978 <para>We use Sparcs running Solaris 2.7 and x86 boxes running
979 FreeBSD and Linux, so those are the best supported platforms,
980 unsurprisingly.</para>
982 <para>Here's everything that's known about GHC ports. We
983 identify platforms by their “canonical”
984 CPU/Manufacturer/OS triple.</para>
988 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:</term>
989 <indexterm><primary>alpha-dec-osf</primary></indexterm>
990 <indexterm><primary>alpha-dec-linux</primary></indexterm>
991 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
992 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
993 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
996 <para>The OSF port is currently working (as of GHC version
997 5.02.1) and well supported. The native code generator is
998 currently non-working. Other operating systems will
999 require some minor porting.</para>
1004 <term>sparc-sun-sunos4</term>
1005 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
1007 <para>Probably works with minor tweaks, hasn't been tested
1013 <term>sparc-sun-solaris2</term>
1014 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
1016 <para>Fully supported (at least for Solaris 2.7),
1017 including native-code generator.</para>
1022 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
1023 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1025 <para>A registerised port is available for version 4.08,
1026 but GHC hasn't been built on that platform since (as far
1027 as we know). No native-code generator.</para>
1032 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1033 <indexterm><primary>i386-*-linux</primary></indexterm>
1035 <para>GHC works registerised and has a native code
1036 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1037 or later. NOTE about <literal>glibc</literal> versions:
1038 GHC binaries built on a system running <literal>glibc
1039 2.0</literal> won't work on a system running
1040 <literal>glibc 2.1</literal>, and vice versa. In general,
1041 don't expect compatibility between
1042 <literal>glibc</literal> versions, even if the shared
1043 library version hasn't changed.</para>
1048 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or
1050 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1052 <para>GHC works registerised. Pre-built packages are
1053 available in the native package format, so if you just
1054 need binaries you're better off just installing the
1055 package (it might even be on your installation
1061 <term>i386-unknown-openbsd (PCs running OpenBSD)</term>
1062 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1064 <para>Supported, with native code generator. Packages are
1065 available through the ports system in the native package
1071 <term>i386-unknown-netbsd (PCs running NetBSD and
1073 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1075 <para>Will require some minor porting effort, but should
1076 work registerised.</para>
1081 <term>i386-unknown-mingw32 (PCs running Windows)</term>
1082 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1084 <para>Fully supported under Win9x, WinNT, Win2k, and
1085 WinXP. Includes a native code generator. Building from
1086 source requires a recent <ulink
1087 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1088 to be installed.</para>
1093 <term>ia64-unknown-linux</term>
1094 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1096 <para>GHC currently works unregisterised. A registerised
1097 port is in progress.</para>
1102 <term>mips-sgi-irix5</term>
1103 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1105 <para>Port has worked in the past, but hasn't been tested
1106 for some time (and will certainly have rotted in various
1107 ways). As usual, we don't have access to machines and
1108 there hasn't been an overwhelming demand for this port,
1109 but feel free to get in touch.</para>
1114 <term>powerpc-ibm-aix</term>
1115 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1117 <para>Port currently doesn't work, needs some minimal
1118 porting effort. As usual, we don't have access to
1119 machines and there hasn't been an overwhelming demand for
1120 this port, but feel free to get in touch.</para>
1125 <term>powerpc-apple-darwin</term>
1126 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1128 <para>Supported registerised. No native code
1134 <term>powerpc-apple-linux</term>
1135 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1137 <para>Not supported (yet).</para>
1142 <para>Various other systems have had GHC ported to them in the
1143 distant past, including various Motorola 68k boxes. The 68k
1144 support still remains, but porting to one of these systems will
1145 certainly be a non-trivial task.</para>
1149 <title>What machines the other tools run on</title>
1151 <para>Unless you hear otherwise, the other tools work if GHC
1157 <sect1 id="sec-pre-supposed">
1158 <title>Installing pre-supposed utilities</title>
1160 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1161 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1163 <para>Here are the gory details about some utility programs you
1164 may need; <command>perl</command>, <command>gcc</command> and
1165 <command>happy</command> are the only important
1166 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1167 important if you're going for Parallel Haskell.) The
1168 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1169 script will tell you if you are missing something.</para>
1175 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1176 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1178 <para>GHC is required to build many of the tools, including
1179 GHC itself. If you need to port GHC to your platform
1180 because there isn't a binary distribution of GHC available,
1181 then see <xref linkend="sec-porting-ghc">.</para>
1183 <para>Which version of GHC you need will depend on the
1184 packages you intend to build. GHC itself will normally
1185 build using one of several older versions of itself - check
1186 the announcement or release notes for details.</para>
1192 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1193 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1195 <para><emphasis>You have to have Perl to proceed!</emphasis>
1196 Perl version 5 at least is required. GHC has been known to
1197 tickle bugs in Perl, so if you find that Perl crashes when
1198 running GHC try updating (or downgrading) your Perl
1199 installation. Versions of Perl that we use and are known to
1200 be fairly stable are 5.005 and 5.6.1.</para>
1202 <para>For Win32 platforms, you should use the binary
1203 supplied in the InstallShield (copy it to
1204 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1207 <para>Perl should be put somewhere so that it can be invoked
1208 by the <literal>#!</literal> script-invoking
1209 mechanism. The full pathname may need to be less than 32
1210 characters long on some systems.</para>
1215 <term>GNU C (<command>gcc</command>)</term>
1216 <indexterm><primary>pre-supposed: GCC (GNU C
1217 compiler)</primary></indexterm> <indexterm><primary>GCC (GNU C
1218 compiler), pre-supposed</primary></indexterm>
1220 <para>We recommend using GCC version 2.95.2 on all
1221 platforms. Failing that, version 2.7.2 is stable on most
1222 platforms. Earlier versions of GCC can be assumed not to
1223 work, and versions in between 2.7.2 and 2.95.2 (including
1224 <command>egcs</command>) have varying degrees of stability
1225 depending on the platform.</para>
1227 <para>GCC 3.2 is currently known to have problems building
1228 GHC on Sparc, but is stable on x86.</para>
1230 <para>GCC 3.3 currently cannot be used to build GHC, due to
1231 some problems with the new C preprocessor.</para>
1233 <para>If your GCC dies with “internal error” on
1234 some GHC source file, please let us know, so we can report
1235 it and get things improved. (Exception: on iX86
1236 boxes—you may need to fiddle with GHC's
1237 <option>-monly-N-regs</option> option; see the User's
1243 <term>GNU Make</term>
1244 <indexterm><primary>make</primary><secondary>GNU</secondary>
1247 <para>The fptools build system makes heavy use of features
1248 specific to GNU <command>make</command>, so you must have
1249 this installed in order to build any of the fptools
1256 <indexterm><primary>Happy</primary></indexterm>
1258 <para>Happy is a parser generator tool for Haskell, and is
1259 used to generate GHC's parsers. Happy is written in
1260 Haskell, and is a project in the CVS repository
1261 (<literal>fptools/happy</literal>). It can be built from
1262 source, but bear in mind that you'll need GHC installed in
1263 order to build it. To avoid the chicken/egg problem,
1264 install a binary distribution of either Happy or GHC to get
1265 started. Happy distributions are available from <ulink
1266 url="http://www.haskell.org/happy/">Happy's Web
1267 Page</ulink>.</para>
1272 <term>Autoconf</term>
1273 <indexterm><primary>pre-supposed: Autoconf</primary></indexterm>
1274 <indexterm><primary>Autoconf, pre-supposed</primary></indexterm>
1276 <para>GNU Autoconf is needed if you intend to build from the
1277 CVS sources, it is <emphasis>not</emphasis> needed if you
1278 just intend to build a standard source distribution.</para>
1280 <para>Version 2.52 or later of autoconf is required.
1281 NB. vesrion 2.13 will no longer work, as of GHC version
1284 <para>Autoconf builds the <command>configure</command>
1285 script from <filename>configure.in</filename> and
1286 <filename>aclocal.m4</filename>. If you modify either of
1287 these files, you'll need <command>autoconf</command> to
1288 rebuild <filename>configure</filename>.</para>
1293 <term><command>sed</command></term>
1294 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1295 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1297 <para>You need a working <command>sed</command> if you are
1298 going to build from sources. The build-configuration stuff
1299 needs it. GNU sed version 2.0.4 is no good! It has a bug
1300 in it that is tickled by the build-configuration. 2.0.5 is
1301 OK. Others are probably OK too (assuming we don't create too
1302 elaborate configure scripts.)</para>
1307 <para>One <literal>fptools</literal> project is worth a quick note
1308 at this point, because it is useful for all the others:
1309 <literal>glafp-utils</literal> contains several utilities which
1310 aren't particularly Glasgow-ish, but Occasionally Indispensable.
1311 Like <command>lndir</command> for creating symbolic link
1314 <sect2 id="pre-supposed-gph-tools">
1315 <title>Tools for building parallel GHC (GPH)</title>
1319 <term>PVM version 3:</term>
1320 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1321 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1323 <para>PVM is the Parallel Virtual Machine on which
1324 Parallel Haskell programs run. (You only need this if you
1325 plan to run Parallel Haskell. Concurrent Haskell, which
1326 runs concurrent threads on a uniprocessor doesn't need
1327 it.) Underneath PVM, you can have (for example) a network
1328 of workstations (slow) or a multiprocessor box
1331 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1332 It is readily available on the net; I think I got it from
1333 <literal>research.att.com</literal>, in
1334 <filename>netlib</filename>.</para>
1336 <para>A PVM installation is slightly quirky, but easy to
1337 do. Just follow the <filename>Readme</filename>
1338 instructions.</para>
1343 <term><command>bash</command>:</term>
1344 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1346 <para>Sadly, the <command>gr2ps</command> script, used to
1347 convert “parallelism profiles” to PostScript,
1348 is written in Bash (GNU's Bourne Again shell). This bug
1349 will be fixed (someday).</para>
1355 <sect2 id="pre-supposed-other-tools">
1356 <title>Other useful tools</title>
1361 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1362 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1364 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1365 to build a couple of utilities in
1366 <literal>glafp-utils</literal>. Depending on your
1367 operating system, the supplied <command>lex</command> may
1368 or may not work; you should get the GNU version.</para>
1373 <para>More tools are required if you want to format the documentation
1374 that comes with GHC and other fptools projects. See <xref
1375 linkend="building-docs">.</para>
1379 <sect1 id="sec-building-from-source">
1380 <title>Building from source</title>
1382 <indexterm><primary>Building from source</primary></indexterm>
1383 <indexterm><primary>Source, building from</primary></indexterm>
1385 <para>You've been rash enough to want to build some of the Glasgow
1386 Functional Programming tools (GHC, Happy, nofib, etc.) from
1387 source. You've slurped the source, from the CVS repository or
1388 from a source distribution, and now you're sitting looking at a
1389 huge mound of bits, wondering what to do next.</para>
1391 <para>Gingerly, you type <command>make</command>. Wrong
1394 <para>This rest of this guide is intended for duffers like me, who
1395 aren't really interested in Makefiles and systems configurations,
1396 but who need a mental model of the interlocking pieces so that
1397 they can make them work, extend them consistently when adding new
1398 software, and lay hands on them gently when they don't
1401 <sect2 id="quick-start">
1402 <title>Quick Start</title>
1404 <para>If you are starting from a source distribution, and just
1405 want a completely standard build, then the following should
1408 <screen>$ ./configure
1413 <para>For GHC, this will do a 2-stage bootstrap build of the
1414 compiler, with profiling libraries, and install the
1417 <para>If you want to do anything at all non-standard, or you
1418 want to do some development, read on...</para>
1421 <sect2 id="sec-source-tree">
1422 <title>Your source tree</title>
1424 <para>The source code is held in your <emphasis>source
1425 tree</emphasis>. The root directory of your source tree
1426 <emphasis>must</emphasis> contain the following directories and
1431 <para><filename>Makefile</filename>: the root
1436 <para><filename>mk/</filename>: the directory that contains
1437 the main Makefile code, shared by all the
1438 <literal>fptools</literal> software.</para>
1442 <para><filename>configure.in</filename>,
1443 <filename>config.sub</filename>,
1444 <filename>config.guess</filename>: these files support the
1445 configuration process.</para>
1449 <para><filename>install-sh</filename>.</para>
1453 <para>All the other directories are individual
1454 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1455 system—for example, the Glasgow Haskell Compiler
1456 (<literal>ghc</literal>), the Happy parser generator
1457 (<literal>happy</literal>), the <literal>nofib</literal>
1458 benchmark suite, and so on. You can have zero or more of these.
1459 Needless to say, some of them are needed to build others.</para>
1461 <para>The important thing to remember is that even if you want
1462 only one project (<literal>happy</literal>, say), you must have
1463 a source tree whose root directory contains
1464 <filename>Makefile</filename>, <filename>mk/</filename>,
1465 <filename>configure.in</filename>, and the project(s) you want
1466 (<filename>happy/</filename> in this case). You cannot get by
1467 with just the <filename>happy/</filename> directory.</para>
1471 <title>Build trees</title>
1472 <indexterm><primary>build trees</primary></indexterm>
1473 <indexterm><primary>link trees, for building</primary></indexterm>
1475 <para>If you just want to build the software once on a single
1476 platform, then your source tree can also be your build tree, and
1477 you can skip the rest of this section.</para>
1479 <para>We often want to build multiple versions of our software
1480 for different architectures, or with different options
1481 (e.g. profiling). It's very desirable to share a single copy of
1482 the source code among all these builds.</para>
1484 <para>So for every source tree we have zero or more
1485 <emphasis>build trees</emphasis>. Each build tree is initially
1486 an exact copy of the source tree, except that each file is a
1487 symbolic link to the source file, rather than being a copy of
1488 the source file. There are “standard” Unix
1489 utilities that make such copies, so standard that they go by
1491 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1492 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1493 are two (If you don't have either, the source distribution
1494 includes sources for the X11
1495 <command>lndir</command>—check out
1496 <filename>fptools/glafp-utils/lndir</filename>). See <Xref
1497 LinkEnd="sec-storysofar"> for a typical invocation.</para>
1499 <para>The build tree does not need to be anywhere near the
1500 source tree in the file system. Indeed, one advantage of
1501 separating the build tree from the source is that the build tree
1502 can be placed in a non-backed-up partition, saving your systems
1503 support people from backing up untold megabytes of
1504 easily-regenerated, and rapidly-changing, gubbins. The golden
1505 rule is that (with a single exception—<XRef
1506 LinkEnd="sec-build-config">) <emphasis>absolutely everything in
1507 the build tree is either a symbolic link to the source tree, or
1508 else is mechanically generated</emphasis>. It should be
1509 perfectly OK for your build tree to vanish overnight; an hour or
1510 two compiling and you're on the road again.</para>
1512 <para>You need to be a bit careful, though, that any new files
1513 you create (if you do any development work) are in the source
1514 tree, not a build tree!</para>
1516 <para>Remember, that the source files in the build tree are
1517 <emphasis>symbolic links</emphasis> to the files in the source
1518 tree. (The build tree soon accumulates lots of built files like
1519 <filename>Foo.o</filename>, as well.) You can
1520 <emphasis>delete</emphasis> a source file from the build tree
1521 without affecting the source tree (though it's an odd thing to
1522 do). On the other hand, if you <emphasis>edit</emphasis> a
1523 source file from the build tree, you'll edit the source-tree
1524 file directly. (You can set up Emacs so that if you edit a
1525 source file from the build tree, Emacs will silently create an
1526 edited copy of the source file in the build tree, leaving the
1527 source file unchanged; but the danger is that you think you've
1528 edited the source file whereas actually all you've done is edit
1529 the build-tree copy. More commonly you do want to edit the
1530 source file.)</para>
1532 <para>Like the source tree, the top level of your build tree
1533 must be (a linked copy of) the root directory of the
1534 <literal>fptools</literal> suite. Inside Makefiles, the root of
1535 your build tree is called
1536 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1537 In the rest of this document path names are relative to
1538 <constant>$(FPTOOLS_TOP)</constant> unless
1539 otherwise stated. For example, the file
1540 <filename>ghc/mk/target.mk</filename> is actually
1541 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</filename>.</para>
1544 <sect2 id="sec-build-config">
1545 <title>Getting the build you want</title>
1547 <para>When you build <literal>fptools</literal> you will be
1548 compiling code on a particular <emphasis>host
1549 platform</emphasis>, to run on a particular <emphasis>target
1550 platform</emphasis> (usually the same as the host
1551 platform)<indexterm><primary>platform</primary></indexterm>.
1552 The difficulty is that there are minor differences between
1553 different platforms; minor, but enough that the code needs to be
1554 a bit different for each. There are some big differences too:
1555 for a different architecture we need to build GHC with a
1556 different native-code generator.</para>
1558 <para>There are also knobs you can turn to control how the
1559 <literal>fptools</literal> software is built. For example, you
1560 might want to build GHC optimised (so that it runs fast) or
1561 unoptimised (so that you can compile it fast after you've
1562 modified it. Or, you might want to compile it with debugging on
1563 (so that extra consistency-checking code gets included) or off.
1566 <para>All of this stuff is called the
1567 <emphasis>configuration</emphasis> of your build. You set the
1568 configuration using a three-step process.</para>
1572 <term>Step 1: get ready for configuration.</term>
1574 <para>NOTE: if you're starting from a source distribution,
1575 rather than CVS sources, you can skip this step.</para>
1577 <para>Change directory to
1578 <constant>$(FPTOOLS_TOP)</constant> and
1580 <command>autoconf</command><indexterm><primary>autoconf</primary></indexterm>
1581 (with no arguments). This GNU program converts
1582 <filename><constant>$(FPTOOLS_TOP)</constant>/configure.in</filename>
1583 to a shell script called
1584 <filename><constant>$(FPTOOLS_TOP)</constant>/configure</filename>.
1587 <para>Some projects, including GHC, have their own
1588 configure script. If there's an
1589 <constant>$(FPTOOLS_TOP)/<project>/configure.in</constant>,
1590 then you need to run <command>autoconf</command> in that
1591 directory too.</para>
1593 <para>Both these steps are completely
1594 platform-independent; they just mean that the
1595 human-written file (<filename>configure.in</filename>) can
1596 be short, although the resulting shell script,
1597 <command>configure</command>, and
1598 <filename>mk/config.h.in</filename>, are long.</para>
1603 <term>Step 2: system configuration.</term>
1605 <para>Runs the newly-created <command>configure</command>
1606 script, thus:</para>
1609 ./configure <optional><parameter>args</parameter></optional>
1612 <para><command>configure</command>'s mission is to scurry
1613 round your computer working out what architecture it has,
1614 what operating system, whether it has the
1615 <Function>vfork</Function> system call, where
1616 <command>yacc</command> is kept, whether
1617 <command>gcc</command> is available, where various obscure
1618 <literal>#include</literal> files are, whether it's a
1619 leap year, and what the systems manager had for lunch. It
1620 communicates these snippets of information in two
1627 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1629 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1630 substituting for things between
1631 “<literal>@</literal>” brackets. So,
1632 “<literal>@HaveGcc@</literal>” will be
1633 replaced by “<literal>YES</literal>” or
1634 “<literal>NO</literal>” depending on what
1635 <command>configure</command> finds.
1636 <filename>mk/config.mk</filename> is included by every
1637 Makefile (directly or indirectly), so the
1638 configuration information is thereby communicated to
1639 all Makefiles.</para>
1643 <para> It translates
1644 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1646 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1647 The latter is <literal>#include</literal>d by
1648 various C programs, which can thereby make use of
1649 configuration information.</para>
1653 <para><command>configure</command> takes some optional
1654 arguments. Use <literal>./configure --help</literal> to
1655 get a list of the available arguments. Here are some of
1656 the ones you might need:</para>
1660 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1661 <indexterm><primary><literal>--with-ghc</literal></primary>
1664 <para>Specifies the path to an installed GHC which
1665 you would like to use. This compiler will be used
1666 for compiling GHC-specific code (eg. GHC itself).
1667 This option <emphasis>cannot</emphasis> be specified
1668 using <filename>build.mk</filename> (see later),
1669 because <command>configure</command> needs to
1670 auto-detect the version of GHC you're using. The
1671 default is to look for a compiler named
1672 <literal>ghc</literal> in your path.</para>
1677 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1678 <indexterm><primary><literal>--with-hc</literal></primary>
1681 <para>Specifies the path to any installed Haskell
1682 compiler. This compiler will be used for compiling
1683 generic Haskell code. The default is to use
1684 <literal>ghc</literal>.</para>
1689 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1690 <indexterm><primary><literal>--with-gcc</literal></primary>
1693 <para>Specifies the path to the installed GCC. This
1694 compiler will be used to compile all C files,
1695 <emphasis>except</emphasis> any generated by the
1696 installed Haskell compiler, which will have its own
1697 idea of which C compiler (if any) to use. The
1698 default is to use <literal>gcc</literal>.</para>
1703 <para><command>configure</command> caches the results of
1704 its run in <filename>config.cache</filename>. Quite often
1705 you don't want that; you're running
1706 <command>configure</command> a second time because
1707 something has changed. In that case, simply delete
1708 <filename>config.cache</filename>.</para>
1713 <term>Step 3: build configuration.</term>
1715 <para>Next, you say how this build of
1716 <literal>fptools</literal> is to differ from the standard
1717 defaults by creating a new file
1718 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1719 <emphasis>in the build tree</emphasis>. This file is the
1720 one and only file you edit in the build tree, precisely
1721 because it says how this build differs from the source.
1722 (Just in case your build tree does die, you might want to
1723 keep a private directory of <filename>build.mk</filename>
1724 files, and use a symbolic link in each build tree to point
1725 to the appropriate one.) So
1726 <filename>mk/build.mk</filename> never exists in the
1727 source tree—you create one in each build tree from
1728 the template. We'll discuss what to put in it
1734 <para>And that's it for configuration. Simple, eh?</para>
1736 <para>What do you put in your build-specific configuration file
1737 <filename>mk/build.mk</filename>? <emphasis>For almost all
1738 purposes all you will do is put make variable definitions that
1739 override those in</emphasis>
1740 <filename>mk/config.mk.in</filename>. The whole point of
1741 <filename>mk/config.mk.in</filename>—and its derived
1742 counterpart <filename>mk/config.mk</filename>—is to define
1743 the build configuration. It is heavily commented, as you will
1744 see if you look at it. So generally, what you do is look at
1745 <filename>mk/config.mk.in</filename>, and add definitions in
1746 <filename>mk/build.mk</filename> that override any of the
1747 <filename>config.mk</filename> definitions that you want to
1748 change. (The override occurs because the main boilerplate file,
1749 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1750 includes <filename>build.mk</filename> after
1751 <filename>config.mk</filename>.)</para>
1753 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1754 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1756 <para>For example, <filename>config.mk.in</filename> contains
1757 the definition:</para>
1760 GhcHcOpts=-O -Rghc-timing
1763 <para>The accompanying comment explains that this is the list of
1764 flags passed to GHC when building GHC itself. For doing
1765 development, it is wise to add <literal>-DDEBUG</literal>, to
1766 enable debugging code. So you would add the following to
1767 <filename>build.mk</filename>:</para>
1769 <para>or, if you prefer,</para>
1772 GhcHcOpts += -DDEBUG
1775 <para>GNU <command>make</command> allows existing definitions to
1776 have new text appended using the “<literal>+=</literal>”
1777 operator, which is quite a convenient feature.)</para>
1779 <para>If you want to remove the <literal>-O</literal> as well (a
1780 good idea when developing, because the turn-around cycle gets a
1781 lot quicker), you can just override
1782 <literal>GhcLibHcOpts</literal> altogether:</para>
1785 GhcHcOpts=-DDEBUG -Rghc-timing
1788 <para>When reading <filename>config.mk.in</filename>, remember
1789 that anything between “@...@” signs is going to be substituted
1790 by <command>configure</command> later. You
1791 <emphasis>can</emphasis> override the resulting definition if
1792 you want, but you need to be a bit surer what you are doing.
1793 For example, there's a line that says:</para>
1799 <para>This defines the Make variables <constant>YACC</constant>
1800 to the pathname for a <command>yacc</command> that
1801 <command>configure</command> finds somewhere. If you have your
1802 own pet <command>yacc</command> you want to use instead, that's
1803 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1809 <para>You do not <emphasis>have</emphasis> to have a
1810 <filename>mk/build.mk</filename> file at all; if you don't,
1811 you'll get all the default settings from
1812 <filename>mk/config.mk.in</filename>.</para>
1814 <para>You can also use <filename>build.mk</filename> to override
1815 anything that <command>configure</command> got wrong. One place
1816 where this happens often is with the definition of
1817 <constant>FPTOOLS_TOP_ABS</constant>: this
1818 variable is supposed to be the canonical path to the top of your
1819 source tree, but if your system uses an automounter then the
1820 correct directory is hard to find automatically. If you find
1821 that <command>configure</command> has got it wrong, just put the
1822 correct definition in <filename>build.mk</filename>.</para>
1826 <sect2 id="sec-storysofar">
1827 <title>The story so far</title>
1829 <para>Let's summarise the steps you need to carry to get
1830 yourself a fully-configured build tree from scratch.</para>
1834 <para> Get your source tree from somewhere (CVS repository
1835 or source distribution). Say you call the root directory
1836 <filename>myfptools</filename> (it does not have to be
1837 called <filename>fptools</filename>). Make sure that you
1838 have the essential files (see <XRef
1839 LinkEnd="sec-source-tree">).</para>
1844 <para>(Optional) Use <command>lndir</command> or
1845 <command>mkshadowdir</command> to create a build tree.</para>
1849 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1852 <para>(N.B. <command>mkshadowdir</command>'s first argument
1853 is taken relative to its second.) You probably want to give
1854 the build tree a name that suggests its main defining
1855 characteristic (in your mind at least), in case you later
1860 <para>Change directory to the build tree. Everything is
1861 going to happen there now.</para>
1864 $ cd /scratch/joe-bloggs/myfptools-sun4
1870 <para>Prepare for system configuration:</para>
1876 <para>(You can skip this step if you are starting from a
1877 source distribution, and you already have
1878 <filename>configure</filename> and
1879 <filename>mk/config.h.in</filename>.)</para>
1881 <para>Some projects, including GHC itself, have their own
1882 configure scripts, so it is necessary to run autoconf again
1883 in the appropriate subdirectories. eg:</para>
1886 $ (cd ghc; autoconf)
1891 <para>Do system configuration:</para>
1897 <para>Don't forget to check whether you need to add any
1898 arguments to <literal>configure</literal>; for example, a
1899 common requirement is to specify which GHC to use with
1900 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1904 <para>Create the file <filename>mk/build.mk</filename>,
1905 adding definitions for your desired configuration
1914 <para>You can make subsequent changes to
1915 <filename>mk/build.mk</filename> as often as you like. You do
1916 not have to run any further configuration programs to make these
1917 changes take effect. In theory you should, however, say
1918 <command>gmake clean</command>, <command>gmake all</command>,
1919 because configuration option changes could affect
1920 anything—but in practice you are likely to know what's
1925 <title>Making things</title>
1927 <para>At this point you have made yourself a fully-configured
1928 build tree, so you are ready to start building real
1931 <para>The first thing you need to know is that <emphasis>you
1932 must use GNU <command>make</command>, usually called
1933 <command>gmake</command>, not standard Unix
1934 <command>make</command></emphasis>. If you use standard Unix
1935 <command>make</command> you will get all sorts of error messages
1936 (but no damage) because the <literal>fptools</literal>
1937 <command>Makefiles</command> use GNU <command>make</command>'s
1938 facilities extensively.</para>
1940 <para>To just build the whole thing, <command>cd</command> to
1941 the top of your <literal>fptools</literal> tree and type
1942 <command>gmake</command>. This will prepare the tree and build
1943 the various projects in the correct order.</para>
1946 <sect2 id="sec-bootstrapping">
1947 <title>Bootstrapping GHC</title>
1949 <para>GHC requires a 2-stage bootstrap in order to provide
1950 full functionality, including GHCi. By a 2-stage bootstrap, we
1951 mean that the compiler is built once using the installed GHC,
1952 and then again using the compiler built in the first stage. You
1953 can also build a stage 3 compiler, but this normally isn't
1954 necessary except to verify that the stage 2 compiler is working
1957 <para>Note that when doing a bootstrap, the stage 1 compiler
1958 must be built, followed by the runtime system and libraries, and
1959 then the stage 2 compiler. The correct ordering is implemented
1960 by the top-level fptools <filename>Makefile</filename>, so if
1961 you want everything to work automatically it's best to start
1962 <command>make</command> from the top of the tree. When building
1963 GHC, the top-level fptools <filename>Makefile</filename> is set
1964 up to do a 2-stage bootstrap by default (when you say
1965 <command>make</command>). Some other targets it supports
1972 <para>Build everything as normal, including the stage 1
1980 <para>Build the stage 2 compiler only.</para>
1987 <para>Build the stage 3 compiler only.</para>
1992 <term>bootstrap</term> <term>bootstrap2</term>
1994 <para>Build stage 1 followed by stage 2.</para>
1999 <term>bootstrap3</term>
2001 <para>Build stages 1, 2 and 3.</para>
2006 <term>install</term>
2008 <para>Install everything, including the compiler built in
2009 stage 2. To override the stage, say <literal>make install
2010 stage=<replaceable>n</replaceable></literal> where
2011 <replaceable>n</replaceable> is the stage to install.</para>
2016 <para>The top-level <filename>Makefile</filename> also arranges
2017 to do the appropriate <literal>make boot</literal> steps (see
2018 below) before actually building anything.</para>
2020 <para>The <literal>stage1</literal>, <literal>stage2</literal>
2021 and <literal>stage3</literal> targets also work in the
2022 <literal>ghc/compiler</literal> directory, but don't forget that
2023 each stage requires its own <literal>make boot</literal> step:
2024 for example, you must do</para>
2026 <screen>$ make boot stage=2</screen>
2028 <para>before <literal>make stage2</literal> in
2029 <literal>ghc/compiler</literal>.</para>
2032 <sect2 id="sec-standard-targets">
2033 <title>Standard Targets</title>
2034 <indexterm><primary>targets, standard makefile</primary></indexterm>
2035 <indexterm><primary>makefile targets</primary></indexterm>
2037 <para>In any directory you should be able to make the following:</para>
2041 <term><literal>boot</literal></term>
2043 <para>does the one-off preparation required to get ready
2044 for the real work. Notably, it does <command>gmake
2045 depend</command> in all directories that contain programs.
2046 It also builds the necessary tools for compilation to
2049 <para>Invoking the <literal>boot</literal> target
2050 explicitly is not normally necessary. From the top-level
2051 <literal>fptools</literal> directory, invoking
2052 <literal>gmake</literal> causes <literal>gmake boot
2053 all</literal> to be invoked in each of the project
2054 subdirectories, in the order specified by
2055 <literal>$(AllTargets)</literal> in
2056 <literal>config.mk</literal>.</para>
2058 <para>If you're working in a subdirectory somewhere and
2059 need to update the dependencies, <literal>gmake
2060 boot</literal> is a good way to do it.</para>
2065 <term><literal>all</literal></term>
2067 <para>makes all the final target(s) for this Makefile.
2068 Depending on which directory you are in a “final
2069 target” may be an executable program, a library
2070 archive, a shell script, or a Postscript file. Typing
2071 <command>gmake</command> alone is generally the same as
2072 typing <command>gmake all</command>.</para>
2077 <term><literal>install</literal></term>
2079 <para>installs the things built by <literal>all</literal>
2080 (except for the documentation). Where does it install
2081 them? That is specified by
2082 <filename>mk/config.mk.in</filename>; you can override it
2083 in <filename>mk/build.mk</filename>, or by running
2084 <command>configure</command> with command-line arguments
2085 like <literal>--bindir=/home/simonpj/bin</literal>; see
2086 <literal>./configure --help</literal> for the full
2092 <term><literal>install-docs</literal></term>
2094 <para>installs the documentation. Otherwise behaves just
2095 like <literal>install</literal>.</para>
2100 <term><literal>uninstall</literal></term>
2102 <para>reverses the effect of
2103 <literal>install</literal>.</para>
2108 <term><literal>clean</literal></term>
2110 <para>Delete all files from the current directory that are
2111 normally created by building the program. Don't delete
2112 the files that record the configuration, or files
2113 generated by <command>gmake boot</command>. Also preserve
2114 files that could be made by building, but normally aren't
2115 because the distribution comes with them.</para>
2120 <term><literal>distclean</literal></term>
2122 <para>Delete all files from the current directory that are
2123 created by configuring or building the program. If you
2124 have unpacked the source and built the program without
2125 creating any other files, <literal>make
2126 distclean</literal> should leave only the files that were
2127 in the distribution.</para>
2132 <term><literal>mostlyclean</literal></term>
2134 <para>Like <literal>clean</literal>, but may refrain from
2135 deleting a few files that people normally don't want to
2141 <term><literal>maintainer-clean</literal></term>
2143 <para>Delete everything from the current directory that
2144 can be reconstructed with this Makefile. This typically
2145 includes everything deleted by
2146 <literal>distclean</literal>, plus more: C source files
2147 produced by Bison, tags tables, Info files, and so
2150 <para>One exception, however: <literal>make
2151 maintainer-clean</literal> should not delete
2152 <filename>configure</filename> even if
2153 <filename>configure</filename> can be remade using a rule
2154 in the <filename>Makefile</filename>. More generally,
2155 <literal>make maintainer-clean</literal> should not delete
2156 anything that needs to exist in order to run
2157 <filename>configure</filename> and then begin to build the
2163 <term><literal>check</literal></term>
2165 <para>run the test suite.</para>
2170 <para>All of these standard targets automatically recurse into
2171 sub-directories. Certain other standard targets do not:</para>
2175 <term><literal>configure</literal></term>
2177 <para>is only available in the root directory
2178 <constant>$(FPTOOLS_TOP)</constant>; it has
2179 been discussed in <XRef
2180 LinkEnd="sec-build-config">.</para>
2185 <term><literal>depend</literal></term>
2187 <para>make a <filename>.depend</filename> file in each
2188 directory that needs it. This <filename>.depend</filename>
2189 file contains mechanically-generated dependency
2190 information; for example, suppose a directory contains a
2191 Haskell source module <filename>Foo.lhs</filename> which
2192 imports another module <literal>Baz</literal>. Then the
2193 generated <filename>.depend</filename> file will contain
2194 the dependency:</para>
2200 <para>which says that the object file
2201 <filename>Foo.o</filename> depends on the interface file
2202 <filename>Baz.hi</filename> generated by compiling module
2203 <literal>Baz</literal>. The <filename>.depend</filename>
2204 file is automatically included by every Makefile.</para>
2209 <term><literal>binary-dist</literal></term>
2211 <para>make a binary distribution. This is the target we
2212 use to build the binary distributions of GHC and
2218 <term><literal>dist</literal></term>
2220 <para>make a source distribution. Note that this target
2221 does “make distclean” as part of its work;
2222 don't use it if you want to keep what you've built.</para>
2227 <para>Most <filename>Makefile</filename>s have targets other
2228 than these. You can discover them by looking in the
2229 <filename>Makefile</filename> itself.</para>
2233 <title>Using a project from the build tree</title>
2235 <para>If you want to build GHC (say) and just use it direct from
2236 the build tree without doing <literal>make install</literal>
2237 first, you can run the in-place driver script:
2238 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2240 <para> Do <emphasis>NOT</emphasis> use
2241 <filename>ghc/compiler/ghc</filename>, or
2242 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2243 scripts intended for installation, and contain hard-wired paths
2244 to the installed libraries, rather than the libraries in the
2247 <para>Happy can similarly be run from the build tree, using
2248 <filename>happy/src/happy-inplace</filename>.</para>
2252 <title>Fast Making</title>
2254 <indexterm><primary>fastmake</primary></indexterm>
2255 <indexterm><primary>dependencies, omitting</primary></indexterm>
2256 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2258 <para>Sometimes the dependencies get in the way: if you've made
2259 a small change to one file, and you're absolutely sure that it
2260 won't affect anything else, but you know that
2261 <command>make</command> is going to rebuild everything anyway,
2262 the following hack may be useful:</para>
2268 <para>This tells the make system to ignore dependencies and just
2269 build what you tell it to. In other words, it's equivalent to
2270 temporarily removing the <filename>.depend</filename> file in
2271 the current directory (where <command>mkdependHS</command> and
2272 friends store their dependency information).</para>
2274 <para>A bit of history: GHC used to come with a
2275 <command>fastmake</command> script that did the above job, but
2276 GNU make provides the features we need to do it without
2277 resorting to a script. Also, we've found that fastmaking is
2278 less useful since the advent of GHC's recompilation checker (see
2279 the User's Guide section on "Separate Compilation").</para>
2283 <sect1 id="sec-makefile-arch">
2284 <title>The <filename>Makefile</filename> architecture</title>
2285 <indexterm><primary>makefile architecture</primary></indexterm>
2287 <para><command>make</command> is great if everything
2288 works—you type <command>gmake install</command> and lo! the
2289 right things get compiled and installed in the right places. Our
2290 goal is to make this happen often, but somehow it often doesn't;
2291 instead some weird error message eventually emerges from the
2292 bowels of a directory you didn't know existed.</para>
2294 <para>The purpose of this section is to give you a road-map to
2295 help you figure out what is going right and what is going
2299 <title>Debugging</title>
2301 <para>Debugging <filename>Makefile</filename>s is something of a
2302 black art, but here's a couple of tricks that we find
2303 particularly useful. The following command allows you to see
2304 the contents of any make variable in the context of the current
2305 <filename>Makefile</filename>:</para>
2307 <screen>$ make show VALUE=HS_SRCS</screen>
2309 <para>where you can replace <literal>HS_SRCS</literal> with the
2310 name of any variable you wish to see the value of.</para>
2312 <para>GNU make has a <option>-d</option> option which generates
2313 a dump of the decision procedure used to arrive at a conclusion
2314 about which files should be recompiled. Sometimes useful for
2315 tracking down problems with superfluous or missing
2316 recompilations.</para>
2320 <title>A small project</title>
2322 <para>To get started, let us look at the
2323 <filename>Makefile</filename> for an imaginary small
2324 <literal>fptools</literal> project, <literal>small</literal>.
2325 Each project in <literal>fptools</literal> has its own directory
2326 in <constant>FPTOOLS_TOP</constant>, so the
2327 <literal>small</literal> project will have its own directory
2328 <constant>FPOOLS_TOP/small/</constant>. Inside the
2329 <filename>small/</filename> directory there will be a
2330 <filename>Makefile</filename>, looking something like
2333 <indexterm><primary>Makefile, minimal</primary></indexterm>
2336 # Makefile for fptools project "small"
2339 include $(TOP)/mk/boilerplate.mk
2341 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2344 include $(TOP)/target.mk
2347 <para>this <filename>Makefile</filename> has three
2352 <para>The first section includes
2355 One of the most important
2356 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2357 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2362 a file of “boilerplate” code from the level
2363 above (which in this case will be
2364 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2365 As its name suggests, <filename>boilerplate.mk</filename>
2366 consists of a large quantity of standard
2367 <filename>Makefile</filename> code. We discuss this
2368 boilerplate in more detail in <XRef LinkEnd="sec-boiler">.
2369 <indexterm><primary>include, directive in
2370 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2371 inclusion</primary></indexterm></para>
2373 <para>Before the <literal>include</literal> statement, you
2374 must define the <command>make</command> variable
2375 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2376 to be the directory containing the <filename>mk</filename>
2377 directory in which the <filename>boilerplate.mk</filename>
2378 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2381 include ../mk/boilerplate.mk # NO NO NO
2385 <para>Why? Because the <filename>boilerplate.mk</filename>
2386 file needs to know where it is, so that it can, in turn,
2387 <literal>include</literal> other files. (Unfortunately,
2388 when an <literal>include</literal>d file does an
2389 <literal>include</literal>, the filename is treated relative
2390 to the directory in which <command>gmake</command> is being
2391 run, not the directory in which the
2392 <literal>include</literal>d sits.) In general,
2393 <emphasis>every file <filename>foo.mk</filename> assumes
2395 <filename><constant>$(TOP)</constant>/mk/foo.mk</filename>
2396 refers to itself.</emphasis> It is up to the
2397 <filename>Makefile</filename> doing the
2398 <literal>include</literal> to ensure this is the case.</para>
2400 <para>Files intended for inclusion in other
2401 <filename>Makefile</filename>s are written to have the
2402 following property: <emphasis>after
2403 <filename>foo.mk</filename> is <literal>include</literal>d,
2404 it leaves <constant>TOP</constant> containing the same value
2405 as it had just before the <literal>include</literal>
2406 statement</emphasis>. In our example, this invariant
2407 guarantees that the <literal>include</literal> for
2408 <filename>target.mk</filename> will look in the same
2409 directory as that for <filename>boilerplate.mk</filename>.</para>
2413 <para> The second section defines the following standard
2414 <command>make</command> variables:
2415 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2416 (the source files from which is to be built), and
2417 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2418 (the executable binary to be built). We will discuss in
2419 more detail what the “standard variables” are,
2420 and how they affect what happens, in <XRef
2421 LinkEnd="sec-targets">.</para>
2423 <para>The definition for <constant>SRCS</constant> uses the
2424 useful GNU <command>make</command> construct
2425 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2426 which expands to a list of all the files matching the
2427 pattern <literal>pat</literal> in the current directory. In
2428 this example, <constant>SRCS</constant> is set to the list
2429 of all the <filename>.lhs</filename> and
2430 <filename>.c</filename> files in the directory. (Let's
2431 suppose there is one of each, <filename>Foo.lhs</filename>
2432 and <filename>Baz.c</filename>.)</para>
2436 <para>The last section includes a second file of standard
2438 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2439 It contains the rules that tell <command>gmake</command> how
2440 to make the standard targets (<Xref
2441 LinkEnd="sec-standard-targets">). Why, you ask, can't this
2442 standard code be part of
2443 <filename>boilerplate.mk</filename>? Good question. We
2444 discuss the reason later, in <Xref
2445 LinkEnd="sec-boiler-arch">.</para>
2447 <para>You do not <emphasis>have</emphasis> to
2448 <literal>include</literal> the
2449 <filename>target.mk</filename> file. Instead, you can write
2450 rules of your own for all the standard targets. Usually,
2451 though, you will find quite a big payoff from using the
2452 canned rules in <filename>target.mk</filename>; the price
2453 tag is that you have to understand what canned rules get
2454 enabled, and what they do (<Xref
2455 LinkEnd="sec-targets">).</para>
2459 <para>In our example <filename>Makefile</filename>, most of the
2460 work is done by the two <literal>include</literal>d files. When
2461 you say <command>gmake all</command>, the following things
2466 <para><command>gmake</command> figures out that the object
2467 files are <filename>Foo.o</filename> and
2468 <filename>Baz.o</filename>.</para>
2472 <para>It uses a boilerplate pattern rule to compile
2473 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2474 using a Haskell compiler. (Which one? That is set in the
2475 build configuration.)</para>
2479 <para>It uses another standard pattern rule to compile
2480 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2481 using a C compiler. (Ditto.)</para>
2485 <para>It links the resulting <filename>.o</filename> files
2486 together to make <literal>small</literal>, using the Haskell
2487 compiler to do the link step. (Why not use
2488 <command>ld</command>? Because the Haskell compiler knows
2489 what standard libraries to link in. How did
2490 <command>gmake</command> know to use the Haskell compiler to
2491 do the link, rather than the C compiler? Because we set the
2492 variable <constant>HS_PROG</constant> rather than
2493 <constant>C_PROG</constant>.)</para>
2497 <para>All <filename>Makefile</filename>s should follow the above
2498 three-section format.</para>
2502 <title>A larger project</title>
2504 <para>Larger projects are usually structured into a number of
2505 sub-directories, each of which has its own
2506 <filename>Makefile</filename>. (In very large projects, this
2507 sub-structure might be iterated recursively, though that is
2508 rare.) To give you the idea, here's part of the directory
2509 structure for the (rather large) GHC project:</para>
2519 ...source files for documentation...
2522 ...source files for driver...
2525 parser/...source files for parser...
2526 renamer/...source files for renamer...
2530 <para>The sub-directories <filename>docs</filename>,
2531 <filename>driver</filename>, <filename>compiler</filename>, and
2532 so on, each contains a sub-component of GHC, and each has its
2533 own <filename>Makefile</filename>. There must also be a
2534 <filename>Makefile</filename> in
2535 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>.
2536 It does most of its work by recursively invoking
2537 <command>gmake</command> on the <filename>Makefile</filename>s
2538 in the sub-directories. We say that
2539 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2540 <filename>Makefile</filename></emphasis>, because it does little
2541 except organise its children, while the
2542 <filename>Makefile</filename>s in the sub-directories are all
2543 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2544 principle the sub-directories might themselves contain a
2545 non-leaf <filename>Makefile</filename> and several
2546 sub-sub-directories, but that does not happen in GHC.)</para>
2548 <para>The <filename>Makefile</filename> in
2549 <filename>ghc/compiler</filename> is considered a leaf
2550 <filename>Makefile</filename> even though the
2551 <filename>ghc/compiler</filename> has sub-directories, because
2552 these sub-directories do not themselves have
2553 <filename>Makefile</filename>s in them. They are just used to
2554 structure the collection of modules that make up GHC, but all
2555 are managed by the single <filename>Makefile</filename> in
2556 <filename>ghc/compiler</filename>.</para>
2558 <para>You will notice that <filename>ghc/</filename> also
2559 contains a directory <filename>ghc/mk/</filename>. It contains
2560 GHC-specific <filename>Makefile</filename> boilerplate code.
2561 More precisely:</para>
2565 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2566 at the top of <filename>ghc/Makefile</filename>, and of all
2567 the leaf <filename>Makefile</filename>s in the
2568 sub-directories. It in turn <literal>include</literal>s the
2569 main boilerplate file
2570 <filename>mk/boilerplate.mk</filename>.</para>
2574 <para><filename>ghc/mk/target.mk</filename> is
2575 <literal>include</literal>d at the bottom of
2576 <filename>ghc/Makefile</filename>, and of all the leaf
2577 <filename>Makefile</filename>s in the sub-directories. It
2578 in turn <literal>include</literal>s the file
2579 <filename>mk/target.mk</filename>.</para>
2583 <para>So these two files are the place to look for GHC-wide
2584 customisation of the standard boilerplate.</para>
2587 <sect2 id="sec-boiler-arch">
2588 <title>Boilerplate architecture</title>
2589 <indexterm><primary>boilerplate architecture</primary></indexterm>
2591 <para>Every <filename>Makefile</filename> includes a
2592 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2593 file at the top, and
2594 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2595 file at the bottom. In this section we discuss what is in these
2596 files, and why there have to be two of them. In general:</para>
2600 <para><filename>boilerplate.mk</filename> consists of:</para>
2604 <para><emphasis>Definitions of millions of
2605 <command>make</command> variables</emphasis> that
2606 collectively specify the build configuration. Examples:
2607 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2608 the options to feed to the Haskell compiler;
2609 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2610 the sub-directories to enable within the
2611 <literal>nofib</literal> project;
2612 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2613 the name of the Haskell compiler to use when compiling
2614 GHC in the <literal>ghc</literal> project.</para>
2618 <para><emphasis>Standard pattern rules</emphasis> that
2619 tell <command>gmake</command> how to construct one file
2620 from another.</para>
2624 <para><filename>boilerplate.mk</filename> needs to be
2625 <literal>include</literal>d at the <emphasis>top</emphasis>
2626 of each <filename>Makefile</filename>, so that the user can
2627 replace the boilerplate definitions or pattern rules by
2628 simply giving a new definition or pattern rule in the
2629 <filename>Makefile</filename>. <command>gmake</command>
2630 simply takes the last definition as the definitive one.</para>
2632 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2633 definitions, it is also quite common to
2634 <emphasis>augment</emphasis> them. For example, a
2635 <filename>Makefile</filename> might say:</para>
2641 <para>thereby adding “<option>-O</option>” to
2643 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2647 <para><filename>target.mk</filename> contains
2648 <command>make</command> rules for the standard targets
2649 described in <Xref LinkEnd="sec-standard-targets">. These
2650 rules are selectively included, depending on the setting of
2651 certain <command>make</command> variables. These variables
2652 are usually set in the middle section of the
2653 <filename>Makefile</filename> between the two
2654 <literal>include</literal>s.</para>
2656 <para><filename>target.mk</filename> must be included at the
2657 end (rather than being part of
2658 <filename>boilerplate.mk</filename>) for several tiresome
2664 <para><command>gmake</command> commits target and
2665 dependency lists earlier than it should. For example,
2666 <FIlename>target.mk</FIlename> has a rule that looks
2670 $(HS_PROG) : $(OBJS)
2671 $(HC) $(LD_OPTS) $< -o $@
2674 <para>If this rule was in
2675 <filename>boilerplate.mk</filename> then
2676 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2678 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2679 would not have their final values at the moment
2680 <command>gmake</command> encountered the rule. Alas,
2681 <command>gmake</command> takes a snapshot of their
2682 current values, and wires that snapshot into the rule.
2683 (In contrast, the commands executed when the rule
2684 “fires” are only substituted at the moment
2685 of firing.) So, the rule must follow the definitions
2686 given in the <filename>Makefile</filename> itself.</para>
2690 <para>Unlike pattern rules, ordinary rules cannot be
2691 overriden or replaced by subsequent rules for the same
2692 target (at least, not without an error message).
2693 Including ordinary rules in
2694 <filename>boilerplate.mk</filename> would prevent the
2695 user from writing rules for specific targets in specific
2700 <para>There are a couple of other reasons I've
2701 forgotten, but it doesn't matter too much.</para>
2708 <sect2 id="sec-boiler">
2709 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2710 <indexterm><primary>boilerplate.mk</primary></indexterm>
2712 <para>If you look at
2713 <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename>
2714 you will find that it consists of the following sections, each
2715 held in a separate file:</para>
2719 <term><filename>config.mk</filename></term>
2720 <indexterm><primary>config.mk</primary></indexterm>
2722 <para>is the build configuration file we discussed at
2723 length in <Xref LinkEnd="sec-build-config">.</para>
2728 <term><filename>paths.mk</filename></term>
2729 <indexterm><primary>paths.mk</primary></indexterm>
2731 <para>defines <command>make</command> variables for
2732 pathnames and file lists. This file contains code for
2733 automatically compiling lists of source files and deriving
2734 lists of object files from those. The results can be
2735 overriden in the <filename>Makefile</filename>, but in
2736 most cases the automatic setup should do the right
2739 <para>The following variables may be set in the
2740 <filename>Makefile</filename> to affect how the automatic
2741 source file search is done:</para>
2745 <term><literal>ALL_DIRS</literal></term>
2746 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2749 <para>Set to a list of directories to search in
2750 addition to the current directory for source
2756 <term><literal>EXCLUDE_SRCS</literal></term>
2757 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2760 <para>Set to a list of source files (relative to the
2761 current directory) to omit from the automatic
2762 search. The source searching machinery is clever
2763 enough to know that if you exclude a source file
2764 from which other sources are derived, then the
2765 derived sources should also be excluded. For
2766 example, if you set <literal>EXCLUDED_SRCS</literal>
2767 to include <filename>Foo.y</filename>, then
2768 <filename>Foo.hs</filename> will also be
2774 <term><literal>EXTRA_SRCS</literal></term>
2775 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2778 <para>Set to a list of extra source files (perhaps
2779 in directories not listed in
2780 <literal>ALL_DIRS</literal>) that should be
2786 <para>The results of the automatic source file search are
2787 placed in the following make variables:</para>
2791 <term><literal>SRCS</literal></term>
2792 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2794 <para>All source files found, sorted and without
2795 duplicates, including those which might not exist
2796 yet but will be derived from other existing sources.
2797 <literal>SRCS</literal> <emphasis>can</emphasis> be
2798 overriden if necessary, in which case the variables
2799 below will follow suit.</para>
2804 <term><literal>HS_SRCS</literal></term>
2805 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2807 <para>all Haskell source files in the current
2808 directory, including those derived from other source
2809 files (eg. Happy sources also give rise to Haskell
2815 <term><literal>HS_OBJS</literal></term>
2816 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2818 <para>Object files derived from
2819 <literal>HS_SRCS</literal>.</para>
2824 <term><literal>HS_IFACES</literal></term>
2825 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2827 <para>Interface files (<literal>.hi</literal> files)
2828 derived from <literal>HS_SRCS</literal>.</para>
2833 <term><literal>C_SRCS</literal></term>
2834 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2836 <para>All C source files found.</para>
2841 <term><literal>C_OBJS</literal></term>
2842 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2844 <para>Object files derived from
2845 <literal>C_SRCS</literal>.</para>
2850 <term><literal>SCRIPT_SRCS</literal></term>
2851 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2853 <para>All script source files found
2854 (<literal>.lprl</literal> files).</para>
2859 <term><literal>SCRIPT_OBJS</literal></term>
2860 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2862 <para><quote>object</quote> files derived from
2863 <literal>SCRIPT_SRCS</literal>
2864 (<literal>.prl</literal> files).</para>
2869 <term><literal>HSC_SRCS</literal></term>
2870 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2872 <para>All <literal>hsc2hs</literal> source files
2873 (<literal>.hsc</literal> files).</para>
2878 <term><literal>HAPPY_SRCS</literal></term>
2879 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2881 <para>All <literal>happy</literal> source files
2882 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2887 <term><literal>OBJS</literal></term>
2888 <indexterm><primary>OBJS</primary></indexterm>
2890 <para>the concatenation of
2891 <literal>$(HS_OBJS)</literal>,
2892 <literal>$(C_OBJS)</literal>, and
2893 <literal>$(SCRIPT_OBJS)</literal>.</para>
2898 <para>Any or all of these definitions can easily be
2899 overriden by giving new definitions in your
2900 <filename>Makefile</filename>.</para>
2902 <para>What, exactly, does <filename>paths.mk</filename>
2903 consider a <quote>source file</quote> to be? It's based
2904 on the file's suffix (e.g. <filename>.hs</filename>,
2905 <filename>.lhs</filename>, <filename>.c</filename>,
2906 <filename>.hy</filename>, etc), but this is the kind of
2907 detail that changes, so rather than enumerate the source
2908 suffices here the best thing to do is to look in
2909 <filename>paths.mk</filename>.</para>
2914 <term><filename>opts.mk</filename></term>
2915 <indexterm><primary>opts.mk</primary></indexterm>
2917 <para>defines <command>make</command> variables for option
2918 strings to pass to each program. For example, it defines
2919 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2920 the option strings to pass to the Haskell compiler. See
2921 <Xref LinkEnd="sec-suffix">.</para>
2926 <term><filename>suffix.mk</filename></term>
2927 <indexterm><primary>suffix.mk</primary></indexterm>
2929 <para>defines standard pattern rules—see <Xref
2930 LinkEnd="sec-suffix">.</para>
2935 <para>Any of the variables and pattern rules defined by the
2936 boilerplate file can easily be overridden in any particular
2937 <filename>Makefile</filename>, because the boilerplate
2938 <literal>include</literal> comes first. Definitions after this
2939 <literal>include</literal> directive simply override the default
2940 ones in <filename>boilerplate.mk</filename>.</para>
2943 <sect2 id="sec-suffix">
2944 <title>Pattern rules and options</title>
2945 <indexterm><primary>Pattern rules</primary></indexterm>
2948 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2949 defines standard <emphasis>pattern rules</emphasis> that say how
2950 to build one kind of file from another, for example, how to
2951 build a <filename>.o</filename> file from a
2952 <filename>.c</filename> file. (GNU <command>make</command>'s
2953 <emphasis>pattern rules</emphasis> are more powerful and easier
2954 to use than Unix <command>make</command>'s <emphasis>suffix
2955 rules</emphasis>.)</para>
2957 <para>Almost all the rules look something like this:</para>
2962 $(CC) $(CC_OPTS) -c $< -o $@
2965 <para>Here's how to understand the rule. It says that
2966 <emphasis>something</emphasis><filename>.o</filename> (say
2967 <filename>Foo.o</filename>) can be built from
2968 <emphasis>something</emphasis><filename>.c</filename>
2969 (<filename>Foo.c</filename>), by invoking the C compiler (path
2970 name held in <constant>$(CC)</constant>), passing to it
2971 the options <constant>$(CC_OPTS)</constant> and
2972 the rule's dependent file of the rule
2973 <literal>$<</literal> (<filename>Foo.c</filename> in
2974 this case), and putting the result in the rule's target
2975 <literal>$@</literal> (<filename>Foo.o</filename> in this
2978 <para>Every program is held in a <command>make</command>
2979 variable defined in <filename>mk/config.mk</filename>—look
2980 in <filename>mk/config.mk</filename> for the complete list. One
2981 important one is the Haskell compiler, which is called
2982 <constant>$(HC)</constant>.</para>
2984 <para>Every program's options are are held in a
2985 <command>make</command> variables called
2986 <constant><prog>_OPTS</constant>. the
2987 <constant><prog>_OPTS</constant> variables are
2988 defined in <filename>mk/opts.mk</filename>. Almost all of them
2989 are defined like this:</para>
2992 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
2995 <para>The four variables from which
2996 <constant>CC_OPTS</constant> is built have the following
3001 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
3003 <para>options passed to all C compilations.</para>
3008 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
3010 <para>options passed to C compilations for way
3011 <literal><way></literal>. For example,
3012 <constant>WAY_mp_CC_OPTS</constant>
3013 gives options to pass to the C compiler when compiling way
3014 <literal>mp</literal>. The variable
3015 <constant>WAY_CC_OPTS</constant> holds
3016 options to pass to the C compiler when compiling the
3017 standard way. (<Xref LinkEnd="sec-ways"> dicusses
3018 multi-way compilation.)</para>
3023 <term><constant><module>_CC_OPTS</constant>:</term>
3025 <para>options to pass to the C compiler that are specific
3026 to module <literal><module></literal>. For example,
3027 <constant>SMap_CC_OPTS</constant> gives the
3028 specific options to pass to the C compiler when compiling
3029 <filename>SMap.c</filename>.</para>
3034 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3036 <para>extra options to pass to all C compilations. This
3037 is intended for command line use, thus:</para>
3040 gmake libHS.a EXTRA_CC_OPTS="-v"
3047 <sect2 id="sec-targets">
3048 <title>The main <filename>mk/target.mk</filename> file</title>
3049 <indexterm><primary>target.mk</primary></indexterm>
3051 <para><filename>target.mk</filename> contains canned rules for
3052 all the standard targets described in <Xref
3053 LinkEnd="sec-standard-targets">. It is complicated by the fact
3054 that you don't want all of these rules to be active in every
3055 <filename>Makefile</filename>. Rather than have a plethora of
3056 tiny files which you can include selectively, there is a single
3057 file, <filename>target.mk</filename>, which selectively includes
3058 rules based on whether you have defined certain variables in
3059 your <filename>Makefile</filename>. This section explains what
3060 rules you get, what variables control them, and what the rules
3061 do. Hopefully, you will also get enough of an idea of what is
3062 supposed to happen that you can read and understand any weird
3063 special cases yourself.</para>
3067 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3069 <para>If <constant>HS_PROG</constant> is defined,
3070 you get rules with the following targets:</para>
3074 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3076 <para>itself. This rule links
3077 <constant>$(OBJS)</constant> with the Haskell
3078 runtime system to get an executable called
3079 <constant>$(HS_PROG)</constant>.</para>
3084 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3087 <constant>$(HS_PROG)</constant> in
3088 <constant>$(bindir)</constant>.</para>
3097 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3099 <para>is similar to <constant>HS_PROG</constant>,
3100 except that the link step links
3101 <constant>$(C_OBJS)</constant> with the C
3102 runtime system.</para>
3107 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3109 <para>is similar to <constant>HS_PROG</constant>,
3110 except that it links
3111 <constant>$(LIB_OBJS)</constant> to make the
3112 library archive <constant>$(LIBRARY)</constant>,
3113 and <literal>install</literal> installs it in
3114 <constant>$(libdir)</constant>.</para>
3119 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3121 <para>…</para>
3126 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3128 <para>…</para>
3133 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3135 <para>If <constant>HS_SRCS</constant> is defined
3136 and non-empty, a rule for the target
3137 <literal>depend</literal> is included, which generates
3138 dependency information for Haskell programs. Similarly
3139 for <constant>C_SRCS</constant>.</para>
3144 <para>All of these rules are “double-colon” rules,
3148 install :: $(HS_PROG)
3149 ...how to install it...
3152 <para>GNU <command>make</command> treats double-colon rules as
3153 separate entities. If there are several double-colon rules for
3154 the same target it takes each in turn and fires it if its
3155 dependencies say to do so. This means that you can, for
3156 example, define both <constant>HS_PROG</constant> and
3157 <constant>LIBRARY</constant>, which will generate two rules for
3158 <literal>install</literal>. When you type <command>gmake
3159 install</command> both rules will be fired, and both the program
3160 and the library will be installed, just as you wanted.</para>
3163 <sect2 id="sec-subdirs">
3164 <title>Recursion</title>
3165 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3166 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3168 <para>In leaf <filename>Makefile</filename>s the variable
3169 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3170 is undefined. In non-leaf <filename>Makefile</filename>s,
3171 <constant>SUBDIRS</constant> is set to the list of
3172 sub-directories that contain subordinate
3173 <filename>Makefile</filename>s. <emphasis>It is up to you to
3174 set <constant>SUBDIRS</constant> in the
3175 <filename>Makefile</filename>.</emphasis> There is no automation
3176 here—<constant>SUBDIRS</constant> is too important to
3179 <para>When <constant>SUBDIRS</constant> is defined,
3180 <filename>target.mk</filename> includes a rather neat rule for
3181 the standard targets (<Xref LinkEnd="sec-standard-targets"> that
3182 simply invokes <command>make</command> recursively in each of
3183 the sub-directories.</para>
3185 <para><emphasis>These recursive invocations are guaranteed to
3186 occur in the order in which the list of directories is specified
3187 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3188 be important. For example, when you say <command>gmake
3189 boot</command> it can be important that the recursive invocation
3190 of <command>make boot</command> is done in one sub-directory
3191 (the include files, say) before another (the source files).
3192 Generally, put the most independent sub-directory first, and the
3193 most dependent last.</para>
3196 <sect2 id="sec-ways">
3197 <title>Way management</title>
3198 <indexterm><primary>way management</primary></indexterm>
3200 <para>We sometimes want to build essentially the same system in
3201 several different “ways”. For example, we want to build GHC's
3202 <literal>Prelude</literal> libraries with and without profiling,
3203 so that there is an appropriately-built library archive to link
3204 with when the user compiles his program. It would be possible
3205 to have a completely separate build tree for each such “way”,
3206 but it would be horribly bureaucratic, especially since often
3207 only parts of the build tree need to be constructed in multiple
3211 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3212 contains some clever magic to allow you to build several
3213 versions of a system; and to control locally how many versions
3214 are built and how they differ. This section explains the
3217 <para>The files for a particular way are distinguished by
3218 munging the suffix. The <quote>normal way</quote> is always
3219 built, and its files have the standard suffices
3220 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3221 In addition, you can build one or more extra ways, each
3222 distinguished by a <emphasis>way tag</emphasis>. The object
3223 files and interface files for one of these extra ways are
3224 distinguished by their suffix. For example, way
3225 <literal>mp</literal> has files
3226 <filename>.mp_o</filename> and
3227 <filename>.mp_hi</filename>. Library archives have their
3228 way tag the other side of the dot, for boring reasons; thus,
3229 <filename>libHS_mp.a</filename>.</para>
3231 <para>A <command>make</command> variable called
3232 <constant>way</constant> holds the current way tag.
3233 <emphasis><constant>way</constant> is only ever set on the
3234 command line of <command>gmake</command></emphasis> (usually in
3235 a recursive invocation of <command>gmake</command> by the
3236 system). It is never set inside a
3237 <filename>Makefile</filename>. So it is a global constant for
3238 any one invocation of <command>gmake</command>. Two other
3239 <command>make</command> variables,
3240 <constant>way_</constant> and
3241 <constant>_way</constant> are immediately derived from
3242 <constant>$(way)</constant> and never altered. If
3243 <constant>way</constant> is not set, then neither are
3244 <constant>way_</constant> and
3245 <constant>_way</constant>, and the invocation of
3246 <command>make</command> will build the <quote>normal
3247 way</quote>. If <constant>way</constant> is set, then the other
3248 two variables are set in sympathy. For example, if
3249 <constant>$(way)</constant> is “<literal>mp</literal>”,
3250 then <constant>way_</constant> is set to
3251 “<literal>mp_</literal>” and
3252 <constant>_way</constant> is set to
3253 “<literal>_mp</literal>”. These three variables are
3254 then used when constructing file names.</para>
3256 <para>So how does <command>make</command> ever get recursively
3257 invoked with <constant>way</constant> set? There are two ways
3258 in which this happens:</para>
3262 <para>For some (but not all) of the standard targets, when
3263 in a leaf sub-directory, <command>make</command> is
3264 recursively invoked for each way tag in
3265 <constant>$(WAYS)</constant>. You set
3266 <constant>WAYS</constant> in the
3267 <filename>Makefile</filename> to the list of way tags you
3268 want these targets built for. The mechanism here is very
3269 much like the recursive invocation of
3270 <command>make</command> in sub-directories (<Xref
3271 LinkEnd="sec-subdirs">). It is up to you to set
3272 <constant>WAYS</constant> in your
3273 <filename>Makefile</filename>; this is how you control what
3274 ways will get built.</para>
3278 <para>For a useful collection of targets (such as
3279 <filename>libHS_mp.a</filename>,
3280 <filename>Foo.mp_o</filename>) there is a rule which
3281 recursively invokes <command>make</command> to make the
3282 specified target, setting the <constant>way</constant>
3283 variable. So if you say <command>gmake
3284 Foo.mp_o</command> you should see a recursive
3285 invocation <command>gmake Foo.mp_o way=mp</command>,
3286 and <emphasis>in this recursive invocation the pattern rule
3287 for compiling a Haskell file into a <filename>.o</filename>
3288 file will match</emphasis>. The key pattern rules (in
3289 <filename>suffix.mk</filename>) look like this:
3293 $(HC) $(HC_OPTS) $< -o $@
3300 <para>You can invoke <command>make</command> with a
3301 particular <literal>way</literal> setting yourself, in order
3302 to build files related to a particular
3303 <literal>way</literal> in the current directory. eg.
3309 will build files for the profiling way only in the current
3316 <title>When the canned rule isn't right</title>
3318 <para>Sometimes the canned rule just doesn't do the right thing.
3319 For example, in the <literal>nofib</literal> suite we want the
3320 link step to print out timing information. The thing to do here
3321 is <emphasis>not</emphasis> to define
3322 <constant>HS_PROG</constant> or
3323 <constant>C_PROG</constant>, and instead define a special
3324 purpose rule in your own <filename>Makefile</filename>. By
3325 using different variable names you will avoid the canned rules
3326 being included, and conflicting with yours.</para>
3330 <sect1 id="building-docs">
3331 <title>Building the documentation</title>
3333 <sect2 id="pre-supposed-doc-tools">
3334 <title>Tools for building the Documentation</title>
3336 <para>The following additional tools are required if you want to
3337 format the documentation that comes with the
3338 <literal>fptools</literal> projects:</para>
3342 <term>DocBook</term>
3343 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3344 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3346 <para>Much of our documentation is written in SGML, using
3347 the DocBook DTD. Instructions on installing and
3348 configuring the DocBook tools are below.</para>
3354 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3355 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3357 <para>A decent TeX distribution is required if you want to
3358 produce printable documentation. We recomment teTeX,
3359 which includes just about everything you need.</para>
3364 <term>Haddock</term>
3365 <indexterm><primary>Haddock</primary>
3368 <para>Haddock is a Haskell documentation tool that we use
3369 for automatically generating documentation from the
3370 library source code. It is an <literal>fptools</literal>
3371 project in itself. To build documentation for the
3372 libraries (<literal>fptools/libraries</literal>) you
3373 should check out and build Haddock in
3374 <literal>fptools/haddock</literal>. Haddock requires GHC
3382 <title>Installing the DocBook tools</title>
3385 <title>Installing the DocBook tools on Linux</title>
3387 <para>If you're on a recent RedHat system (7.0+), you probably
3388 have working DocBook tools already installed. The configure
3389 script should detect your setup and you're away.</para>
3391 <para>If you don't have DocBook tools installed, and you are
3392 using a system that can handle RedHat RPM packages, you can
3393 probably use the <ULink
3394 URL="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3395 DocBook tools</ULink>, which is the most shrink-wrapped SGML
3396 suite that we could find. You need all the RPMs except for
3397 psgml (i.e. <Filename>docbook</Filename>,
3398 <Filename>jade</Filename>, <Filename>jadetex</Filename>,
3399 <Filename>sgmlcommon</Filename> and
3400 <Filename>stylesheets</Filename>). Note that most of these
3401 RPMs are architecture neutral, so are likely to be found in a
3402 <Filename>noarch</Filename> directory. The SuSE RPMs also
3403 work; the RedHat ones <Emphasis>don't</Emphasis> in RedHat 6.2
3404 (7.0 and later should be OK), but they are easy to fix: just
3406 <Filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</Filename>
3407 to <Filename>/usr/lib/sgml/lib/dblib.dsl</Filename>. </para>
3411 <title>Installing DocBook on FreeBSD</title>
3413 <para>On FreeBSD systems, the easiest way to get DocBook up
3414 and running is to install it from the ports tree or a
3415 pre-compiled package (packages are available from your local
3416 FreeBSD mirror site).</para>
3418 <para>To use the ports tree, do this:
3420 $ cd /usr/ports/textproc/docproj
3423 This installs the FreeBSD documentation project tools, which
3424 includes everything needed to format the GHC
3425 documentation.</para>
3429 <title>Installing from binaries on Windows</title>
3431 <Para>It's a good idea to use Norman Walsh's <ULink
3432 URL="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3433 notes</ULink> as a guide. You should get version 3.1 of
3434 DocBook, and note that his file <Filename>test.sgm</Filename>
3435 won't work, as it needs version 3.0. You should unpack Jade
3436 into <Filename>\Jade</Filename>, along with the entities,
3437 DocBook into <Filename>\docbook</Filename>, and the DocBook
3438 stylesheets into <Filename>\docbook\stylesheets</Filename> (so
3439 they actually end up in
3440 <Filename>\docbook\stylesheets\docbook</Filename>).</para>
3445 <title>Installing the DocBook tools from source</title>
3450 <para>Install <ULink
3451 URL="http://openjade.sourceforge.net/">OpenJade</ULink>
3452 (Windows binaries are available as well as sources). If you
3453 want DVI, PS, or PDF then install JadeTeX from the
3454 <Filename>dsssl</Filename> subdirectory. (If you get the
3458 ! LaTeX Error: Unknown option implicit=false' for package hyperref'.
3461 your version of <Command>hyperref</Command> is out of date;
3462 download it from CTAN
3463 (<Filename>macros/latex/contrib/supported/hyperref</Filename>),
3464 and make it, ensuring that you have first removed or renamed
3465 your old copy. If you start getting file not found errors
3466 when making the test for <Command>hyperref</Command>, you
3467 can abort at that point and proceed straight to
3468 <Command>make install</Command>, or enter them as
3469 <Filename>../</Filename><Emphasis>filename</Emphasis>.)</para>
3471 <para>Make links from <Filename>virtex</Filename> to
3472 <Filename>jadetex</Filename> and
3473 <Filename>pdfvirtex</Filename> to
3474 <Filename>pdfjadetex</Filename> (otherwise DVI, PostScript
3475 and PDF output will not work). Copy
3476 <Filename>dsssl/*.{dtd,dsl}</Filename> and
3477 <Filename>catalog</Filename> to
3478 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3482 <title>DocBook and the DocBook stylesheets</title>
3484 <para>Get a Zip of <ULink
3485 URL="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ULink>
3486 and install the contents in
3487 <Filename>/usr/[local/]/lib/sgml</Filename>.</para>
3489 <para>Get the <ULink
3490 URL="http://nwalsh.com/docbook/dsssl/">DocBook
3491 stylesheets</ULink> and install in
3492 <Filename>/usr/[local/]lib/sgml/stylesheets</Filename>
3493 (thereby creating a subdirectory docbook). For indexing,
3494 copy or link <Filename>collateindex.pl</Filename> from the
3495 DocBook stylesheets archive in <Filename>bin</Filename> into
3496 a directory on your <Constant>PATH</Constant>.</para>
3498 <para>Download the <ULink
3499 URL="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3500 entities</ULink> into
3501 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3507 <title>Configuring the DocBook tools</title>
3509 <Para>Once the DocBook tools are installed, the configure script
3510 will detect them and set up the build system accordingly. If you
3511 have a system that isn't supported, let us know, and we'll try
3516 <title>Remaining problems</title>
3518 <para>If you install from source, you'll get a pile of warnings
3521 <Screen>DTDDECL catalog entries are not supported</Screen>
3523 every time you build anything. These can safely be ignored, but
3524 if you find them tedious you can get rid of them by removing all
3525 the <Constant>DTDDECL</Constant> entries from
3526 <Filename>docbook.cat</Filename>.</para>
3530 <title>Building the documentation</title>
3532 <para>To build documentation in a certain format, you can
3533 say, for example,</para>
3539 <para>to build HTML documentation below the current directory.
3540 The available formats are: <literal>dvi</literal>,
3541 <literal>ps</literal>, <literal>pdf</literal>,
3542 <literal>html</literal>, and <literal>rtf</literal>. Note that
3543 not all documentation can be built in all of these formats: HTML
3544 documentation is generally supported everywhere, and DocBook
3545 documentation might support the other formats (depending on what
3546 other tools you have installed).</para>
3548 <para>All of these targets are recursive; that is, saying
3549 <literal>make html</literal> will make HTML docs for all the
3550 documents recursively below the current directory.</para>
3552 <para>Because there are many different formats that the DocBook
3553 documentation can be generated in, you have to select which ones
3554 you want by setting the <literal>SGMLDocWays</literal> variable
3555 to a list of them. For example, in
3556 <filename>build.mk</filename> you might have a line:</para>
3559 SGMLDocWays = html ps
3562 <para>This will cause the documentation to be built in the requested
3563 formats as part of the main build (the default is not to build
3564 any documentation at all).</para>
3568 <title>Installing the documentation</title>
3570 <para>To install the documentation, use:</para>
3576 <para>This will install the documentation into
3577 <literal>$(datadir)</literal> (which defaults to
3578 <literal>$(prefix)/share</literal>). The exception is HTML
3579 documentation, which goes into
3580 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3582 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3583 to a list of formats, the <literal>install-docs</literal> target
3584 won't do anything for SGML documentation.</para>
3590 <sect1 id="sec-porting-ghc">
3591 <title>Porting GHC</title>
3593 <para>This section describes how to port GHC to a currenly
3594 unsupported platform. There are two distinct
3595 possibilities:</para>
3599 <para>The hardware architecture for your system is already
3600 supported by GHC, but you're running an OS that isn't
3601 supported (or perhaps has been supported in the past, but
3602 currently isn't). This is the easiest type of porting job,
3603 but it still requires some careful bootstrapping. Proceed to
3604 <xref linkend="sec-booting-from-hc">.</para>
3608 <para>Your system's hardware architecture isn't supported by
3609 GHC. This will be a more difficult port (though by comparison
3610 perhaps not as difficult as porting gcc). Proceed to <xref
3611 linkend="unregisterised-porting">.</para>
3615 <sect2 id="sec-booting-from-hc">
3616 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3618 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3619 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3620 <indexterm><primary>porting GHC</primary></indexterm>
3622 <para>Bootstrapping GHC on a system without GHC already
3623 installed is achieved by taking the intermediate C files (known
3624 as HC files) from a GHC compilation on a supported system to the
3625 target machine, and compiling them using gcc to get a working
3628 <para><emphasis>NOTE: GHC versions 5.xx and later are
3629 significantly harder to bootstrap from C than earlier versions.
3630 We recommend starting from version 4.08.2 if you need to
3631 bootstrap in this way.</emphasis></para>
3633 <para>HC files are architecture-dependent (but not
3634 OS-dependent), so you have to get a set that were generated on
3635 similar hardware. There may be some supplied on the GHC
3636 download page, otherwise you'll have to compile some up
3637 yourself, or start from <emphasis>unregisterised</emphasis> HC
3638 files - see <xref linkend="unregisterised-porting">.</para>
3640 <para>The following steps should result in a working GHC build
3641 with full libraries:</para>
3645 <para>Unpack the HC files on top of a fresh source tree
3646 (make sure the source tree version matches the version of
3647 the HC files <emphasis>exactly</emphasis>!). This will
3648 place matching <filename>.hc</filename> files next to the
3649 corresponding Haskell source (<filename>.hs</filename> or
3650 <filename>.lhs</filename>) in the compiler subdirectory
3651 <filename>ghc/compiler</filename> and in the libraries
3652 (subdirectories of <filename>hslibs</filename> and
3653 <literal>libraries</literal>).</para>
3657 <para>The actual build process is fully automated by the
3658 <filename>hc-build</filename> script located in the
3659 <filename>distrib</filename> directory. If you eventually
3660 want to install GHC into the directory
3661 <replaceable>dir</replaceable>, the following
3662 command will execute the whole build process (it won't
3663 install yet):</para>
3666 foo% distrib/hc-build --prefix=<replaceable>dir</replaceable>
3668 <indexterm><primary>--hc-build</primary></indexterm>
3670 <para>By default, the installation directory is
3671 <filename>/usr/local</filename>. If that is what you want,
3672 you may omit the argument to <filename>hc-build</filename>.
3673 Generally, any option given to <filename>hc-build</filename>
3674 is passed through to the configuration script
3675 <filename>configure</filename>. If
3676 <filename>hc-build</filename> successfully completes the
3677 build process, you can install the resulting system, as
3687 <sect2 id="unregisterised-porting">
3688 <title>Porting GHC to a new architecture</title>
3690 <para>The first step in porting to a new architecture is to get
3691 an <firstterm>unregisterised</firstterm> build working. An
3692 unregisterised build is one that compiles via vanilla C only.
3693 By contrast, a registerised build uses the following
3694 architecture-specific hacks for speed:</para>
3698 <para>Global register variables: certain abstract machine
3699 <quote>registers</quote> are mapped to real machine
3700 registers, depending on how many machine registers are
3702 <filename>ghc/includes/MachRegs.h</filename>).</para>
3706 <para>Assembly-mangling: when compiling via C, we feed the
3707 assembly generated by gcc though a Perl script known as the
3708 <firstterm>mangler</firstterm> (see
3709 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3710 mangler rearranges the assembly to support tail-calls and
3711 various other optimisations.</para>
3715 <para>In an unregisterised build, neither of these hacks are
3716 used — the idea is that the C code generated by the
3717 compiler should compile using gcc only. The lack of these
3718 optimisations costs about a factor of two in performance, but
3719 since unregisterised compilation is usually just a step on the
3720 way to a full registerised port, we don't mind too much.</para>
3723 <title>Building an unregisterised port</title>
3725 <para>The first step is to get some unregisterised HC files.
3726 Either (a) download them from the GHC site (if there are
3727 some available for the right version of GHC), or
3728 (b) build them yourself on any machine with a working
3729 GHC. If at all possible this should be a machine with the
3730 same word size as the target.</para>
3732 <para>There is a script available which should automate the
3733 process of doing the 2-stage bootstrap necessary to get the
3734 unregisterised HC files - it's available in <ulink
3735 url="http://cvs.haskell.org/cgi-bin/cvsweb.cgi/fptools/distrib/cross-port"><filename>fptools/distrib/cross-port</filename></ulink>
3738 <para>Now take these unregisterised HC files to the target
3739 platform and bootstrap a compiler from them as per the
3740 instructions in <xref linkend="sec-booting-from-hc">. In
3741 <filename>build.mk</filename>, you need to tell the build
3742 system that the compiler you're building is
3743 (a) unregisterised itself, and (b) builds
3744 unregisterised binaries. This varies depending on the GHC
3745 version you're bootstraping:</para>
3748 # build.mk for GHC 4.08.x
3749 GhcWithRegisterised=NO
3753 # build.mk for GHC 5.xx and 6.x
3754 GhcUnregisterised=YES
3757 <para>Versions 5.xx and 6.x only: use the option
3758 <option>--enable-hc-boot-unregisterised</option> instead of
3759 <option>--enable-hc-boot</option> when running
3760 <filename>./configure</filename>.</para>
3762 <para>The build may not go through cleanly. We've tried to
3763 stick to writing portable code in most parts of the compiler,
3764 so it should compile on any POSIXish system with gcc, but in
3765 our experience most systems differ from the standards in one
3766 way or another. Deal with any problems as they arise - if you
3767 get stuck, ask the experts on
3768 <email>glasgow-haskell-users@haskell.org</email>.</para>
3770 <para>Once you have the unregisterised compiler up and
3771 running, you can use it to start a registerised port. The
3772 following sections describe the various parts of the system
3773 that will need architecture-specific tweaks in order to get a
3774 registerised build going.</para>
3776 <para>Lots of useful information about the innards of GHC is
3777 available in the <ulink
3778 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3779 Commentary</ulink>, which might be helpful if you run into
3780 some code which needs tweaking for your system.</para>
3784 <title>Porting the RTS</title>
3786 <para>The following files need architecture-specific code for a
3787 registerised build:</para>
3791 <term><filename>ghc/includes/MachRegs.h</filename></term>
3792 <indexterm><primary><filename>MachRegs.h</filename></primary>
3795 <para>Defines the STG-register to machine-register
3796 mapping. You need to know your platform's C calling
3797 convention, and which registers are generally available
3798 for mapping to global register variables. There are
3799 plenty of useful comments in this file.</para>
3803 <term><filename>ghc/includes/TailCalls.h</filename></term>
3804 <indexterm><primary><filename>TailCalls.h</filename></primary>
3807 <para>Macros that cooperate with the mangler (see <xref
3808 linkend="sec-mangler">) to make proper tail-calls
3813 <term><filename>ghc/rts/Adjustor.c</filename></term>
3814 <indexterm><primary><filename>Adjustor.c</filename></primary>
3818 <literal>foreign import "wrapper"</literal>
3820 <literal>foreign export dynamic</literal>).
3821 Not essential for getting GHC bootstrapped, so this file
3822 can be deferred until later if necessary.</para>
3826 <term><filename>ghc/rts/StgCRun.c</filename></term>
3827 <indexterm><primary><filename>StgCRun.c</filename></primary>
3830 <para>The little assembly layer between the C world and
3831 the Haskell world. See the comments and code for the
3832 other architectures in this file for pointers.</para>
3836 <term><filename>ghc/rts/MBlock.h</filename></term>
3837 <term><filename>ghc/rts/MBlock.c</filename></term>
3838 <indexterm><primary><filename>MBlock.h</filename></primary>
3840 <indexterm><primary><filename>MBlock.c</filename></primary>
3843 <para>These files are really OS-specific rather than
3844 architecture-specific. In <filename>MBlock.h</filename>
3845 is specified the absolute location at which the RTS
3846 should try to allocate memory on your platform (try to
3847 find an area which doesn't conflict with code or dynamic
3848 libraries). In <filename>Mblock.c</filename> you might
3849 need to tweak the call to <literal>mmap()</literal> for
3856 <sect3 id="sec-mangler">
3857 <title>The mangler</title>
3859 <para>The mangler is an evil Perl-script that rearranges the
3860 assembly code output from gcc to do two main things:</para>
3864 <para>Remove function prologues and epilogues, and all
3865 movement of the C stack pointer. This is to support
3866 tail-calls: every code block in Haskell code ends in an
3867 explicit jump, so we don't want the C-stack overflowing
3868 while we're jumping around between code blocks.</para>
3871 <para>Move the <firstterm>info table</firstterm> for a
3872 closure next to the entry code for that closure. In
3873 unregisterised code, info tables contain a pointer to the
3874 entry code, but in registerised compilation we arrange
3875 that the info table is shoved right up against the entry
3876 code, and addressed backwards from the entry code pointer
3877 (this saves a word in the info table and an extra
3878 indirection when jumping to the closure entry
3883 <para>The mangler is abstracted to a certain extent over some
3884 architecture-specific things such as the particular assembler
3885 directives used to herald symbols. Take a look at the
3886 definitions for other architectures and use these as a
3887 starting point.</para>
3891 <title>The native code generator</title>
3893 <para>The native code generator isn't essential to getting a
3894 registerised build going, but it's a desirable thing to have
3895 because it can cut compilation times in half. The native code
3896 generator is described in some detail in the <ulink
3897 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3898 commentary</ulink>.</para>
3904 <para>To support GHCi, you need to port the dynamic linker
3905 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
3906 currently supports the ELF and PEi386 object file formats - if
3907 your platform uses one of these then you probably don't have
3908 to do anything except fiddle with the
3909 <literal>#ifdef</literal>s at the top of
3910 <filename>Linker.c</filename> to tell it about your OS.</para>
3912 <para>If your system uses a different object file format, then
3913 you have to write a linker — good luck!</para>
3919 <sect1 id="sec-build-pitfalls">
3920 <title>Known pitfalls in building Glasgow Haskell
3922 <indexterm><primary>problems, building</primary></indexterm>
3923 <indexterm><primary>pitfalls, in building</primary></indexterm>
3924 <indexterm><primary>building pitfalls</primary></indexterm></title>
3927 WARNINGS about pitfalls and known “problems”:
3936 One difficulty that comes up from time to time is running out of space
3937 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
3938 compensate for the vagaries of different sysadmin approaches to temp
3940 <indexterm><primary>tmp, running out of space in</primary></indexterm>
3942 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
3943 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
3946 The best way around it is to say
3949 export TMPDIR=<dir>
3952 in your <filename>build.mk</filename> file.
3953 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
3962 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
3963 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
3971 When compiling via C, you'll sometimes get “warning: assignment from
3972 incompatible pointer type” out of GCC. Harmless.
3979 Similarly, <command>ar</command>chiving warning messages like the following are not
3983 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
3984 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
3994 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
3995 get an “Out of heap space” error message. These can vary with the
3996 vagaries of different systems, it seems. The solution is simple:
4003 If you're compiling with GHC 4.00 or later, then the
4004 <emphasis>maximum</emphasis> heap size must have been reached. This
4005 is somewhat unlikely, since the maximum is set to 64M by default.
4006 Anyway, you can raise it with the
4007 <option>-optCrts-M<size></option> flag (add this flag to
4008 <constant><module>_HC_OPTS</constant>
4009 <command>make</command> variable in the appropriate
4010 <filename>Makefile</filename>).
4017 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
4026 and try again: <command>gmake</command>. (see <Xref LinkEnd="sec-suffix"> for information about
4027 <constant><module>_HC_OPTS</constant>.)
4029 Alternatively, just cut to the chase:
4033 % make EXTRA_HC_OPTS=-optCrts-M128M
4042 If you try to compile some Haskell, and you get errors from GCC about
4043 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4044 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4046 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4047 this bug also suggests that you have an old GCC.
4055 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4059 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4060 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4062 ? # or, on some machines: ar s $i
4067 We'd be interested to know if this is still necessary.
4075 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4076 a bit from one Unix to another. One particular gotcha is macro calls
4081 SLIT("Hello, world")
4085 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4086 arguments, so you get
4090 :731: macro `SLIT' used with too many (2) args
4094 Alas, <command>cpp</command> doesn't tell you the offending file!
4096 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4107 <Sect1 id="winbuild"><Title>Notes for building under Windows</Title>
4110 This section summarises how to get the utilities you need on your
4111 Win95/98/NT/2000 machine to use CVS and build GHC. Similar notes for
4112 installing and running GHC may be found in the user guide. In general,
4113 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4114 You should read the GHC installation guide sections on Windows (in the user
4115 guide) before continuing to read these notes.
4119 <sect2 id="cygwin-and-mingw"><Title>Cygwin and MinGW</Title>
4121 <para> The Windows situation for building GHC is rather confusing. This section
4122 tries to clarify, and to establish terminology.</para>
4124 <sect3 id="ghc-mingw"><title>GHC-mingw</title>
4126 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4127 is a collection of header
4128 files and import libraries that allow one to use <command>gcc</command> and produce
4129 native Win32 programs that do not rely on any third-party DLLs. The
4130 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4131 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4134 <para>The GHC that we distribute includes, inside the distribution itself, the MinGW <command>gcc</command>,
4135 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4136 GHC compiles Haskell to C (or to
4137 assembly code), and then invokes these MinGW tools to generate an executable binary.
4138 The resulting binaries can run on any Win32 system.
4140 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4142 <para> The down-side of GHC-mingw is that the MinGW libraries do not support anything like the full
4143 Posix interface. So programs compiled with GHC-mingw cannot import the (Haskell) Posix
4144 library; they have to do
4145 their input output using standard Haskell I/O libraries, or native Win32 bindings.
4149 <sect3 id="ghc-cygwin"><title>GHC-cygwin</title>
4151 <para>There <emphasis>is</emphasis> a way to get the full Posix interface, which is to use Cygwin.
4152 <ulink url="http://www.cygwin.com">Cygwin</ulink> is a complete Unix simulation that runs on Win32.
4153 Cygwin comes with a shell, and all the usual Unix commands: <command>mv</command>, <command>rm</command>,
4154 <command>ls</command>, plus of course <command>gcc</command>, <command>ld</command> and so on.
4155 A C program compiled with the Cygwin <command>gcc</command> certainly can use all of Posix.
4157 <para>So why doesn't GHC use the Cygwin <command>gcc</command> and libraries? Because
4158 Cygwin comes with a DLL <emphasis>that must be linked with every runnable Cygwin-compiled program</emphasis>.
4159 A program compiled by the Cygwin tools cannot run at all unless Cygwin is installed.
4160 If GHC targeted Cygwin, users would have to install Cygwin just to run the Haskell programs
4161 that GHC compiled; and the Cygwin DLL would have to be in the DLL load path.
4162 Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4163 does not change, but the implementation certainly does. Even the interfaces to functions
4164 it exports seem to change occasionally. So programs compiled by GHC might only run with
4165 particular versions of Cygwin. All of this seems very undesirable.
4168 Nevertheless, it is certainly possible to build a version of GHC that targets Cygwin;
4169 we will call that <emphasis>GHC-cygwin</emphasis>. The up-side of GHC-cygwin is
4170 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4174 <sect3><title>HOST_OS vs TARGET_OS</title>
4177 In the source code you'll find various ifdefs looking like:
4179 #ifdef mingw32_HOST_OS
4185 #ifdef mingw32_TARGET_OS
4189 These macros are set by the configure script (via the file config.h).
4190 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4193 The "host" system is the one on which GHC itself will be run.
4196 The "target" system is the one for which the program compiled by GHC will be run.
4199 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4200 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4205 <sect3><title>Summary</title>
4207 <para>Notice that "GHC-mingw" means "GHC that <emphasis>targets</emphasis> MinGW". It says nothing about
4208 how that GHC was <emphasis>built</emphasis>. It is entirely possible to have a GHC-mingw that was built
4209 by compiling GHC's Haskell sources with a GHC-cygwin, or vice versa.</para>
4211 <para>We distribute only a GHC-mingw built by a GHC-mingw; supporting
4212 GHC-cygwin too is beyond our resources. The GHC we distribute
4213 therefore does not require Cygwin to run, nor do the programs it
4214 compiles require Cygwin.</para>
4216 <para>The instructions that follow describe how to build GHC-mingw. It is
4217 possible to build GHC-cygwin, but it's not a supported route, and the build system might
4220 <para>In your build tree, you build a compiler called <Command>ghc-inplace</Command>. It
4221 uses the <Command>gcc</Command> that you specify using the
4222 <option>--with-gcc</option> flag when you run
4223 <Command>configure</Command> (see below).
4224 The makefiles are careful to use <Command>ghc-inplace</Command> (not <Command>gcc</Command>)
4225 to compile any C files, so that it will in turn invoke the right <Command>gcc</Command> rather that
4226 whatever one happens to be in your path. However, the makefiles do use whatever <Command>ld</Command>
4227 and <Command>ar</Command> happen to be in your path. This is a bit naughty, but (a) they are only
4228 used to glom together .o files into a bigger .o file, or a .a file,
4229 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4230 Cygwin and Mingw use the same .o file format. So its ok.
4235 <Sect2><Title>Installing and configuring Cygwin</Title>
4237 <para>You don't need Cygwin to <emphasis>use</emphasis> GHC,
4238 but you do need it to <emphasis>build</emphasis> GHC.</para>
4240 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4241 The installation process is straightforward; we install it in <Filename>c:/cygwin</Filename>.
4242 During the installation dialogue, make sure that you select:
4243 <command>cvs</command>, <command>openssh</command>,
4244 <command>autoconf</command>,
4245 <command>binutils</command> (includes ld and (I think) ar),
4246 <command>gcc</command>,
4247 <command>flex</command>,
4248 <command>make</command>.
4251 <para> Now set the following user environment variables:
4254 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4255 <constant>PATH</constant></para></listitem>
4259 Set <constant>MAKE_MODE</constant> to <Literal>UNIX</Literal>. If you
4260 don't do this you get very weird messages when you type
4261 <Command>make</Command>, such as:
4263 /c: /c: No such file or directory
4268 <listitem><para> Set <constant>SHELL</constant> to
4269 <Filename>c:/cygwin/bin/sh</Filename>. When you invoke a shell in Emacs, this
4270 <constant>SHELL</constant> is what you get.
4273 <listitem><para> Set <constant>HOME</constant> to point to your
4274 home directory. This is where, for example,
4275 <command>bash</command> will look for your <filename>.bashrc</filename>
4276 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4282 There are a few other things to do:
4286 By default, cygwin provides the command shell <filename>ash</filename>
4287 as <filename>sh.exe</filename>. We have often seen build-system problems that
4288 turn out to be due to bugs in <filename>ash</filename>
4290 and length of command lines). On the other hand <filename>bash</filename> seems
4292 So, in <filename>cygwin/bin</filename>
4293 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4294 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4295 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4296 you can't rename a running program!
4302 Some script files used in the make system start with "<Command>#!/bin/perl</Command>",
4303 (and similarly for <Command>sh</Command>). Notice the hardwired path!
4304 So you need to ensure that your <Filename>/bin</Filename> directory has the following
4307 <listitem> <para><Command>sh</Command></para></listitem>
4308 <listitem> <para><Command>perl</Command></para></listitem>
4309 <listitem> <para><Command>cat</Command></para></listitem>
4311 All these come in Cygwin's <Filename>bin</Filename> directory, which you probably have
4312 installed as <Filename>c:/cygwin/bin</Filename>. By default Cygwin mounts "<Filename>/</Filename>" as
4313 <Filename>c:/cygwin</Filename>, so if you just take the defaults it'll all work ok.
4314 (You can discover where your Cygwin
4315 root directory <Filename>/</Filename> is by typing <Command>mount</Command>.)
4316 Provided <Filename>/bin</Filename> points to the Cygwin <Filename>bin</Filename>
4317 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4318 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4324 <para>Finally, here are some things to be aware of when using Cygwin:
4326 <listitem> <para>Cygwin doesn't deal well with filenames that include
4327 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4331 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4332 magical text in it. So other programs that don't use Cygwin's
4333 I/O libraries won't recognise such files as symlinks.
4334 In particular, programs compiled by GHC are meant to be runnable
4335 without having Cygwin, so they don't use the Cygwin library, so
4336 they don't recognise symlinks.
4340 Win32 has a <command>find</command> command which is not the same as Cygwin's find.
4341 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4342 before the Cygwin one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4343 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4344 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4351 <Sect2><Title>Other things you need to install</Title>
4353 <para>You have to install the following other things to build GHC:
4357 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4358 This is what you will use to compile GHC. Add it in your
4359 <constant>PATH</constant>: the installer tells you the path element
4360 you need to add upon completion.
4366 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4367 Happy is a parser generator used to compile the Haskell grammar. Add it in your
4368 <constant>PATH</constant>.
4374 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4375 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw">).
4376 Just pick up a mingw bundle at
4377 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4378 We install it in <filename>c:/mingw</filename>.
4380 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4381 They are only going to get used by explicit access (via the --with-gcc flag you
4382 give to <Command>configure</Command> later). If you do add them to your path
4383 you are likely to get into a mess because their names overlap with Cygwin binaries.
4389 <para>We use <command>emacs</command> a lot, so we install that too.
4390 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4391 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4392 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4393 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4394 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4395 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4401 <para> Finally, check out a copy of GHC sources from
4402 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
4409 <Sect2><Title>Building GHC</Title>
4412 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source">);
4413 the bullets below only tell
4414 you about Windows-specific wrinkles.</para>
4418 Run <Command>autoconf</Command> both in <filename>fptools</filename>
4419 and in <filename>fptools/ghc</filename>. If you omit the latter step you'll
4420 get an error when you run <filename>./configure</filename>:
4423 creating mk/config.h
4424 mk/config.h is unchanged
4426 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4427 ./configure: ./configure: No such file or directory
4428 configure: error: ./configure failed for ghc
4433 <listitem> <para><command>autoconf</command> seems to create the file <filename>configure</filename>
4434 read-only. So if you need to run autoconf again (which I sometimes do for safety's sake),
4437 /usr/bin/autoconf: cannot create configure: permission denied
4439 Solution: delete <filename>configure</filename> first.
4444 You either need to add <filename>ghc</filename> to your
4445 <constant>PATH</constant> before you invoke
4446 <Command>configure</Command>, or use the <Command>configure</Command>
4447 option <option>--with-ghc=c:/ghc/ghc-some-version/bin/ghc</option>.
4452 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4453 This file occasionally remembers out-of-date configuration information, which
4454 can be really confusing.
4460 After <command>autoconf</command> run <command>./configure</command> in
4461 <filename>fptools/</filename> thus:
4464 ./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc
4466 This is the point at which you specify that you are building GHC-mingw
4467 (see <xref linkend="ghc-mingw">). </para>
4469 <para> Both these options are important! It's possible to get into
4470 trouble using the wrong C compiler!</para>
4472 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4473 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4474 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4475 understand a Cygwin path. For example, if you
4476 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4477 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4478 time it tries to invoke it. Worse, the failure comes with
4479 no error message whatsoever. GHC simply fails silently when first invoked,
4480 typically leaving you with this:
4482 make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4483 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4484 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4485 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4486 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4487 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4488 -optc-fomit-frame-pointer -O2 -static
4489 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4490 make[2]: *** [Adjustor.o] Error 1
4491 make[1]: *** [all] Error 1
4492 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4493 make: *** [all] Error 1
4499 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin">)
4500 you'll have to do something more like:
4502 ./configure --with-gcc=...the Cygwin gcc...
4507 <listitem><para> You almost certainly want to set
4511 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config">).
4512 This tells the build system not to split each library into a myriad of little object files, one
4513 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4514 it dramatically increases the time taken to build the libraries in the first place.
4518 <listitem><para> Do not attempt to build the documentation.
4519 It needs all kinds of wierd Jade stuff that we haven't worked out for
4520 Win32.</para></listitem>