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 <emphasis>Windows users: see the notes in <xref linkend="configure-ssh"> about <command>ssh</command> wrinkles!</emphasis>
250 <para>Send a message to to the CVS repository
251 administrator (currently Jeff Lewis
252 <email>jeff@galconn.com</email>), containing:</para>
255 <para>Your desired user-name.</para>
258 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
259 <literal>.ssh/identity.pub</literal>).</para>
262 <para>He will set up your account.</para>
266 <para>Set the following environment variables:</para>
270 <constant>$HOME</constant>: points to your home directory. This is where CVS
271 will look for its <filename>.cvsrc</filename> file.
277 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
279 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
280 <literal>ssh</literal> assumes that your CVS client
281 understands how to execute shell script
282 ("#!"s,really), which is what
283 <literal>ssh</literal> is. This may not be the case on
284 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
285 <literal>ssh1</literal>.</para>
289 <para><literal>$CVSROOT</literal> to
290 <literal>:ext:</literal><replaceable>your-username</replaceable>
291 <literal>@cvs.haskell.org:/home/cvs/root</literal>
292 where <replaceable>your-username</replaceable> is your user name on
293 <literal>cvs.haskell.org</literal>.
295 <para>The <literal>CVSROOT</literal> environment variable will
296 be recorded in the checked-out tree, so you don't need to set
297 this every time. </para>
303 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
304 if you want to use an Emacs buffer for typing in those long commit messages.
310 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
311 set this to point to <filename>bash.exe</filename>.
322 Put the following in <filename>$HOME/.cvsrc</filename>:
333 These are the default options for the specified CVS commands,
334 and represent better defaults than the usual ones. (Feel
335 free to change them.)
339 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
340 the 8.3 DOS filesystem, but that restriction should have
341 been lifted by now (i.e., you're using VFAT or later filesystems.) If
342 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
350 <para>[Experts.] Once your account is set up, you can get
351 access from other machines without bothering Jeff, thus:</para>
354 <para>Generate a public/private key pair on the new
358 <para>Use ssh to log in to
359 <literal>cvs.haskell.org</literal>, from your old
363 <para>Add the public key for the new machine to the file
364 <literal>$HOME/ssh/authorized_keys</literal> on
365 <literal>cvs.haskell.org</literal>.
366 (<literal>authorized_keys2</literal>, I think, for Version
370 <para>Make sure that the new version of
371 <literal>authorized_keys</literal> still has 600 file
380 <sect2 id="cvs-first">
381 <title>Checking Out a Source Tree</title>
385 <para>Make sure you set your <literal>CVSROOT</literal>
386 environment variable according to either of the remote
387 methods above. The Approved Way to check out a source tree
388 is as follows:</para>
391 $ cvs checkout fpconfig
394 <para>At this point you have a new directory called
395 <literal>fptools</literal> which contains the basic stuff
396 for the fptools suite, including the configuration files and
397 some other junk. </para>
399 <para>[Windows users.] The following messages appear to be harmless:
401 setsockopt IPTOS_LOWDELAY: Invalid argument
402 setsockopt IPTOS_THROUGHPUT: Invalid argument
407 <para>You can call the fptools directory whatever you like,
408 CVS won't mind: </para>
411 $ mv fptools <replaceable>directory</replaceable>
414 <para> NB: after you've read the CVS manual you might be
415 tempted to try</para>
417 $ cvs checkout -d <replaceable>directory</replaceable> fpconfig
420 <para>instead of checking out <literal>fpconfig</literal>
421 and then renaming it. But this doesn't work, and will
422 result in checking out the entire repository instead of just
423 the <literal>fpconfig</literal> bit.</para>
425 $ cd <replaceable>directory</replaceable>
426 $ cvs checkout ghc hslibs libraries
429 <para>The second command here checks out the relevant
430 modules you want to work on. For a GHC build, for instance,
431 you need at least the <literal>ghc</literal>,
432 <literal>hslibs</literal> and <literal>libraries</literal>
433 modules (for a full list of the projects available, see
434 <xref linkend="projects">).</para>
436 <para>Remember that if you do not have
437 <literal>happy</literal> installed, you need to check it out
443 <sect2 id="cvs-committing">
444 <title>Committing Changes</title>
446 <para>This is only if you have read-write access to the
447 repository. For anoncvs users, CVS will issue a "read-only
448 repository" error if you try to commit changes.</para>
452 <para>Build the software, if necessary. Unless you're just
453 working on documentation, you'll probably want to build the
454 software in order to test any changes you make.</para>
458 <para>Make changes. Preferably small ones first.</para>
462 <para>Test them. You can see exactly what changes you've
463 made by using the <literal>cvs diff</literal> command:</para>
467 <para>lists all the changes (using the
468 <literal>diff</literal> command) in and below the current
469 directory. In emacs, <literal>C-c C-v =</literal> runs
470 <literal>cvs diff</literal> on the current buffer and shows
471 you the results.</para>
475 <para>If you changed something in the
476 <literal>fptools/libraries</literal> subdirectories, also run
477 <literal>make html</literal> to check if the documentation can
478 be generated successfully, too.</para>
482 <para>Before checking in a change, you need to update your
489 <para>This pulls in any changes that other people have made,
490 and merges them with yours. If there are any conflicts, CVS
491 will tell you, and you'll have to resolve them before you
492 can check your changes in. The documentation describes what
493 to do in the event of a conflict.</para>
495 <para>It's not always necessary to do a full cvs update
496 before checking in a change, since CVS will always tell you
497 if you try to check in a file that someone else has changed.
498 However, you should still update at regular intervals to
499 avoid making changes that don't work in conjuction with
500 changes that someone else made. Keeping an eye on what goes
501 by on the mailing list can help here.</para>
505 <para>When you're happy that your change isn't going to
506 break anything, check it in. For a one-file change:</para>
509 $ cvs commit <replaceable>filename</replaceable>
512 <para>CVS will then pop up an editor for you to enter a
513 "commit message", this is just a short description
514 of what your change does, and will be kept in the history of
517 <para>If you're using emacs, simply load up the file into a
518 buffer and type <literal>C-x C-q</literal>, and emacs will
519 prompt for a commit message and then check in the file for
522 <para>For a multiple-file change, things are a bit
523 trickier. There are several ways to do this, but this is the
524 way I find easiest. First type the commit message into a
525 temporary file. Then either</para>
528 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
531 <para>or, if nothing else has changed in this part of the
535 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
538 <para>where <replaceable>directory</replaceable> is a common
539 parent directory for all your changes, and
540 <replaceable>commit-message</replaceable> is the name of the
541 file containing the commit message.</para>
543 <para>Shortly afterwards, you'll get some mail from the
544 relevant mailing list saying which files changed, and giving
545 the commit message. For a multiple-file change, you should
546 still get only <emphasis>one</emphasis> message.</para>
551 <sect2 id="cvs-update">
552 <title>Updating Your Source Tree</title>
554 <para>It can be tempting to cvs update just part of a source
555 tree to bring in some changes that someone else has made, or
556 before committing your own changes. This is NOT RECOMMENDED!
557 Quite often changes in one part of the tree are dependent on
558 changes in another part of the tree (the
559 <literal>mk/*.mk</literal> files are a good example where
560 problems crop up quite often). Having an inconsistent tree is a
561 major cause of headaches. </para>
563 <para>So, to avoid a lot of hassle, follow this recipe for
564 updating your tree:</para>
568 $ cvs update -P 2>&1 | tee log</screen>
570 <para>Look at the log file, and fix any conflicts (denoted by a
571 <quote>C</quote> in the first column). New directories may have
572 appeared in the repository; CVS doesn't check these out by
573 default, so to get new directories you have to explicitly do
575 $ cvs update -d</screen>
576 in each project subdirectory. Don't do this at the top level,
577 because then <emphasis>all</emphasis> the projects will be
580 <para>If you're using multiple build trees, then for every build
581 tree you have pointing at this source tree, you need to update
582 the links in case any new files have appeared: </para>
585 $ cd <replaceable>build-tree</replaceable>
586 $ lndir <replaceable>source-tree</replaceable>
589 <para>Some files might have been removed, so you need to remove
590 the links pointing to these non-existent files:</para>
593 $ find . -xtype l -exec rm '{}' \;
596 <para>To be <emphasis>really</emphasis> safe, you should do
599 <screen>$ gmake all</screen>
601 <para>from the top-level, to update the dependencies and build
602 any changed files. </para>
605 <sect2 id="cvs-tags">
606 <title>GHC Tag Policy</title>
608 <para>If you want to check out a particular version of GHC,
609 you'll need to know how we tag versions in the repository. The
610 policy (as of 4.04) is:</para>
614 <para>The tree is branched before every major release. The
615 branch tag is <literal>ghc-x-xx-branch</literal>, where
616 <literal>x-xx</literal> is the version number of the release
617 with the <literal>'.'</literal> replaced by a
618 <literal>'-'</literal>. For example, the 4.04 release lives
619 on <literal>ghc-4-04-branch</literal>.</para>
623 <para>The release itself is tagged with
624 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
625 called <literal>ghc-4-06</literal>.</para>
629 <para>We didn't always follow these guidelines, so to see
630 what tags there are for previous versions, do <literal>cvs
631 log</literal> on a file that's been around for a while (like
632 <literal>fptools/ghc/README</literal>).</para>
636 <para>So, to check out a fresh GHC 4.06 tree you would
640 $ cvs co -r ghc-4-06 fpconfig
642 $ cvs co -r ghc-4-06 ghc hslibs
646 <sect2 id="cvs-hints">
647 <title>General Hints</title>
651 <para>As a general rule: commit changes in small units,
652 preferably addressing one issue or implementing a single
653 feature. Provide a descriptive log message so that the
654 repository records exactly which changes were required to
655 implement a given feature/fix a bug. I've found this
656 <emphasis>very</emphasis> useful in the past for finding out
657 when a particular bug was introduced: you can just wind back
658 the CVS tree until the bug disappears.</para>
662 <para>Keep the sources at least *buildable* at any given
663 time. No doubt bugs will creep in, but it's quite easy to
664 ensure that any change made at least leaves the tree in a
665 buildable state. We do nightly builds of GHC to keep an eye
666 on what things work/don't work each day and how we're doing
667 in relation to previous verions. This idea is truely wrecked
668 if the compiler won't build in the first place!</para>
672 <para>To check out extra bits into an already-checked-out
673 tree, use the following procedure. Suppose you have a
674 checked-out fptools tree containing just ghc, and you want
675 to add nofib to it:</para>
686 $ cvs update -d nofib
689 <para>(the -d flag tells update to create a new
690 directory). If you just want part of the nofib suite, you
695 $ cvs checkout nofib/spectral
698 <para>This works because <literal>nofib</literal> is a
699 module in its own right, and spectral is a subdirectory of
700 the nofib module. The path argument to checkout must always
701 start with a module name. There's no equivalent form of this
702 command using <literal>update</literal>.</para>
708 <sect1 id="projects">
709 <title>What projects are there?</title>
711 <para>The <literal>fptools</literal> suite consists of several
712 <firstterm>projects</firstterm>, most of which can be downloaded,
713 built and installed individually. Each project corresponds to a
714 subdirectory in the source tree, and if checking out from CVS then
715 each project can be checked out individually by sitting in the top
716 level of your source tree and typing <command>cvs checkout
717 <replaceable>project</replaceable></command>.</para>
719 <para>Here is a list of the projects currently available:</para>
723 <term><literal>ghc</literal></term>
724 <indexterm><primary><literal>ghc</literal></primary>
725 <secondary>project</secondary></indexterm>
727 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
728 Haskell Compiler</ulink> (minus libraries). Absolutely
729 required for building GHC.</para>
734 <term><literal>glafp-utils</literal></term>
735 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
737 <para>Utility programs, some of which are used by the
738 build/installation system. Required for pretty much
744 <term><literal>green-card</literal></term>
745 <indexterm><primary><literal>green-card</literal></primary><secondary>project</secondary></indexterm>
748 url="http://www.haskell.org/greencard/">Green Card</ulink>
749 system for generating Haskell foreign function
755 <term><literal>haggis</literal></term>
756 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
759 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
760 Haskell GUI framework.</para>
765 <term><literal>haddock</literal></term>
766 <indexterm><primary><literal>haddock</literal></primary><secondary>project</secondary></indexterm>
769 url="http://www.haskell.org/haddock/">Haddock</ulink>
770 documentation tool.</para>
775 <term><literal>happy</literal></term>
776 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
779 url="http://www.haskell.org/happy/">Happy</ulink> Parser
785 <term><literal>hdirect</literal></term>
786 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
789 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
790 Haskell interoperability tool.</para>
795 <term><literal>hood</literal></term>
796 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
798 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
799 Object Observation Debugger</ulink>.</para>
804 <term><literal>hslibs</literal></term>
805 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
807 <para>Supplemental libraries for GHC
808 (<emphasis>required</emphasis> for building GHC).</para>
813 <term><literal>libraries</literal></term>
814 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
816 <para>Hierarchical Haskell library suite
817 (<emphasis>required</emphasis> for building GHC).</para>
822 <term><literal>mhms</literal></term>
823 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
825 <para>The Modular Haskell Metric System.</para>
830 <term><literal>nofib</literal></term>
831 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
833 <para>The NoFib suite: A collection of Haskell programs used
834 primarily for benchmarking.</para>
839 <term><literal>testsuite</literal></term>
840 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
842 <para>A testing framework, including GHC's regression test
848 <para>So, to build GHC you need at least the
849 <literal>ghc</literal>, <literal>libraries</literal> and
850 <literal>hslibs</literal> projects (a GHC source distribution will
851 already include the bits you need).</para>
854 <sect1 id="sec-build-checks">
855 <title>Things to check before you start</title>
857 <para>Here's a list of things to check before you get
863 <indexterm><primary>Disk space needed</primary></indexterm>
864 <para>Disk space needed: from about 100Mb for a basic GHC
865 build, up to probably 500Mb for a GHC build with everything
866 included (libraries built several different ways,
871 <para>Use an appropriate machine / operating system. <xref
872 linkend="sec-port-info"> lists the supported platforms; if
873 yours isn't amongst these then you can try porting GHC (see
874 <xref linkend="sec-porting-ghc">).</para>
878 <para>Be sure that the “pre-supposed” utilities are
879 installed. <Xref LinkEnd="sec-pre-supposed">
884 <para>If you have any problem when building or installing the
885 Glasgow tools, please check the “known pitfalls” (<Xref
886 LinkEnd="sec-build-pitfalls">). Also check the FAQ for the
887 version you're building, which is part of the User's Guide and
888 available on the <ulink URL="http://www.haskell.org/ghc/" >GHC web
891 <indexterm><primary>bugs</primary><secondary>known</secondary></indexterm>
893 <para>If you feel there is still some shortcoming in our
894 procedure or instructions, please report it.</para>
896 <para>For GHC, please see the <ulink
897 url="http://www.haskell.org/ghc/docs/latest/set/bug-reporting.html">bug-reporting
898 section of the GHC Users' Guide</ulink>, to maximise the
899 usefulness of your report.</para>
901 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
902 <para>If in doubt, please send a message to
903 <email>glasgow-haskell-bugs@haskell.org</email>.
904 <indexterm><primary>bugs</primary><secondary>mailing
905 list</secondary></indexterm></para>
910 <sect1 id="sec-port-info">
911 <title>What machines the Glasgow tools run on</title>
913 <indexterm><primary>ports</primary><secondary>GHC</secondary></indexterm>
914 <indexterm><primary>GHC</primary><secondary>ports</secondary></indexterm>
915 <indexterm><primary>platforms</primary><secondary>supported</secondary></indexterm>
917 <para>The main question is whether or not the Haskell compiler
918 (GHC) runs on your platform.</para>
920 <para>A “platform” is a
921 architecture/manufacturer/operating-system combination, such as
922 <literal>sparc-sun-solaris2</literal>. Other common ones are
923 <literal>alpha-dec-osf2</literal>,
924 <literal>hppa1.1-hp-hpux9</literal>,
925 <literal>i386-unknown-linux</literal>,
926 <literal>i386-unknown-solaris2</literal>,
927 <literal>i386-unknown-freebsd</literal>,
928 <literal>i386-unknown-cygwin32</literal>,
929 <literal>m68k-sun-sunos4</literal>,
930 <literal>mips-sgi-irix5</literal>,
931 <literal>sparc-sun-sunos4</literal>,
932 <literal>sparc-sun-solaris2</literal>,
933 <literal>powerpc-ibm-aix</literal>.</para>
935 <para>Some libraries may only work on a limited number of
936 platforms; for example, a sockets library is of no use unless the
937 operating system supports the underlying BSDisms.</para>
940 <title>What platforms the Haskell compiler (GHC) runs on</title>
942 <indexterm><primary>fully-supported platforms</primary></indexterm>
943 <indexterm><primary>native-code generator</primary></indexterm>
944 <indexterm><primary>registerised ports</primary></indexterm>
945 <indexterm><primary>unregisterised ports</primary></indexterm>
947 <para>The GHC hierarchy of Porting Goodness: (a) Best is a
948 native-code generator; (b) next best is a
949 “registerised” port; (c) the bare minimum is an
950 “unregisterised” port.
951 (“Unregisterised” is so terrible that we won't say
952 more about it).</para>
954 <para>We use Sparcs running Solaris 2.7 and x86 boxes running
955 FreeBSD and Linux, so those are the best supported platforms,
956 unsurprisingly.</para>
958 <para>Here's everything that's known about GHC ports. We
959 identify platforms by their “canonical”
960 CPU/Manufacturer/OS triple.</para>
964 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:</term>
965 <indexterm><primary>alpha-dec-osf</primary></indexterm>
966 <indexterm><primary>alpha-dec-linux</primary></indexterm>
967 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
968 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
969 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
972 <para>The OSF port is currently working (as of GHC version
973 5.02.1) and well supported. The native code generator is
974 currently non-working. Other operating systems will
975 require some minor porting.</para>
980 <term>sparc-sun-sunos4</term>
981 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
983 <para>Probably works with minor tweaks, hasn't been tested
989 <term>sparc-sun-solaris2</term>
990 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
992 <para>Fully supported (at least for Solaris 2.7),
993 including native-code generator.</para>
998 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
999 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1001 <para>A registerised port is available for version 4.08,
1002 but GHC hasn't been built on that platform since (as far
1003 as we know). No native-code generator.</para>
1008 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1009 <indexterm><primary>i386-*-linux</primary></indexterm>
1011 <para>GHC works registerised and has a native code
1012 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1013 or later. NOTE about <literal>glibc</literal> versions:
1014 GHC binaries built on a system running <literal>glibc
1015 2.0</literal> won't work on a system running
1016 <literal>glibc 2.1</literal>, and vice versa. In general,
1017 don't expect compatibility between
1018 <literal>glibc</literal> versions, even if the shared
1019 library version hasn't changed.</para>
1024 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2 or
1026 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1028 <para>GHC works registerised. Pre-built packages are
1029 available in the native package format, so if you just
1030 need binaries you're better off just installing the
1031 package (it might even be on your installation
1037 <term>i386-unknown-openbsd (PCs running OpenBSD)</term>
1038 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1040 <para>Supported, with native code generator. Packages are
1041 available through the ports system in the native package
1047 <term>i386-unknown-netbsd (PCs running NetBSD and
1049 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1051 <para>Will require some minor porting effort, but should
1052 work registerised.</para>
1057 <term>i386-unknown-mingw32 (PCs running Windows)</term>
1058 <indexterm><primary>i386-unknown-mingw32</primary></indexterm>
1060 <para>Fully supported under Win9x, WinNT, Win2k, and
1061 WinXP. Includes a native code generator. Building from
1062 source requires a recent <ulink
1063 url="http://www.cygwin.com/">Cygwin</ulink> distribution
1064 to be installed.</para>
1069 <term>ia64-unknown-linux</term>
1070 <indexterm><primary>ia64-unknown-linux</primary></indexterm>
1072 <para>GHC currently works unregisterised. A registerised
1073 port is in progress.</para>
1078 <term>mips-sgi-irix5</term>
1079 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1081 <para>Port has worked in the past, but hasn't been tested
1082 for some time (and will certainly have rotted in various
1083 ways). As usual, we don't have access to machines and
1084 there hasn't been an overwhelming demand for this port,
1085 but feel free to get in touch.</para>
1090 <term>powerpc-ibm-aix</term>
1091 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1093 <para>Port currently doesn't work, needs some minimal
1094 porting effort. As usual, we don't have access to
1095 machines and there hasn't been an overwhelming demand for
1096 this port, but feel free to get in touch.</para>
1101 <term>powerpc-apple-darwin</term>
1102 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1104 <para>Supported registerised. No native code
1110 <term>powerpc-apple-linux</term>
1111 <indexterm><primary>powerpc-apple-linux</primary></indexterm>
1113 <para>Not supported (yet).</para>
1118 <para>Various other systems have had GHC ported to them in the
1119 distant past, including various Motorola 68k boxes. The 68k
1120 support still remains, but porting to one of these systems will
1121 certainly be a non-trivial task.</para>
1125 <title>What machines the other tools run on</title>
1127 <para>Unless you hear otherwise, the other tools work if GHC
1133 <sect1 id="sec-pre-supposed">
1134 <title>Installing pre-supposed utilities</title>
1136 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1137 <indexterm><primary>utilities, pre-supposed</primary></indexterm>
1139 <para>Here are the gory details about some utility programs you
1140 may need; <command>perl</command>, <command>gcc</command> and
1141 <command>happy</command> are the only important
1142 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is
1143 important if you're going for Parallel Haskell.) The
1144 <command>configure</command><indexterm><primary>configure</primary></indexterm>
1145 script will tell you if you are missing something.</para>
1151 <indexterm><primary>pre-supposed: GHC</primary></indexterm>
1152 <indexterm><primary>GHC, pre-supposed</primary></indexterm>
1154 <para>GHC is required to build many of the tools, including
1155 GHC itself. If you need to port GHC to your platform
1156 because there isn't a binary distribution of GHC available,
1157 then see <xref linkend="sec-porting-ghc">.</para>
1159 <para>Which version of GHC you need will depend on the
1160 packages you intend to build. GHC itself will normally
1161 build using one of several older versions of itself - check
1162 the announcement or release notes for details.</para>
1168 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1169 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1171 <para><emphasis>You have to have Perl to proceed!</emphasis>
1172 Perl version 5 at least is required. GHC has been known to
1173 tickle bugs in Perl, so if you find that Perl crashes when
1174 running GHC try updating (or downgrading) your Perl
1175 installation. Versions of Perl that we use and are known to
1176 be fairly stable are 5.005 and 5.6.1.</para>
1178 <para>For Win32 platforms, you should use the binary
1179 supplied in the InstallShield (copy it to
1180 <filename>/bin</filename>). The Cygwin-supplied Perl seems
1183 <para>Perl should be put somewhere so that it can be invoked
1184 by the <literal>#!</literal> script-invoking
1185 mechanism. The full pathname may need to be less than 32
1186 characters long on some systems.</para>
1191 <term>GNU C (<command>gcc</command>)</term>
1192 <indexterm><primary>pre-supposed: GCC (GNU C
1193 compiler)</primary></indexterm> <indexterm><primary>GCC (GNU C
1194 compiler), pre-supposed</primary></indexterm>
1196 <para>We recommend using GCC version 2.95.2 on all
1197 platforms. Failing that, version 2.7.2 is stable on most
1198 platforms. Earlier versions of GCC can be assumed not to
1199 work, and versions in between 2.7.2 and 2.95.2 (including
1200 <command>egcs</command>) have varying degrees of stability
1201 depending on the platform.</para>
1203 <para>GCC 3.2 is currently known to have problems building
1204 GHC on Sparc, but is stable on x86.</para>
1206 <para>GCC 3.3 currently cannot be used to build GHC, due to
1207 some problems with the new C preprocessor.</para>
1209 <para>If your GCC dies with “internal error” on
1210 some GHC source file, please let us know, so we can report
1211 it and get things improved. (Exception: on iX86
1212 boxes—you may need to fiddle with GHC's
1213 <option>-monly-N-regs</option> option; see the User's
1219 <term>GNU Make</term>
1220 <indexterm><primary>make</primary><secondary>GNU</secondary>
1223 <para>The fptools build system makes heavy use of features
1224 specific to GNU <command>make</command>, so you must have
1225 this installed in order to build any of the fptools
1232 <indexterm><primary>Happy</primary></indexterm>
1234 <para>Happy is a parser generator tool for Haskell, and is
1235 used to generate GHC's parsers. Happy is written in
1236 Haskell, and is a project in the CVS repository
1237 (<literal>fptools/happy</literal>). It can be built from
1238 source, but bear in mind that you'll need GHC installed in
1239 order to build it. To avoid the chicken/egg problem,
1240 install a binary distribution of either Happy or GHC to get
1241 started. Happy distributions are available from <ulink
1242 url="http://www.haskell.org/happy/">Happy's Web
1243 Page</ulink>.</para>
1248 <term>Autoconf</term>
1249 <indexterm><primary>pre-supposed: Autoconf</primary></indexterm>
1250 <indexterm><primary>Autoconf, pre-supposed</primary></indexterm>
1252 <para>GNU Autoconf is needed if you intend to build from the
1253 CVS sources, it is <emphasis>not</emphasis> needed if you
1254 just intend to build a standard source distribution.</para>
1256 <para>Version 2.52 or later of autoconf is required.
1257 NB. vesrion 2.13 will no longer work, as of GHC version
1260 <para>Autoconf builds the <command>configure</command>
1261 script from <filename>configure.in</filename> and
1262 <filename>aclocal.m4</filename>. If you modify either of
1263 these files, you'll need <command>autoconf</command> to
1264 rebuild <filename>configure</filename>.</para>
1269 <term><command>sed</command></term>
1270 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1271 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1273 <para>You need a working <command>sed</command> if you are
1274 going to build from sources. The build-configuration stuff
1275 needs it. GNU sed version 2.0.4 is no good! It has a bug
1276 in it that is tickled by the build-configuration. 2.0.5 is
1277 OK. Others are probably OK too (assuming we don't create too
1278 elaborate configure scripts.)</para>
1283 <para>One <literal>fptools</literal> project is worth a quick note
1284 at this point, because it is useful for all the others:
1285 <literal>glafp-utils</literal> contains several utilities which
1286 aren't particularly Glasgow-ish, but Occasionally Indispensable.
1287 Like <command>lndir</command> for creating symbolic link
1290 <sect2 id="pre-supposed-gph-tools">
1291 <title>Tools for building parallel GHC (GPH)</title>
1295 <term>PVM version 3:</term>
1296 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1297 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1299 <para>PVM is the Parallel Virtual Machine on which
1300 Parallel Haskell programs run. (You only need this if you
1301 plan to run Parallel Haskell. Concurrent Haskell, which
1302 runs concurrent threads on a uniprocessor doesn't need
1303 it.) Underneath PVM, you can have (for example) a network
1304 of workstations (slow) or a multiprocessor box
1307 <para>The current version of PVM is 3.3.11; we use 3.3.7.
1308 It is readily available on the net; I think I got it from
1309 <literal>research.att.com</literal>, in
1310 <filename>netlib</filename>.</para>
1312 <para>A PVM installation is slightly quirky, but easy to
1313 do. Just follow the <filename>Readme</filename>
1314 instructions.</para>
1319 <term><command>bash</command>:</term>
1320 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1322 <para>Sadly, the <command>gr2ps</command> script, used to
1323 convert “parallelism profiles” to PostScript,
1324 is written in Bash (GNU's Bourne Again shell). This bug
1325 will be fixed (someday).</para>
1331 <sect2 id="pre-supposed-other-tools">
1332 <title>Other useful tools</title>
1337 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1338 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1340 <para>This is a quite-a-bit-better-than-Lex lexer. Used
1341 to build a couple of utilities in
1342 <literal>glafp-utils</literal>. Depending on your
1343 operating system, the supplied <command>lex</command> may
1344 or may not work; you should get the GNU version.</para>
1349 <para>More tools are required if you want to format the documentation
1350 that comes with GHC and other fptools projects. See <xref
1351 linkend="building-docs">.</para>
1355 <sect1 id="sec-building-from-source">
1356 <title>Building from source</title>
1358 <indexterm><primary>Building from source</primary></indexterm>
1359 <indexterm><primary>Source, building from</primary></indexterm>
1361 <para>You've been rash enough to want to build some of the Glasgow
1362 Functional Programming tools (GHC, Happy, nofib, etc.) from
1363 source. You've slurped the source, from the CVS repository or
1364 from a source distribution, and now you're sitting looking at a
1365 huge mound of bits, wondering what to do next.</para>
1367 <para>Gingerly, you type <command>make</command>. Wrong
1370 <para>This rest of this guide is intended for duffers like me, who
1371 aren't really interested in Makefiles and systems configurations,
1372 but who need a mental model of the interlocking pieces so that
1373 they can make them work, extend them consistently when adding new
1374 software, and lay hands on them gently when they don't
1377 <sect2 id="quick-start">
1378 <title>Quick Start</title>
1380 <para>If you are starting from a source distribution, and just
1381 want a completely standard build, then the following should
1384 <screen>$ ./configure
1389 <para>For GHC, this will do a 2-stage bootstrap build of the
1390 compiler, with profiling libraries, and install the
1393 <para>If you want to do anything at all non-standard, or you
1394 want to do some development, read on...</para>
1397 <sect2 id="sec-source-tree">
1398 <title>Your source tree</title>
1400 <para>The source code is held in your <emphasis>source
1401 tree</emphasis>. The root directory of your source tree
1402 <emphasis>must</emphasis> contain the following directories and
1407 <para><filename>Makefile</filename>: the root
1412 <para><filename>mk/</filename>: the directory that contains
1413 the main Makefile code, shared by all the
1414 <literal>fptools</literal> software.</para>
1418 <para><filename>configure.in</filename>,
1419 <filename>config.sub</filename>,
1420 <filename>config.guess</filename>: these files support the
1421 configuration process.</para>
1425 <para><filename>install-sh</filename>.</para>
1429 <para>All the other directories are individual
1430 <emphasis>projects</emphasis> of the <literal>fptools</literal>
1431 system—for example, the Glasgow Haskell Compiler
1432 (<literal>ghc</literal>), the Happy parser generator
1433 (<literal>happy</literal>), the <literal>nofib</literal>
1434 benchmark suite, and so on. You can have zero or more of these.
1435 Needless to say, some of them are needed to build others.</para>
1437 <para>The important thing to remember is that even if you want
1438 only one project (<literal>happy</literal>, say), you must have
1439 a source tree whose root directory contains
1440 <filename>Makefile</filename>, <filename>mk/</filename>,
1441 <filename>configure.in</filename>, and the project(s) you want
1442 (<filename>happy/</filename> in this case). You cannot get by
1443 with just the <filename>happy/</filename> directory.</para>
1447 <title>Build trees</title>
1448 <indexterm><primary>build trees</primary></indexterm>
1449 <indexterm><primary>link trees, for building</primary></indexterm>
1451 <para>If you just want to build the software once on a single
1452 platform, then your source tree can also be your build tree, and
1453 you can skip the rest of this section.</para>
1455 <para>We often want to build multiple versions of our software
1456 for different architectures, or with different options
1457 (e.g. profiling). It's very desirable to share a single copy of
1458 the source code among all these builds.</para>
1460 <para>So for every source tree we have zero or more
1461 <emphasis>build trees</emphasis>. Each build tree is initially
1462 an exact copy of the source tree, except that each file is a
1463 symbolic link to the source file, rather than being a copy of
1464 the source file. There are “standard” Unix
1465 utilities that make such copies, so standard that they go by
1467 <command>lndir</command><indexterm><primary>lndir</primary></indexterm>,
1468 <command>mkshadowdir</command><indexterm><primary>mkshadowdir</primary></indexterm>
1469 are two (If you don't have either, the source distribution
1470 includes sources for the X11
1471 <command>lndir</command>—check out
1472 <filename>fptools/glafp-utils/lndir</filename>). See <Xref
1473 LinkEnd="sec-storysofar"> for a typical invocation.</para>
1475 <para>The build tree does not need to be anywhere near the
1476 source tree in the file system. Indeed, one advantage of
1477 separating the build tree from the source is that the build tree
1478 can be placed in a non-backed-up partition, saving your systems
1479 support people from backing up untold megabytes of
1480 easily-regenerated, and rapidly-changing, gubbins. The golden
1481 rule is that (with a single exception—<XRef
1482 LinkEnd="sec-build-config">) <emphasis>absolutely everything in
1483 the build tree is either a symbolic link to the source tree, or
1484 else is mechanically generated</emphasis>. It should be
1485 perfectly OK for your build tree to vanish overnight; an hour or
1486 two compiling and you're on the road again.</para>
1488 <para>You need to be a bit careful, though, that any new files
1489 you create (if you do any development work) are in the source
1490 tree, not a build tree!</para>
1492 <para>Remember, that the source files in the build tree are
1493 <emphasis>symbolic links</emphasis> to the files in the source
1494 tree. (The build tree soon accumulates lots of built files like
1495 <filename>Foo.o</filename>, as well.) You can
1496 <emphasis>delete</emphasis> a source file from the build tree
1497 without affecting the source tree (though it's an odd thing to
1498 do). On the other hand, if you <emphasis>edit</emphasis> a
1499 source file from the build tree, you'll edit the source-tree
1500 file directly. (You can set up Emacs so that if you edit a
1501 source file from the build tree, Emacs will silently create an
1502 edited copy of the source file in the build tree, leaving the
1503 source file unchanged; but the danger is that you think you've
1504 edited the source file whereas actually all you've done is edit
1505 the build-tree copy. More commonly you do want to edit the
1506 source file.)</para>
1508 <para>Like the source tree, the top level of your build tree
1509 must be (a linked copy of) the root directory of the
1510 <literal>fptools</literal> suite. Inside Makefiles, the root of
1511 your build tree is called
1512 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>.
1513 In the rest of this document path names are relative to
1514 <constant>$(FPTOOLS_TOP)</constant> unless
1515 otherwise stated. For example, the file
1516 <filename>ghc/mk/target.mk</filename> is actually
1517 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</filename>.</para>
1520 <sect2 id="sec-build-config">
1521 <title>Getting the build you want</title>
1523 <para>When you build <literal>fptools</literal> you will be
1524 compiling code on a particular <emphasis>host
1525 platform</emphasis>, to run on a particular <emphasis>target
1526 platform</emphasis> (usually the same as the host
1527 platform)<indexterm><primary>platform</primary></indexterm>.
1528 The difficulty is that there are minor differences between
1529 different platforms; minor, but enough that the code needs to be
1530 a bit different for each. There are some big differences too:
1531 for a different architecture we need to build GHC with a
1532 different native-code generator.</para>
1534 <para>There are also knobs you can turn to control how the
1535 <literal>fptools</literal> software is built. For example, you
1536 might want to build GHC optimised (so that it runs fast) or
1537 unoptimised (so that you can compile it fast after you've
1538 modified it. Or, you might want to compile it with debugging on
1539 (so that extra consistency-checking code gets included) or off.
1542 <para>All of this stuff is called the
1543 <emphasis>configuration</emphasis> of your build. You set the
1544 configuration using a three-step process.</para>
1548 <term>Step 1: get ready for configuration.</term>
1550 <para>NOTE: if you're starting from a source distribution,
1551 rather than CVS sources, you can skip this step.</para>
1553 <para>Change directory to
1554 <constant>$(FPTOOLS_TOP)</constant> and
1556 <command>autoconf</command><indexterm><primary>autoconf</primary></indexterm>
1557 (with no arguments). This GNU program converts
1558 <filename><constant>$(FPTOOLS_TOP)</constant>/configure.in</filename>
1559 to a shell script called
1560 <filename><constant>$(FPTOOLS_TOP)</constant>/configure</filename>.
1563 <para>Some projects, including GHC, have their own
1564 configure script. If there's an
1565 <constant>$(FPTOOLS_TOP)/<project>/configure.in</constant>,
1566 then you need to run <command>autoconf</command> in that
1567 directory too.</para>
1569 <para>Both these steps are completely
1570 platform-independent; they just mean that the
1571 human-written file (<filename>configure.in</filename>) can
1572 be short, although the resulting shell script,
1573 <command>configure</command>, and
1574 <filename>mk/config.h.in</filename>, are long.</para>
1579 <term>Step 2: system configuration.</term>
1581 <para>Runs the newly-created <command>configure</command>
1582 script, thus:</para>
1585 ./configure <optional><parameter>args</parameter></optional>
1588 <para><command>configure</command>'s mission is to scurry
1589 round your computer working out what architecture it has,
1590 what operating system, whether it has the
1591 <Function>vfork</Function> system call, where
1592 <command>yacc</command> is kept, whether
1593 <command>gcc</command> is available, where various obscure
1594 <literal>#include</literal> files are, whether it's a
1595 leap year, and what the systems manager had for lunch. It
1596 communicates these snippets of information in two
1603 <filename>mk/config.mk.in</filename><indexterm><primary>config.mk.in</primary></indexterm>
1605 <filename>mk/config.mk</filename><indexterm><primary>config.mk</primary></indexterm>,
1606 substituting for things between
1607 “<literal>@</literal>” brackets. So,
1608 “<literal>@HaveGcc@</literal>” will be
1609 replaced by “<literal>YES</literal>” or
1610 “<literal>NO</literal>” depending on what
1611 <command>configure</command> finds.
1612 <filename>mk/config.mk</filename> is included by every
1613 Makefile (directly or indirectly), so the
1614 configuration information is thereby communicated to
1615 all Makefiles.</para>
1619 <para> It translates
1620 <filename>mk/config.h.in</filename><indexterm><primary>config.h.in</primary></indexterm>
1622 <filename>mk/config.h</filename><indexterm><primary>config.h</primary></indexterm>.
1623 The latter is <literal>#include</literal>d by
1624 various C programs, which can thereby make use of
1625 configuration information.</para>
1629 <para><command>configure</command> takes some optional
1630 arguments. Use <literal>./configure --help</literal> to
1631 get a list of the available arguments. Here are some of
1632 the ones you might need:</para>
1636 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1637 <indexterm><primary><literal>--with-ghc</literal></primary>
1640 <para>Specifies the path to an installed GHC which
1641 you would like to use. This compiler will be used
1642 for compiling GHC-specific code (eg. GHC itself).
1643 This option <emphasis>cannot</emphasis> be specified
1644 using <filename>build.mk</filename> (see later),
1645 because <command>configure</command> needs to
1646 auto-detect the version of GHC you're using. The
1647 default is to look for a compiler named
1648 <literal>ghc</literal> in your path.</para>
1653 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1654 <indexterm><primary><literal>--with-hc</literal></primary>
1657 <para>Specifies the path to any installed Haskell
1658 compiler. This compiler will be used for compiling
1659 generic Haskell code. The default is to use
1660 <literal>ghc</literal>.</para>
1665 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1666 <indexterm><primary><literal>--with-gcc</literal></primary>
1669 <para>Specifies the path to the installed GCC. This
1670 compiler will be used to compile all C files,
1671 <emphasis>except</emphasis> any generated by the
1672 installed Haskell compiler, which will have its own
1673 idea of which C compiler (if any) to use. The
1674 default is to use <literal>gcc</literal>.</para>
1679 <para><command>configure</command> caches the results of
1680 its run in <filename>config.cache</filename>. Quite often
1681 you don't want that; you're running
1682 <command>configure</command> a second time because
1683 something has changed. In that case, simply delete
1684 <filename>config.cache</filename>.</para>
1689 <term>Step 3: build configuration.</term>
1691 <para>Next, you say how this build of
1692 <literal>fptools</literal> is to differ from the standard
1693 defaults by creating a new file
1694 <filename>mk/build.mk</filename><indexterm><primary>build.mk</primary></indexterm>
1695 <emphasis>in the build tree</emphasis>. This file is the
1696 one and only file you edit in the build tree, precisely
1697 because it says how this build differs from the source.
1698 (Just in case your build tree does die, you might want to
1699 keep a private directory of <filename>build.mk</filename>
1700 files, and use a symbolic link in each build tree to point
1701 to the appropriate one.) So
1702 <filename>mk/build.mk</filename> never exists in the
1703 source tree—you create one in each build tree from
1704 the template. We'll discuss what to put in it
1710 <para>And that's it for configuration. Simple, eh?</para>
1712 <para>What do you put in your build-specific configuration file
1713 <filename>mk/build.mk</filename>? <emphasis>For almost all
1714 purposes all you will do is put make variable definitions that
1715 override those in</emphasis>
1716 <filename>mk/config.mk.in</filename>. The whole point of
1717 <filename>mk/config.mk.in</filename>—and its derived
1718 counterpart <filename>mk/config.mk</filename>—is to define
1719 the build configuration. It is heavily commented, as you will
1720 see if you look at it. So generally, what you do is look at
1721 <filename>mk/config.mk.in</filename>, and add definitions in
1722 <filename>mk/build.mk</filename> that override any of the
1723 <filename>config.mk</filename> definitions that you want to
1724 change. (The override occurs because the main boilerplate file,
1725 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1726 includes <filename>build.mk</filename> after
1727 <filename>config.mk</filename>.)</para>
1729 <para>For your convenience, there's a file called <filename>build.mk.sample</filename>
1730 that can serve as a starting point for your <filename>build.mk</filename>.</para>
1732 <para>For example, <filename>config.mk.in</filename> contains
1733 the definition:</para>
1736 GhcHcOpts=-O -Rghc-timing
1739 <para>The accompanying comment explains that this is the list of
1740 flags passed to GHC when building GHC itself. For doing
1741 development, it is wise to add <literal>-DDEBUG</literal>, to
1742 enable debugging code. So you would add the following to
1743 <filename>build.mk</filename>:</para>
1745 <para>or, if you prefer,</para>
1748 GhcHcOpts += -DDEBUG
1751 <para>GNU <command>make</command> allows existing definitions to
1752 have new text appended using the “<literal>+=</literal>”
1753 operator, which is quite a convenient feature.)</para>
1755 <para>If you want to remove the <literal>-O</literal> as well (a
1756 good idea when developing, because the turn-around cycle gets a
1757 lot quicker), you can just override
1758 <literal>GhcLibHcOpts</literal> altogether:</para>
1761 GhcHcOpts=-DDEBUG -Rghc-timing
1764 <para>When reading <filename>config.mk.in</filename>, remember
1765 that anything between “@...@” signs is going to be substituted
1766 by <command>configure</command> later. You
1767 <emphasis>can</emphasis> override the resulting definition if
1768 you want, but you need to be a bit surer what you are doing.
1769 For example, there's a line that says:</para>
1775 <para>This defines the Make variables <constant>YACC</constant>
1776 to the pathname for a <command>yacc</command> that
1777 <command>configure</command> finds somewhere. If you have your
1778 own pet <command>yacc</command> you want to use instead, that's
1779 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1785 <para>You do not <emphasis>have</emphasis> to have a
1786 <filename>mk/build.mk</filename> file at all; if you don't,
1787 you'll get all the default settings from
1788 <filename>mk/config.mk.in</filename>.</para>
1790 <para>You can also use <filename>build.mk</filename> to override
1791 anything that <command>configure</command> got wrong. One place
1792 where this happens often is with the definition of
1793 <constant>FPTOOLS_TOP_ABS</constant>: this
1794 variable is supposed to be the canonical path to the top of your
1795 source tree, but if your system uses an automounter then the
1796 correct directory is hard to find automatically. If you find
1797 that <command>configure</command> has got it wrong, just put the
1798 correct definition in <filename>build.mk</filename>.</para>
1802 <sect2 id="sec-storysofar">
1803 <title>The story so far</title>
1805 <para>Let's summarise the steps you need to carry to get
1806 yourself a fully-configured build tree from scratch.</para>
1810 <para> Get your source tree from somewhere (CVS repository
1811 or source distribution). Say you call the root directory
1812 <filename>myfptools</filename> (it does not have to be
1813 called <filename>fptools</filename>). Make sure that you
1814 have the essential files (see <XRef
1815 LinkEnd="sec-source-tree">).</para>
1820 <para>(Optional) Use <command>lndir</command> or
1821 <command>mkshadowdir</command> to create a build tree.</para>
1825 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1828 <para>(N.B. <command>mkshadowdir</command>'s first argument
1829 is taken relative to its second.) You probably want to give
1830 the build tree a name that suggests its main defining
1831 characteristic (in your mind at least), in case you later
1836 <para>Change directory to the build tree. Everything is
1837 going to happen there now.</para>
1840 $ cd /scratch/joe-bloggs/myfptools-sun4
1846 <para>Prepare for system configuration:</para>
1852 <para>(You can skip this step if you are starting from a
1853 source distribution, and you already have
1854 <filename>configure</filename> and
1855 <filename>mk/config.h.in</filename>.)</para>
1857 <para>Some projects, including GHC itself, have their own
1858 configure scripts, so it is necessary to run autoconf again
1859 in the appropriate subdirectories. eg:</para>
1862 $ (cd ghc; autoconf)
1867 <para>Do system configuration:</para>
1873 <para>Don't forget to check whether you need to add any
1874 arguments to <literal>configure</literal>; for example, a
1875 common requirement is to specify which GHC to use with
1876 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1880 <para>Create the file <filename>mk/build.mk</filename>,
1881 adding definitions for your desired configuration
1890 <para>You can make subsequent changes to
1891 <filename>mk/build.mk</filename> as often as you like. You do
1892 not have to run any further configuration programs to make these
1893 changes take effect. In theory you should, however, say
1894 <command>gmake clean</command>, <command>gmake all</command>,
1895 because configuration option changes could affect
1896 anything—but in practice you are likely to know what's
1901 <title>Making things</title>
1903 <para>At this point you have made yourself a fully-configured
1904 build tree, so you are ready to start building real
1907 <para>The first thing you need to know is that <emphasis>you
1908 must use GNU <command>make</command>, usually called
1909 <command>gmake</command>, not standard Unix
1910 <command>make</command></emphasis>. If you use standard Unix
1911 <command>make</command> you will get all sorts of error messages
1912 (but no damage) because the <literal>fptools</literal>
1913 <command>Makefiles</command> use GNU <command>make</command>'s
1914 facilities extensively.</para>
1916 <para>To just build the whole thing, <command>cd</command> to
1917 the top of your <literal>fptools</literal> tree and type
1918 <command>gmake</command>. This will prepare the tree and build
1919 the various projects in the correct order.</para>
1922 <sect2 id="sec-bootstrapping">
1923 <title>Bootstrapping GHC</title>
1925 <para>GHC requires a 2-stage bootstrap in order to provide
1926 full functionality, including GHCi. By a 2-stage bootstrap, we
1927 mean that the compiler is built once using the installed GHC,
1928 and then again using the compiler built in the first stage. You
1929 can also build a stage 3 compiler, but this normally isn't
1930 necessary except to verify that the stage 2 compiler is working
1933 <para>Note that when doing a bootstrap, the stage 1 compiler
1934 must be built, followed by the runtime system and libraries, and
1935 then the stage 2 compiler. The correct ordering is implemented
1936 by the top-level fptools <filename>Makefile</filename>, so if
1937 you want everything to work automatically it's best to start
1938 <command>make</command> from the top of the tree. When building
1939 GHC, the top-level fptools <filename>Makefile</filename> is set
1940 up to do a 2-stage bootstrap by default (when you say
1941 <command>make</command>). Some other targets it supports
1948 <para>Build everything as normal, including the stage 1
1956 <para>Build the stage 2 compiler only.</para>
1963 <para>Build the stage 3 compiler only.</para>
1968 <term>bootstrap</term> <term>bootstrap2</term>
1970 <para>Build stage 1 followed by stage 2.</para>
1975 <term>bootstrap3</term>
1977 <para>Build stages 1, 2 and 3.</para>
1982 <term>install</term>
1984 <para>Install everything, including the compiler built in
1985 stage 2. To override the stage, say <literal>make install
1986 stage=<replaceable>n</replaceable></literal> where
1987 <replaceable>n</replaceable> is the stage to install.</para>
1992 <para>The top-level <filename>Makefile</filename> also arranges
1993 to do the appropriate <literal>make boot</literal> steps (see
1994 below) before actually building anything.</para>
1996 <para>The <literal>stage1</literal>, <literal>stage2</literal>
1997 and <literal>stage3</literal> targets also work in the
1998 <literal>ghc/compiler</literal> directory, but don't forget that
1999 each stage requires its own <literal>make boot</literal> step:
2000 for example, you must do</para>
2002 <screen>$ make boot stage=2</screen>
2004 <para>before <literal>make stage2</literal> in
2005 <literal>ghc/compiler</literal>.</para>
2008 <sect2 id="sec-standard-targets">
2009 <title>Standard Targets</title>
2010 <indexterm><primary>targets, standard makefile</primary></indexterm>
2011 <indexterm><primary>makefile targets</primary></indexterm>
2013 <para>In any directory you should be able to make the following:</para>
2017 <term><literal>boot</literal></term>
2019 <para>does the one-off preparation required to get ready
2020 for the real work. Notably, it does <command>gmake
2021 depend</command> in all directories that contain programs.
2022 It also builds the necessary tools for compilation to
2025 <para>Invoking the <literal>boot</literal> target
2026 explicitly is not normally necessary. From the top-level
2027 <literal>fptools</literal> directory, invoking
2028 <literal>gmake</literal> causes <literal>gmake boot
2029 all</literal> to be invoked in each of the project
2030 subdirectories, in the order specified by
2031 <literal>$(AllTargets)</literal> in
2032 <literal>config.mk</literal>.</para>
2034 <para>If you're working in a subdirectory somewhere and
2035 need to update the dependencies, <literal>gmake
2036 boot</literal> is a good way to do it.</para>
2041 <term><literal>all</literal></term>
2043 <para>makes all the final target(s) for this Makefile.
2044 Depending on which directory you are in a “final
2045 target” may be an executable program, a library
2046 archive, a shell script, or a Postscript file. Typing
2047 <command>gmake</command> alone is generally the same as
2048 typing <command>gmake all</command>.</para>
2053 <term><literal>install</literal></term>
2055 <para>installs the things built by <literal>all</literal>
2056 (except for the documentation). Where does it install
2057 them? That is specified by
2058 <filename>mk/config.mk.in</filename>; you can override it
2059 in <filename>mk/build.mk</filename>, or by running
2060 <command>configure</command> with command-line arguments
2061 like <literal>--bindir=/home/simonpj/bin</literal>; see
2062 <literal>./configure --help</literal> for the full
2068 <term><literal>install-docs</literal></term>
2070 <para>installs the documentation. Otherwise behaves just
2071 like <literal>install</literal>.</para>
2076 <term><literal>uninstall</literal></term>
2078 <para>reverses the effect of
2079 <literal>install</literal>.</para>
2084 <term><literal>clean</literal></term>
2086 <para>Delete all files from the current directory that are
2087 normally created by building the program. Don't delete
2088 the files that record the configuration, or files
2089 generated by <command>gmake boot</command>. Also preserve
2090 files that could be made by building, but normally aren't
2091 because the distribution comes with them.</para>
2096 <term><literal>distclean</literal></term>
2098 <para>Delete all files from the current directory that are
2099 created by configuring or building the program. If you
2100 have unpacked the source and built the program without
2101 creating any other files, <literal>make
2102 distclean</literal> should leave only the files that were
2103 in the distribution.</para>
2108 <term><literal>mostlyclean</literal></term>
2110 <para>Like <literal>clean</literal>, but may refrain from
2111 deleting a few files that people normally don't want to
2117 <term><literal>maintainer-clean</literal></term>
2119 <para>Delete everything from the current directory that
2120 can be reconstructed with this Makefile. This typically
2121 includes everything deleted by
2122 <literal>distclean</literal>, plus more: C source files
2123 produced by Bison, tags tables, Info files, and so
2126 <para>One exception, however: <literal>make
2127 maintainer-clean</literal> should not delete
2128 <filename>configure</filename> even if
2129 <filename>configure</filename> can be remade using a rule
2130 in the <filename>Makefile</filename>. More generally,
2131 <literal>make maintainer-clean</literal> should not delete
2132 anything that needs to exist in order to run
2133 <filename>configure</filename> and then begin to build the
2139 <term><literal>check</literal></term>
2141 <para>run the test suite.</para>
2146 <para>All of these standard targets automatically recurse into
2147 sub-directories. Certain other standard targets do not:</para>
2151 <term><literal>configure</literal></term>
2153 <para>is only available in the root directory
2154 <constant>$(FPTOOLS_TOP)</constant>; it has
2155 been discussed in <XRef
2156 LinkEnd="sec-build-config">.</para>
2161 <term><literal>depend</literal></term>
2163 <para>make a <filename>.depend</filename> file in each
2164 directory that needs it. This <filename>.depend</filename>
2165 file contains mechanically-generated dependency
2166 information; for example, suppose a directory contains a
2167 Haskell source module <filename>Foo.lhs</filename> which
2168 imports another module <literal>Baz</literal>. Then the
2169 generated <filename>.depend</filename> file will contain
2170 the dependency:</para>
2176 <para>which says that the object file
2177 <filename>Foo.o</filename> depends on the interface file
2178 <filename>Baz.hi</filename> generated by compiling module
2179 <literal>Baz</literal>. The <filename>.depend</filename>
2180 file is automatically included by every Makefile.</para>
2185 <term><literal>binary-dist</literal></term>
2187 <para>make a binary distribution. This is the target we
2188 use to build the binary distributions of GHC and
2194 <term><literal>dist</literal></term>
2196 <para>make a source distribution. Note that this target
2197 does “make distclean” as part of its work;
2198 don't use it if you want to keep what you've built.</para>
2203 <para>Most <filename>Makefile</filename>s have targets other
2204 than these. You can discover them by looking in the
2205 <filename>Makefile</filename> itself.</para>
2209 <title>Using a project from the build tree</title>
2211 <para>If you want to build GHC (say) and just use it direct from
2212 the build tree without doing <literal>make install</literal>
2213 first, you can run the in-place driver script:
2214 <filename>ghc/compiler/ghc-inplace</filename>.</para>
2216 <para> Do <emphasis>NOT</emphasis> use
2217 <filename>ghc/compiler/ghc</filename>, or
2218 <filename>ghc/compiler/ghc-6.xx</filename>, as these are the
2219 scripts intended for installation, and contain hard-wired paths
2220 to the installed libraries, rather than the libraries in the
2223 <para>Happy can similarly be run from the build tree, using
2224 <filename>happy/src/happy-inplace</filename>.</para>
2228 <title>Fast Making</title>
2230 <indexterm><primary>fastmake</primary></indexterm>
2231 <indexterm><primary>dependencies, omitting</primary></indexterm>
2232 <indexterm><primary>FAST, makefile variable</primary></indexterm>
2234 <para>Sometimes the dependencies get in the way: if you've made
2235 a small change to one file, and you're absolutely sure that it
2236 won't affect anything else, but you know that
2237 <command>make</command> is going to rebuild everything anyway,
2238 the following hack may be useful:</para>
2244 <para>This tells the make system to ignore dependencies and just
2245 build what you tell it to. In other words, it's equivalent to
2246 temporarily removing the <filename>.depend</filename> file in
2247 the current directory (where <command>mkdependHS</command> and
2248 friends store their dependency information).</para>
2250 <para>A bit of history: GHC used to come with a
2251 <command>fastmake</command> script that did the above job, but
2252 GNU make provides the features we need to do it without
2253 resorting to a script. Also, we've found that fastmaking is
2254 less useful since the advent of GHC's recompilation checker (see
2255 the User's Guide section on "Separate Compilation").</para>
2259 <sect1 id="sec-makefile-arch">
2260 <title>The <filename>Makefile</filename> architecture</title>
2261 <indexterm><primary>makefile architecture</primary></indexterm>
2263 <para><command>make</command> is great if everything
2264 works—you type <command>gmake install</command> and lo! the
2265 right things get compiled and installed in the right places. Our
2266 goal is to make this happen often, but somehow it often doesn't;
2267 instead some weird error message eventually emerges from the
2268 bowels of a directory you didn't know existed.</para>
2270 <para>The purpose of this section is to give you a road-map to
2271 help you figure out what is going right and what is going
2275 <title>Debugging</title>
2277 <para>Debugging <filename>Makefile</filename>s is something of a
2278 black art, but here's a couple of tricks that we find
2279 particularly useful. The following command allows you to see
2280 the contents of any make variable in the context of the current
2281 <filename>Makefile</filename>:</para>
2283 <screen>$ make show VALUE=HS_SRCS</screen>
2285 <para>where you can replace <literal>HS_SRCS</literal> with the
2286 name of any variable you wish to see the value of.</para>
2288 <para>GNU make has a <option>-d</option> option which generates
2289 a dump of the decision procedure used to arrive at a conclusion
2290 about which files should be recompiled. Sometimes useful for
2291 tracking down problems with superfluous or missing
2292 recompilations.</para>
2296 <title>A small project</title>
2298 <para>To get started, let us look at the
2299 <filename>Makefile</filename> for an imaginary small
2300 <literal>fptools</literal> project, <literal>small</literal>.
2301 Each project in <literal>fptools</literal> has its own directory
2302 in <constant>FPTOOLS_TOP</constant>, so the
2303 <literal>small</literal> project will have its own directory
2304 <constant>FPOOLS_TOP/small/</constant>. Inside the
2305 <filename>small/</filename> directory there will be a
2306 <filename>Makefile</filename>, looking something like
2309 <indexterm><primary>Makefile, minimal</primary></indexterm>
2312 # Makefile for fptools project "small"
2315 include $(TOP)/mk/boilerplate.mk
2317 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2320 include $(TOP)/target.mk
2323 <para>this <filename>Makefile</filename> has three
2328 <para>The first section includes
2331 One of the most important
2332 features of GNU <command>make</command> that we use is the ability for a <filename>Makefile</filename> to
2333 include another named file, very like <command>cpp</command>'s <literal>#include</literal>
2338 a file of “boilerplate” code from the level
2339 above (which in this case will be
2340 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>).
2341 As its name suggests, <filename>boilerplate.mk</filename>
2342 consists of a large quantity of standard
2343 <filename>Makefile</filename> code. We discuss this
2344 boilerplate in more detail in <XRef LinkEnd="sec-boiler">.
2345 <indexterm><primary>include, directive in
2346 Makefiles</primary></indexterm> <indexterm><primary>Makefile
2347 inclusion</primary></indexterm></para>
2349 <para>Before the <literal>include</literal> statement, you
2350 must define the <command>make</command> variable
2351 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm>
2352 to be the directory containing the <filename>mk</filename>
2353 directory in which the <filename>boilerplate.mk</filename>
2354 file is. It is <emphasis>not</emphasis> OK to simply say</para>
2357 include ../mk/boilerplate.mk # NO NO NO
2361 <para>Why? Because the <filename>boilerplate.mk</filename>
2362 file needs to know where it is, so that it can, in turn,
2363 <literal>include</literal> other files. (Unfortunately,
2364 when an <literal>include</literal>d file does an
2365 <literal>include</literal>, the filename is treated relative
2366 to the directory in which <command>gmake</command> is being
2367 run, not the directory in which the
2368 <literal>include</literal>d sits.) In general,
2369 <emphasis>every file <filename>foo.mk</filename> assumes
2371 <filename><constant>$(TOP)</constant>/mk/foo.mk</filename>
2372 refers to itself.</emphasis> It is up to the
2373 <filename>Makefile</filename> doing the
2374 <literal>include</literal> to ensure this is the case.</para>
2376 <para>Files intended for inclusion in other
2377 <filename>Makefile</filename>s are written to have the
2378 following property: <emphasis>after
2379 <filename>foo.mk</filename> is <literal>include</literal>d,
2380 it leaves <constant>TOP</constant> containing the same value
2381 as it had just before the <literal>include</literal>
2382 statement</emphasis>. In our example, this invariant
2383 guarantees that the <literal>include</literal> for
2384 <filename>target.mk</filename> will look in the same
2385 directory as that for <filename>boilerplate.mk</filename>.</para>
2389 <para> The second section defines the following standard
2390 <command>make</command> variables:
2391 <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm>
2392 (the source files from which is to be built), and
2393 <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>
2394 (the executable binary to be built). We will discuss in
2395 more detail what the “standard variables” are,
2396 and how they affect what happens, in <XRef
2397 LinkEnd="sec-targets">.</para>
2399 <para>The definition for <constant>SRCS</constant> uses the
2400 useful GNU <command>make</command> construct
2401 <literal>$(wildcard $pat$)</literal><indexterm><primary>wildcard</primary></indexterm>,
2402 which expands to a list of all the files matching the
2403 pattern <literal>pat</literal> in the current directory. In
2404 this example, <constant>SRCS</constant> is set to the list
2405 of all the <filename>.lhs</filename> and
2406 <filename>.c</filename> files in the directory. (Let's
2407 suppose there is one of each, <filename>Foo.lhs</filename>
2408 and <filename>Baz.c</filename>.)</para>
2412 <para>The last section includes a second file of standard
2414 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>.
2415 It contains the rules that tell <command>gmake</command> how
2416 to make the standard targets (<Xref
2417 LinkEnd="sec-standard-targets">). Why, you ask, can't this
2418 standard code be part of
2419 <filename>boilerplate.mk</filename>? Good question. We
2420 discuss the reason later, in <Xref
2421 LinkEnd="sec-boiler-arch">.</para>
2423 <para>You do not <emphasis>have</emphasis> to
2424 <literal>include</literal> the
2425 <filename>target.mk</filename> file. Instead, you can write
2426 rules of your own for all the standard targets. Usually,
2427 though, you will find quite a big payoff from using the
2428 canned rules in <filename>target.mk</filename>; the price
2429 tag is that you have to understand what canned rules get
2430 enabled, and what they do (<Xref
2431 LinkEnd="sec-targets">).</para>
2435 <para>In our example <filename>Makefile</filename>, most of the
2436 work is done by the two <literal>include</literal>d files. When
2437 you say <command>gmake all</command>, the following things
2442 <para><command>gmake</command> figures out that the object
2443 files are <filename>Foo.o</filename> and
2444 <filename>Baz.o</filename>.</para>
2448 <para>It uses a boilerplate pattern rule to compile
2449 <filename>Foo.lhs</filename> to <filename>Foo.o</filename>
2450 using a Haskell compiler. (Which one? That is set in the
2451 build configuration.)</para>
2455 <para>It uses another standard pattern rule to compile
2456 <filename>Baz.c</filename> to <filename>Baz.o</filename>,
2457 using a C compiler. (Ditto.)</para>
2461 <para>It links the resulting <filename>.o</filename> files
2462 together to make <literal>small</literal>, using the Haskell
2463 compiler to do the link step. (Why not use
2464 <command>ld</command>? Because the Haskell compiler knows
2465 what standard libraries to link in. How did
2466 <command>gmake</command> know to use the Haskell compiler to
2467 do the link, rather than the C compiler? Because we set the
2468 variable <constant>HS_PROG</constant> rather than
2469 <constant>C_PROG</constant>.)</para>
2473 <para>All <filename>Makefile</filename>s should follow the above
2474 three-section format.</para>
2478 <title>A larger project</title>
2480 <para>Larger projects are usually structured into a number of
2481 sub-directories, each of which has its own
2482 <filename>Makefile</filename>. (In very large projects, this
2483 sub-structure might be iterated recursively, though that is
2484 rare.) To give you the idea, here's part of the directory
2485 structure for the (rather large) GHC project:</para>
2495 ...source files for documentation...
2498 ...source files for driver...
2501 parser/...source files for parser...
2502 renamer/...source files for renamer...
2506 <para>The sub-directories <filename>docs</filename>,
2507 <filename>driver</filename>, <filename>compiler</filename>, and
2508 so on, each contains a sub-component of GHC, and each has its
2509 own <filename>Makefile</filename>. There must also be a
2510 <filename>Makefile</filename> in
2511 <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>.
2512 It does most of its work by recursively invoking
2513 <command>gmake</command> on the <filename>Makefile</filename>s
2514 in the sub-directories. We say that
2515 <filename>ghc/Makefile</filename> is a <emphasis>non-leaf
2516 <filename>Makefile</filename></emphasis>, because it does little
2517 except organise its children, while the
2518 <filename>Makefile</filename>s in the sub-directories are all
2519 <emphasis>leaf <filename>Makefile</filename>s</emphasis>. (In
2520 principle the sub-directories might themselves contain a
2521 non-leaf <filename>Makefile</filename> and several
2522 sub-sub-directories, but that does not happen in GHC.)</para>
2524 <para>The <filename>Makefile</filename> in
2525 <filename>ghc/compiler</filename> is considered a leaf
2526 <filename>Makefile</filename> even though the
2527 <filename>ghc/compiler</filename> has sub-directories, because
2528 these sub-directories do not themselves have
2529 <filename>Makefile</filename>s in them. They are just used to
2530 structure the collection of modules that make up GHC, but all
2531 are managed by the single <filename>Makefile</filename> in
2532 <filename>ghc/compiler</filename>.</para>
2534 <para>You will notice that <filename>ghc/</filename> also
2535 contains a directory <filename>ghc/mk/</filename>. It contains
2536 GHC-specific <filename>Makefile</filename> boilerplate code.
2537 More precisely:</para>
2541 <para><filename>ghc/mk/boilerplate.mk</filename> is included
2542 at the top of <filename>ghc/Makefile</filename>, and of all
2543 the leaf <filename>Makefile</filename>s in the
2544 sub-directories. It in turn <literal>include</literal>s the
2545 main boilerplate file
2546 <filename>mk/boilerplate.mk</filename>.</para>
2550 <para><filename>ghc/mk/target.mk</filename> is
2551 <literal>include</literal>d at the bottom of
2552 <filename>ghc/Makefile</filename>, and of all the leaf
2553 <filename>Makefile</filename>s in the sub-directories. It
2554 in turn <literal>include</literal>s the file
2555 <filename>mk/target.mk</filename>.</para>
2559 <para>So these two files are the place to look for GHC-wide
2560 customisation of the standard boilerplate.</para>
2563 <sect2 id="sec-boiler-arch">
2564 <title>Boilerplate architecture</title>
2565 <indexterm><primary>boilerplate architecture</primary></indexterm>
2567 <para>Every <filename>Makefile</filename> includes a
2568 <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>
2569 file at the top, and
2570 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
2571 file at the bottom. In this section we discuss what is in these
2572 files, and why there have to be two of them. In general:</para>
2576 <para><filename>boilerplate.mk</filename> consists of:</para>
2580 <para><emphasis>Definitions of millions of
2581 <command>make</command> variables</emphasis> that
2582 collectively specify the build configuration. Examples:
2583 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2584 the options to feed to the Haskell compiler;
2585 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>,
2586 the sub-directories to enable within the
2587 <literal>nofib</literal> project;
2588 <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>,
2589 the name of the Haskell compiler to use when compiling
2590 GHC in the <literal>ghc</literal> project.</para>
2594 <para><emphasis>Standard pattern rules</emphasis> that
2595 tell <command>gmake</command> how to construct one file
2596 from another.</para>
2600 <para><filename>boilerplate.mk</filename> needs to be
2601 <literal>include</literal>d at the <emphasis>top</emphasis>
2602 of each <filename>Makefile</filename>, so that the user can
2603 replace the boilerplate definitions or pattern rules by
2604 simply giving a new definition or pattern rule in the
2605 <filename>Makefile</filename>. <command>gmake</command>
2606 simply takes the last definition as the definitive one.</para>
2608 <para>Instead of <emphasis>replacing</emphasis> boilerplate
2609 definitions, it is also quite common to
2610 <emphasis>augment</emphasis> them. For example, a
2611 <filename>Makefile</filename> might say:</para>
2617 <para>thereby adding “<option>-O</option>” to
2619 <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.</para>
2623 <para><filename>target.mk</filename> contains
2624 <command>make</command> rules for the standard targets
2625 described in <Xref LinkEnd="sec-standard-targets">. These
2626 rules are selectively included, depending on the setting of
2627 certain <command>make</command> variables. These variables
2628 are usually set in the middle section of the
2629 <filename>Makefile</filename> between the two
2630 <literal>include</literal>s.</para>
2632 <para><filename>target.mk</filename> must be included at the
2633 end (rather than being part of
2634 <filename>boilerplate.mk</filename>) for several tiresome
2640 <para><command>gmake</command> commits target and
2641 dependency lists earlier than it should. For example,
2642 <FIlename>target.mk</FIlename> has a rule that looks
2646 $(HS_PROG) : $(OBJS)
2647 $(HC) $(LD_OPTS) $< -o $@
2650 <para>If this rule was in
2651 <filename>boilerplate.mk</filename> then
2652 <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2654 <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm>
2655 would not have their final values at the moment
2656 <command>gmake</command> encountered the rule. Alas,
2657 <command>gmake</command> takes a snapshot of their
2658 current values, and wires that snapshot into the rule.
2659 (In contrast, the commands executed when the rule
2660 “fires” are only substituted at the moment
2661 of firing.) So, the rule must follow the definitions
2662 given in the <filename>Makefile</filename> itself.</para>
2666 <para>Unlike pattern rules, ordinary rules cannot be
2667 overriden or replaced by subsequent rules for the same
2668 target (at least, not without an error message).
2669 Including ordinary rules in
2670 <filename>boilerplate.mk</filename> would prevent the
2671 user from writing rules for specific targets in specific
2676 <para>There are a couple of other reasons I've
2677 forgotten, but it doesn't matter too much.</para>
2684 <sect2 id="sec-boiler">
2685 <title>The main <filename>mk/boilerplate.mk</filename> file</title>
2686 <indexterm><primary>boilerplate.mk</primary></indexterm>
2688 <para>If you look at
2689 <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename>
2690 you will find that it consists of the following sections, each
2691 held in a separate file:</para>
2695 <term><filename>config.mk</filename></term>
2696 <indexterm><primary>config.mk</primary></indexterm>
2698 <para>is the build configuration file we discussed at
2699 length in <Xref LinkEnd="sec-build-config">.</para>
2704 <term><filename>paths.mk</filename></term>
2705 <indexterm><primary>paths.mk</primary></indexterm>
2707 <para>defines <command>make</command> variables for
2708 pathnames and file lists. This file contains code for
2709 automatically compiling lists of source files and deriving
2710 lists of object files from those. The results can be
2711 overriden in the <filename>Makefile</filename>, but in
2712 most cases the automatic setup should do the right
2715 <para>The following variables may be set in the
2716 <filename>Makefile</filename> to affect how the automatic
2717 source file search is done:</para>
2721 <term><literal>ALL_DIRS</literal></term>
2722 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2725 <para>Set to a list of directories to search in
2726 addition to the current directory for source
2732 <term><literal>EXCLUDE_SRCS</literal></term>
2733 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2736 <para>Set to a list of source files (relative to the
2737 current directory) to omit from the automatic
2738 search. The source searching machinery is clever
2739 enough to know that if you exclude a source file
2740 from which other sources are derived, then the
2741 derived sources should also be excluded. For
2742 example, if you set <literal>EXCLUDED_SRCS</literal>
2743 to include <filename>Foo.y</filename>, then
2744 <filename>Foo.hs</filename> will also be
2750 <term><literal>EXTRA_SRCS</literal></term>
2751 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2754 <para>Set to a list of extra source files (perhaps
2755 in directories not listed in
2756 <literal>ALL_DIRS</literal>) that should be
2762 <para>The results of the automatic source file search are
2763 placed in the following make variables:</para>
2767 <term><literal>SRCS</literal></term>
2768 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2770 <para>All source files found, sorted and without
2771 duplicates, including those which might not exist
2772 yet but will be derived from other existing sources.
2773 <literal>SRCS</literal> <emphasis>can</emphasis> be
2774 overriden if necessary, in which case the variables
2775 below will follow suit.</para>
2780 <term><literal>HS_SRCS</literal></term>
2781 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2783 <para>all Haskell source files in the current
2784 directory, including those derived from other source
2785 files (eg. Happy sources also give rise to Haskell
2791 <term><literal>HS_OBJS</literal></term>
2792 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2794 <para>Object files derived from
2795 <literal>HS_SRCS</literal>.</para>
2800 <term><literal>HS_IFACES</literal></term>
2801 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2803 <para>Interface files (<literal>.hi</literal> files)
2804 derived from <literal>HS_SRCS</literal>.</para>
2809 <term><literal>C_SRCS</literal></term>
2810 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2812 <para>All C source files found.</para>
2817 <term><literal>C_OBJS</literal></term>
2818 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2820 <para>Object files derived from
2821 <literal>C_SRCS</literal>.</para>
2826 <term><literal>SCRIPT_SRCS</literal></term>
2827 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2829 <para>All script source files found
2830 (<literal>.lprl</literal> files).</para>
2835 <term><literal>SCRIPT_OBJS</literal></term>
2836 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2838 <para><quote>object</quote> files derived from
2839 <literal>SCRIPT_SRCS</literal>
2840 (<literal>.prl</literal> files).</para>
2845 <term><literal>HSC_SRCS</literal></term>
2846 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2848 <para>All <literal>hsc2hs</literal> source files
2849 (<literal>.hsc</literal> files).</para>
2854 <term><literal>HAPPY_SRCS</literal></term>
2855 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2857 <para>All <literal>happy</literal> source files
2858 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2863 <term><literal>OBJS</literal></term>
2864 <indexterm><primary>OBJS</primary></indexterm>
2866 <para>the concatenation of
2867 <literal>$(HS_OBJS)</literal>,
2868 <literal>$(C_OBJS)</literal>, and
2869 <literal>$(SCRIPT_OBJS)</literal>.</para>
2874 <para>Any or all of these definitions can easily be
2875 overriden by giving new definitions in your
2876 <filename>Makefile</filename>.</para>
2878 <para>What, exactly, does <filename>paths.mk</filename>
2879 consider a <quote>source file</quote> to be? It's based
2880 on the file's suffix (e.g. <filename>.hs</filename>,
2881 <filename>.lhs</filename>, <filename>.c</filename>,
2882 <filename>.hy</filename>, etc), but this is the kind of
2883 detail that changes, so rather than enumerate the source
2884 suffices here the best thing to do is to look in
2885 <filename>paths.mk</filename>.</para>
2890 <term><filename>opts.mk</filename></term>
2891 <indexterm><primary>opts.mk</primary></indexterm>
2893 <para>defines <command>make</command> variables for option
2894 strings to pass to each program. For example, it defines
2895 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2896 the option strings to pass to the Haskell compiler. See
2897 <Xref LinkEnd="sec-suffix">.</para>
2902 <term><filename>suffix.mk</filename></term>
2903 <indexterm><primary>suffix.mk</primary></indexterm>
2905 <para>defines standard pattern rules—see <Xref
2906 LinkEnd="sec-suffix">.</para>
2911 <para>Any of the variables and pattern rules defined by the
2912 boilerplate file can easily be overridden in any particular
2913 <filename>Makefile</filename>, because the boilerplate
2914 <literal>include</literal> comes first. Definitions after this
2915 <literal>include</literal> directive simply override the default
2916 ones in <filename>boilerplate.mk</filename>.</para>
2919 <sect2 id="sec-suffix">
2920 <title>Pattern rules and options</title>
2921 <indexterm><primary>Pattern rules</primary></indexterm>
2924 <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm>
2925 defines standard <emphasis>pattern rules</emphasis> that say how
2926 to build one kind of file from another, for example, how to
2927 build a <filename>.o</filename> file from a
2928 <filename>.c</filename> file. (GNU <command>make</command>'s
2929 <emphasis>pattern rules</emphasis> are more powerful and easier
2930 to use than Unix <command>make</command>'s <emphasis>suffix
2931 rules</emphasis>.)</para>
2933 <para>Almost all the rules look something like this:</para>
2938 $(CC) $(CC_OPTS) -c $< -o $@
2941 <para>Here's how to understand the rule. It says that
2942 <emphasis>something</emphasis><filename>.o</filename> (say
2943 <filename>Foo.o</filename>) can be built from
2944 <emphasis>something</emphasis><filename>.c</filename>
2945 (<filename>Foo.c</filename>), by invoking the C compiler (path
2946 name held in <constant>$(CC)</constant>), passing to it
2947 the options <constant>$(CC_OPTS)</constant> and
2948 the rule's dependent file of the rule
2949 <literal>$<</literal> (<filename>Foo.c</filename> in
2950 this case), and putting the result in the rule's target
2951 <literal>$@</literal> (<filename>Foo.o</filename> in this
2954 <para>Every program is held in a <command>make</command>
2955 variable defined in <filename>mk/config.mk</filename>—look
2956 in <filename>mk/config.mk</filename> for the complete list. One
2957 important one is the Haskell compiler, which is called
2958 <constant>$(HC)</constant>.</para>
2960 <para>Every program's options are are held in a
2961 <command>make</command> variables called
2962 <constant><prog>_OPTS</constant>. the
2963 <constant><prog>_OPTS</constant> variables are
2964 defined in <filename>mk/opts.mk</filename>. Almost all of them
2965 are defined like this:</para>
2968 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
2971 <para>The four variables from which
2972 <constant>CC_OPTS</constant> is built have the following
2977 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
2979 <para>options passed to all C compilations.</para>
2984 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
2986 <para>options passed to C compilations for way
2987 <literal><way></literal>. For example,
2988 <constant>WAY_mp_CC_OPTS</constant>
2989 gives options to pass to the C compiler when compiling way
2990 <literal>mp</literal>. The variable
2991 <constant>WAY_CC_OPTS</constant> holds
2992 options to pass to the C compiler when compiling the
2993 standard way. (<Xref LinkEnd="sec-ways"> dicusses
2994 multi-way compilation.)</para>
2999 <term><constant><module>_CC_OPTS</constant>:</term>
3001 <para>options to pass to the C compiler that are specific
3002 to module <literal><module></literal>. For example,
3003 <constant>SMap_CC_OPTS</constant> gives the
3004 specific options to pass to the C compiler when compiling
3005 <filename>SMap.c</filename>.</para>
3010 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3012 <para>extra options to pass to all C compilations. This
3013 is intended for command line use, thus:</para>
3016 gmake libHS.a EXTRA_CC_OPTS="-v"
3023 <sect2 id="sec-targets">
3024 <title>The main <filename>mk/target.mk</filename> file</title>
3025 <indexterm><primary>target.mk</primary></indexterm>
3027 <para><filename>target.mk</filename> contains canned rules for
3028 all the standard targets described in <Xref
3029 LinkEnd="sec-standard-targets">. It is complicated by the fact
3030 that you don't want all of these rules to be active in every
3031 <filename>Makefile</filename>. Rather than have a plethora of
3032 tiny files which you can include selectively, there is a single
3033 file, <filename>target.mk</filename>, which selectively includes
3034 rules based on whether you have defined certain variables in
3035 your <filename>Makefile</filename>. This section explains what
3036 rules you get, what variables control them, and what the rules
3037 do. Hopefully, you will also get enough of an idea of what is
3038 supposed to happen that you can read and understand any weird
3039 special cases yourself.</para>
3043 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3045 <para>If <constant>HS_PROG</constant> is defined,
3046 you get rules with the following targets:</para>
3050 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3052 <para>itself. This rule links
3053 <constant>$(OBJS)</constant> with the Haskell
3054 runtime system to get an executable called
3055 <constant>$(HS_PROG)</constant>.</para>
3060 <term><literal>install</literal><indexterm><primary>install</primary></indexterm></term>
3063 <constant>$(HS_PROG)</constant> in
3064 <constant>$(bindir)</constant>.</para>
3073 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3075 <para>is similar to <constant>HS_PROG</constant>,
3076 except that the link step links
3077 <constant>$(C_OBJS)</constant> with the C
3078 runtime system.</para>
3083 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3085 <para>is similar to <constant>HS_PROG</constant>,
3086 except that it links
3087 <constant>$(LIB_OBJS)</constant> to make the
3088 library archive <constant>$(LIBRARY)</constant>,
3089 and <literal>install</literal> installs it in
3090 <constant>$(libdir)</constant>.</para>
3095 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3097 <para>…</para>
3102 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3104 <para>…</para>
3109 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3111 <para>If <constant>HS_SRCS</constant> is defined
3112 and non-empty, a rule for the target
3113 <literal>depend</literal> is included, which generates
3114 dependency information for Haskell programs. Similarly
3115 for <constant>C_SRCS</constant>.</para>
3120 <para>All of these rules are “double-colon” rules,
3124 install :: $(HS_PROG)
3125 ...how to install it...
3128 <para>GNU <command>make</command> treats double-colon rules as
3129 separate entities. If there are several double-colon rules for
3130 the same target it takes each in turn and fires it if its
3131 dependencies say to do so. This means that you can, for
3132 example, define both <constant>HS_PROG</constant> and
3133 <constant>LIBRARY</constant>, which will generate two rules for
3134 <literal>install</literal>. When you type <command>gmake
3135 install</command> both rules will be fired, and both the program
3136 and the library will be installed, just as you wanted.</para>
3139 <sect2 id="sec-subdirs">
3140 <title>Recursion</title>
3141 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3142 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm>
3144 <para>In leaf <filename>Makefile</filename>s the variable
3145 <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm>
3146 is undefined. In non-leaf <filename>Makefile</filename>s,
3147 <constant>SUBDIRS</constant> is set to the list of
3148 sub-directories that contain subordinate
3149 <filename>Makefile</filename>s. <emphasis>It is up to you to
3150 set <constant>SUBDIRS</constant> in the
3151 <filename>Makefile</filename>.</emphasis> There is no automation
3152 here—<constant>SUBDIRS</constant> is too important to
3155 <para>When <constant>SUBDIRS</constant> is defined,
3156 <filename>target.mk</filename> includes a rather neat rule for
3157 the standard targets (<Xref LinkEnd="sec-standard-targets"> that
3158 simply invokes <command>make</command> recursively in each of
3159 the sub-directories.</para>
3161 <para><emphasis>These recursive invocations are guaranteed to
3162 occur in the order in which the list of directories is specified
3163 in <constant>SUBDIRS</constant>. </emphasis>This guarantee can
3164 be important. For example, when you say <command>gmake
3165 boot</command> it can be important that the recursive invocation
3166 of <command>make boot</command> is done in one sub-directory
3167 (the include files, say) before another (the source files).
3168 Generally, put the most independent sub-directory first, and the
3169 most dependent last.</para>
3172 <sect2 id="sec-ways">
3173 <title>Way management</title>
3174 <indexterm><primary>way management</primary></indexterm>
3176 <para>We sometimes want to build essentially the same system in
3177 several different “ways”. For example, we want to build GHC's
3178 <literal>Prelude</literal> libraries with and without profiling,
3179 so that there is an appropriately-built library archive to link
3180 with when the user compiles his program. It would be possible
3181 to have a completely separate build tree for each such “way”,
3182 but it would be horribly bureaucratic, especially since often
3183 only parts of the build tree need to be constructed in multiple
3187 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3188 contains some clever magic to allow you to build several
3189 versions of a system; and to control locally how many versions
3190 are built and how they differ. This section explains the
3193 <para>The files for a particular way are distinguished by
3194 munging the suffix. The <quote>normal way</quote> is always
3195 built, and its files have the standard suffices
3196 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3197 In addition, you can build one or more extra ways, each
3198 distinguished by a <emphasis>way tag</emphasis>. The object
3199 files and interface files for one of these extra ways are
3200 distinguished by their suffix. For example, way
3201 <literal>mp</literal> has files
3202 <filename>.mp_o</filename> and
3203 <filename>.mp_hi</filename>. Library archives have their
3204 way tag the other side of the dot, for boring reasons; thus,
3205 <filename>libHS_mp.a</filename>.</para>
3207 <para>A <command>make</command> variable called
3208 <constant>way</constant> holds the current way tag.
3209 <emphasis><constant>way</constant> is only ever set on the
3210 command line of <command>gmake</command></emphasis> (usually in
3211 a recursive invocation of <command>gmake</command> by the
3212 system). It is never set inside a
3213 <filename>Makefile</filename>. So it is a global constant for
3214 any one invocation of <command>gmake</command>. Two other
3215 <command>make</command> variables,
3216 <constant>way_</constant> and
3217 <constant>_way</constant> are immediately derived from
3218 <constant>$(way)</constant> and never altered. If
3219 <constant>way</constant> is not set, then neither are
3220 <constant>way_</constant> and
3221 <constant>_way</constant>, and the invocation of
3222 <command>make</command> will build the <quote>normal
3223 way</quote>. If <constant>way</constant> is set, then the other
3224 two variables are set in sympathy. For example, if
3225 <constant>$(way)</constant> is “<literal>mp</literal>”,
3226 then <constant>way_</constant> is set to
3227 “<literal>mp_</literal>” and
3228 <constant>_way</constant> is set to
3229 “<literal>_mp</literal>”. These three variables are
3230 then used when constructing file names.</para>
3232 <para>So how does <command>make</command> ever get recursively
3233 invoked with <constant>way</constant> set? There are two ways
3234 in which this happens:</para>
3238 <para>For some (but not all) of the standard targets, when
3239 in a leaf sub-directory, <command>make</command> is
3240 recursively invoked for each way tag in
3241 <constant>$(WAYS)</constant>. You set
3242 <constant>WAYS</constant> in the
3243 <filename>Makefile</filename> to the list of way tags you
3244 want these targets built for. The mechanism here is very
3245 much like the recursive invocation of
3246 <command>make</command> in sub-directories (<Xref
3247 LinkEnd="sec-subdirs">). It is up to you to set
3248 <constant>WAYS</constant> in your
3249 <filename>Makefile</filename>; this is how you control what
3250 ways will get built.</para>
3254 <para>For a useful collection of targets (such as
3255 <filename>libHS_mp.a</filename>,
3256 <filename>Foo.mp_o</filename>) there is a rule which
3257 recursively invokes <command>make</command> to make the
3258 specified target, setting the <constant>way</constant>
3259 variable. So if you say <command>gmake
3260 Foo.mp_o</command> you should see a recursive
3261 invocation <command>gmake Foo.mp_o way=mp</command>,
3262 and <emphasis>in this recursive invocation the pattern rule
3263 for compiling a Haskell file into a <filename>.o</filename>
3264 file will match</emphasis>. The key pattern rules (in
3265 <filename>suffix.mk</filename>) look like this:
3269 $(HC) $(HC_OPTS) $< -o $@
3276 <para>You can invoke <command>make</command> with a
3277 particular <literal>way</literal> setting yourself, in order
3278 to build files related to a particular
3279 <literal>way</literal> in the current directory. eg.
3285 will build files for the profiling way only in the current
3292 <title>When the canned rule isn't right</title>
3294 <para>Sometimes the canned rule just doesn't do the right thing.
3295 For example, in the <literal>nofib</literal> suite we want the
3296 link step to print out timing information. The thing to do here
3297 is <emphasis>not</emphasis> to define
3298 <constant>HS_PROG</constant> or
3299 <constant>C_PROG</constant>, and instead define a special
3300 purpose rule in your own <filename>Makefile</filename>. By
3301 using different variable names you will avoid the canned rules
3302 being included, and conflicting with yours.</para>
3306 <sect1 id="building-docs">
3307 <title>Building the documentation</title>
3309 <sect2 id="pre-supposed-doc-tools">
3310 <title>Tools for building the Documentation</title>
3312 <para>The following additional tools are required if you want to
3313 format the documentation that comes with the
3314 <literal>fptools</literal> projects:</para>
3318 <term>DocBook</term>
3319 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
3320 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
3322 <para>Much of our documentation is written in SGML, using
3323 the DocBook DTD. Instructions on installing and
3324 configuring the DocBook tools are below.</para>
3330 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
3331 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
3333 <para>A decent TeX distribution is required if you want to
3334 produce printable documentation. We recomment teTeX,
3335 which includes just about everything you need.</para>
3340 <term>Haddock</term>
3341 <indexterm><primary>Haddock</primary>
3344 <para>Haddock is a Haskell documentation tool that we use
3345 for automatically generating documentation from the
3346 library source code. It is an <literal>fptools</literal>
3347 project in itself. To build documentation for the
3348 libraries (<literal>fptools/libraries</literal>) you
3349 should check out and build Haddock in
3350 <literal>fptools/haddock</literal>. Haddock requires GHC
3358 <title>Installing the DocBook tools</title>
3361 <title>Installing the DocBook tools on Linux</title>
3363 <para>If you're on a recent RedHat system (7.0+), you probably
3364 have working DocBook tools already installed. The configure
3365 script should detect your setup and you're away.</para>
3367 <para>If you don't have DocBook tools installed, and you are
3368 using a system that can handle RedHat RPM packages, you can
3369 probably use the <ULink
3370 URL="http://sourceware.cygnus.com/docbook-tools/">Cygnus
3371 DocBook tools</ULink>, which is the most shrink-wrapped SGML
3372 suite that we could find. You need all the RPMs except for
3373 psgml (i.e. <Filename>docbook</Filename>,
3374 <Filename>jade</Filename>, <Filename>jadetex</Filename>,
3375 <Filename>sgmlcommon</Filename> and
3376 <Filename>stylesheets</Filename>). Note that most of these
3377 RPMs are architecture neutral, so are likely to be found in a
3378 <Filename>noarch</Filename> directory. The SuSE RPMs also
3379 work; the RedHat ones <Emphasis>don't</Emphasis> in RedHat 6.2
3380 (7.0 and later should be OK), but they are easy to fix: just
3382 <Filename>/usr/lib/sgml/stylesheets/nwalsh-modular/lib/dblib.dsl</Filename>
3383 to <Filename>/usr/lib/sgml/lib/dblib.dsl</Filename>. </para>
3387 <title>Installing DocBook on FreeBSD</title>
3389 <para>On FreeBSD systems, the easiest way to get DocBook up
3390 and running is to install it from the ports tree or a
3391 pre-compiled package (packages are available from your local
3392 FreeBSD mirror site).</para>
3394 <para>To use the ports tree, do this:
3396 $ cd /usr/ports/textproc/docproj
3399 This installs the FreeBSD documentation project tools, which
3400 includes everything needed to format the GHC
3401 documentation.</para>
3405 <title>Installing from binaries on Windows</title>
3407 <Para>It's a good idea to use Norman Walsh's <ULink
3408 URL="http://nwalsh.com/docbook/dsssl/doc/install.html">installation
3409 notes</ULink> as a guide. You should get version 3.1 of
3410 DocBook, and note that his file <Filename>test.sgm</Filename>
3411 won't work, as it needs version 3.0. You should unpack Jade
3412 into <Filename>\Jade</Filename>, along with the entities,
3413 DocBook into <Filename>\docbook</Filename>, and the DocBook
3414 stylesheets into <Filename>\docbook\stylesheets</Filename> (so
3415 they actually end up in
3416 <Filename>\docbook\stylesheets\docbook</Filename>).</para>
3421 <title>Installing the DocBook tools from source</title>
3426 <para>Install <ULink
3427 URL="http://openjade.sourceforge.net/">OpenJade</ULink>
3428 (Windows binaries are available as well as sources). If you
3429 want DVI, PS, or PDF then install JadeTeX from the
3430 <Filename>dsssl</Filename> subdirectory. (If you get the
3434 ! LaTeX Error: Unknown option implicit=false' for package hyperref'.
3437 your version of <Command>hyperref</Command> is out of date;
3438 download it from CTAN
3439 (<Filename>macros/latex/contrib/supported/hyperref</Filename>),
3440 and make it, ensuring that you have first removed or renamed
3441 your old copy. If you start getting file not found errors
3442 when making the test for <Command>hyperref</Command>, you
3443 can abort at that point and proceed straight to
3444 <Command>make install</Command>, or enter them as
3445 <Filename>../</Filename><Emphasis>filename</Emphasis>.)</para>
3447 <para>Make links from <Filename>virtex</Filename> to
3448 <Filename>jadetex</Filename> and
3449 <Filename>pdfvirtex</Filename> to
3450 <Filename>pdfjadetex</Filename> (otherwise DVI, PostScript
3451 and PDF output will not work). Copy
3452 <Filename>dsssl/*.{dtd,dsl}</Filename> and
3453 <Filename>catalog</Filename> to
3454 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3458 <title>DocBook and the DocBook stylesheets</title>
3460 <para>Get a Zip of <ULink
3461 URL="http://www.oasis-open.org/docbook/sgml/3.1/index.html">DocBook</ULink>
3462 and install the contents in
3463 <Filename>/usr/[local/]/lib/sgml</Filename>.</para>
3465 <para>Get the <ULink
3466 URL="http://nwalsh.com/docbook/dsssl/">DocBook
3467 stylesheets</ULink> and install in
3468 <Filename>/usr/[local/]lib/sgml/stylesheets</Filename>
3469 (thereby creating a subdirectory docbook). For indexing,
3470 copy or link <Filename>collateindex.pl</Filename> from the
3471 DocBook stylesheets archive in <Filename>bin</Filename> into
3472 a directory on your <Constant>PATH</Constant>.</para>
3474 <para>Download the <ULink
3475 URL="http://www.oasis-open.org/cover/ISOEnts.zip">ISO
3476 entities</ULink> into
3477 <Filename>/usr/[local/]lib/sgml</Filename>.</para>
3483 <title>Configuring the DocBook tools</title>
3485 <Para>Once the DocBook tools are installed, the configure script
3486 will detect them and set up the build system accordingly. If you
3487 have a system that isn't supported, let us know, and we'll try
3492 <title>Remaining problems</title>
3494 <para>If you install from source, you'll get a pile of warnings
3497 <Screen>DTDDECL catalog entries are not supported</Screen>
3499 every time you build anything. These can safely be ignored, but
3500 if you find them tedious you can get rid of them by removing all
3501 the <Constant>DTDDECL</Constant> entries from
3502 <Filename>docbook.cat</Filename>.</para>
3506 <title>Building the documentation</title>
3508 <para>To build documentation in a certain format, you can
3509 say, for example,</para>
3515 <para>to build HTML documentation below the current directory.
3516 The available formats are: <literal>dvi</literal>,
3517 <literal>ps</literal>, <literal>pdf</literal>,
3518 <literal>html</literal>, and <literal>rtf</literal>. Note that
3519 not all documentation can be built in all of these formats: HTML
3520 documentation is generally supported everywhere, and DocBook
3521 documentation might support the other formats (depending on what
3522 other tools you have installed).</para>
3524 <para>All of these targets are recursive; that is, saying
3525 <literal>make html</literal> will make HTML docs for all the
3526 documents recursively below the current directory.</para>
3528 <para>Because there are many different formats that the DocBook
3529 documentation can be generated in, you have to select which ones
3530 you want by setting the <literal>SGMLDocWays</literal> variable
3531 to a list of them. For example, in
3532 <filename>build.mk</filename> you might have a line:</para>
3535 SGMLDocWays = html ps
3538 <para>This will cause the documentation to be built in the requested
3539 formats as part of the main build (the default is not to build
3540 any documentation at all).</para>
3544 <title>Installing the documentation</title>
3546 <para>To install the documentation, use:</para>
3552 <para>This will install the documentation into
3553 <literal>$(datadir)</literal> (which defaults to
3554 <literal>$(prefix)/share</literal>). The exception is HTML
3555 documentation, which goes into
3556 <literal>$(datadir)/html</literal>, to keep things tidy.</para>
3558 <para>Note that unless you set <literal>$(SGMLDocWays)</literal>
3559 to a list of formats, the <literal>install-docs</literal> target
3560 won't do anything for SGML documentation.</para>
3566 <sect1 id="sec-porting-ghc">
3567 <title>Porting GHC</title>
3569 <para>This section describes how to port GHC to a currenly
3570 unsupported platform. There are two distinct
3571 possibilities:</para>
3575 <para>The hardware architecture for your system is already
3576 supported by GHC, but you're running an OS that isn't
3577 supported (or perhaps has been supported in the past, but
3578 currently isn't). This is the easiest type of porting job,
3579 but it still requires some careful bootstrapping. Proceed to
3580 <xref linkend="sec-booting-from-hc">.</para>
3584 <para>Your system's hardware architecture isn't supported by
3585 GHC. This will be a more difficult port (though by comparison
3586 perhaps not as difficult as porting gcc). Proceed to <xref
3587 linkend="unregisterised-porting">.</para>
3591 <sect2 id="sec-booting-from-hc">
3592 <title>Booting/porting from C (<filename>.hc</filename>) files</title>
3594 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3595 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3596 <indexterm><primary>porting GHC</primary></indexterm>
3598 <para>Bootstrapping GHC on a system without GHC already
3599 installed is achieved by taking the intermediate C files (known
3600 as HC files) from a GHC compilation on a supported system to the
3601 target machine, and compiling them using gcc to get a working
3604 <para><emphasis>NOTE: GHC versions 5.xx and later are
3605 significantly harder to bootstrap from C than earlier versions.
3606 We recommend starting from version 4.08.2 if you need to
3607 bootstrap in this way.</emphasis></para>
3609 <para>HC files are architecture-dependent (but not
3610 OS-dependent), so you have to get a set that were generated on
3611 similar hardware. There may be some supplied on the GHC
3612 download page, otherwise you'll have to compile some up
3613 yourself, or start from <emphasis>unregisterised</emphasis> HC
3614 files - see <xref linkend="unregisterised-porting">.</para>
3616 <para>The following steps should result in a working GHC build
3617 with full libraries:</para>
3621 <para>Unpack the HC files on top of a fresh source tree
3622 (make sure the source tree version matches the version of
3623 the HC files <emphasis>exactly</emphasis>!). This will
3624 place matching <filename>.hc</filename> files next to the
3625 corresponding Haskell source (<filename>.hs</filename> or
3626 <filename>.lhs</filename>) in the compiler subdirectory
3627 <filename>ghc/compiler</filename> and in the libraries
3628 (subdirectories of <filename>hslibs</filename> and
3629 <literal>libraries</literal>).</para>
3633 <para>The actual build process is fully automated by the
3634 <filename>hc-build</filename> script located in the
3635 <filename>distrib</filename> directory. If you eventually
3636 want to install GHC into the directory
3637 <replaceable>dir</replaceable>, the following
3638 command will execute the whole build process (it won't
3639 install yet):</para>
3642 foo% distrib/hc-build --prefix=<replaceable>dir</replaceable>
3644 <indexterm><primary>--hc-build</primary></indexterm>
3646 <para>By default, the installation directory is
3647 <filename>/usr/local</filename>. If that is what you want,
3648 you may omit the argument to <filename>hc-build</filename>.
3649 Generally, any option given to <filename>hc-build</filename>
3650 is passed through to the configuration script
3651 <filename>configure</filename>. If
3652 <filename>hc-build</filename> successfully completes the
3653 build process, you can install the resulting system, as
3663 <sect2 id="unregisterised-porting">
3664 <title>Porting GHC to a new architecture</title>
3666 <para>The first step in porting to a new architecture is to get
3667 an <firstterm>unregisterised</firstterm> build working. An
3668 unregisterised build is one that compiles via vanilla C only.
3669 By contrast, a registerised build uses the following
3670 architecture-specific hacks for speed:</para>
3674 <para>Global register variables: certain abstract machine
3675 <quote>registers</quote> are mapped to real machine
3676 registers, depending on how many machine registers are
3678 <filename>ghc/includes/MachRegs.h</filename>).</para>
3682 <para>Assembly-mangling: when compiling via C, we feed the
3683 assembly generated by gcc though a Perl script known as the
3684 <firstterm>mangler</firstterm> (see
3685 <filename>ghc/driver/mangler/ghc-asm.lprl</filename>). The
3686 mangler rearranges the assembly to support tail-calls and
3687 various other optimisations.</para>
3691 <para>In an unregisterised build, neither of these hacks are
3692 used — the idea is that the C code generated by the
3693 compiler should compile using gcc only. The lack of these
3694 optimisations costs about a factor of two in performance, but
3695 since unregisterised compilation is usually just a step on the
3696 way to a full registerised port, we don't mind too much.</para>
3699 <title>Building an unregisterised port</title>
3701 <para>The first step is to get some unregisterised HC files.
3702 Either (a) download them from the GHC site (if there are
3703 some available for the right version of GHC), or
3704 (b) build them yourself on any machine with a working
3705 GHC. If at all possible this should be a machine with the
3706 same word size as the target.</para>
3708 <para>There is a script available which should automate the
3709 process of doing the 2-stage bootstrap necessary to get the
3710 unregisterised HC files - it's available in <ulink
3711 url="http://cvs.haskell.org/cgi-bin/cvsweb.cgi/fptools/distrib/cross-port"><filename>fptools/distrib/cross-port</filename></ulink>
3714 <para>Now take these unregisterised HC files to the target
3715 platform and bootstrap a compiler from them as per the
3716 instructions in <xref linkend="sec-booting-from-hc">. In
3717 <filename>build.mk</filename>, you need to tell the build
3718 system that the compiler you're building is
3719 (a) unregisterised itself, and (b) builds
3720 unregisterised binaries. This varies depending on the GHC
3721 version you're bootstraping:</para>
3724 # build.mk for GHC 4.08.x
3725 GhcWithRegisterised=NO
3729 # build.mk for GHC 5.xx and 6.x
3730 GhcUnregisterised=YES
3733 <para>Versions 5.xx and 6.x only: use the option
3734 <option>--enable-hc-boot-unregisterised</option> instead of
3735 <option>--enable-hc-boot</option> when running
3736 <filename>./configure</filename>.</para>
3738 <para>The build may not go through cleanly. We've tried to
3739 stick to writing portable code in most parts of the compiler,
3740 so it should compile on any POSIXish system with gcc, but in
3741 our experience most systems differ from the standards in one
3742 way or another. Deal with any problems as they arise - if you
3743 get stuck, ask the experts on
3744 <email>glasgow-haskell-users@haskell.org</email>.</para>
3746 <para>Once you have the unregisterised compiler up and
3747 running, you can use it to start a registerised port. The
3748 following sections describe the various parts of the system
3749 that will need architecture-specific tweaks in order to get a
3750 registerised build going.</para>
3752 <para>Lots of useful information about the innards of GHC is
3753 available in the <ulink
3754 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3755 Commentary</ulink>, which might be helpful if you run into
3756 some code which needs tweaking for your system.</para>
3760 <title>Porting the RTS</title>
3762 <para>The following files need architecture-specific code for a
3763 registerised build:</para>
3767 <term><filename>ghc/includes/MachRegs.h</filename></term>
3768 <indexterm><primary><filename>MachRegs.h</filename></primary>
3771 <para>Defines the STG-register to machine-register
3772 mapping. You need to know your platform's C calling
3773 convention, and which registers are generally available
3774 for mapping to global register variables. There are
3775 plenty of useful comments in this file.</para>
3779 <term><filename>ghc/includes/TailCalls.h</filename></term>
3780 <indexterm><primary><filename>TailCalls.h</filename></primary>
3783 <para>Macros that cooperate with the mangler (see <xref
3784 linkend="sec-mangler">) to make proper tail-calls
3789 <term><filename>ghc/rts/Adjustor.c</filename></term>
3790 <indexterm><primary><filename>Adjustor.c</filename></primary>
3794 <literal>foreign import "wrapper"</literal>
3796 <literal>foreign export dynamic</literal>).
3797 Not essential for getting GHC bootstrapped, so this file
3798 can be deferred until later if necessary.</para>
3802 <term><filename>ghc/rts/StgCRun.c</filename></term>
3803 <indexterm><primary><filename>StgCRun.c</filename></primary>
3806 <para>The little assembly layer between the C world and
3807 the Haskell world. See the comments and code for the
3808 other architectures in this file for pointers.</para>
3812 <term><filename>ghc/rts/MBlock.h</filename></term>
3813 <term><filename>ghc/rts/MBlock.c</filename></term>
3814 <indexterm><primary><filename>MBlock.h</filename></primary>
3816 <indexterm><primary><filename>MBlock.c</filename></primary>
3819 <para>These files are really OS-specific rather than
3820 architecture-specific. In <filename>MBlock.h</filename>
3821 is specified the absolute location at which the RTS
3822 should try to allocate memory on your platform (try to
3823 find an area which doesn't conflict with code or dynamic
3824 libraries). In <filename>Mblock.c</filename> you might
3825 need to tweak the call to <literal>mmap()</literal> for
3832 <sect3 id="sec-mangler">
3833 <title>The mangler</title>
3835 <para>The mangler is an evil Perl-script that rearranges the
3836 assembly code output from gcc to do two main things:</para>
3840 <para>Remove function prologues and epilogues, and all
3841 movement of the C stack pointer. This is to support
3842 tail-calls: every code block in Haskell code ends in an
3843 explicit jump, so we don't want the C-stack overflowing
3844 while we're jumping around between code blocks.</para>
3847 <para>Move the <firstterm>info table</firstterm> for a
3848 closure next to the entry code for that closure. In
3849 unregisterised code, info tables contain a pointer to the
3850 entry code, but in registerised compilation we arrange
3851 that the info table is shoved right up against the entry
3852 code, and addressed backwards from the entry code pointer
3853 (this saves a word in the info table and an extra
3854 indirection when jumping to the closure entry
3859 <para>The mangler is abstracted to a certain extent over some
3860 architecture-specific things such as the particular assembler
3861 directives used to herald symbols. Take a look at the
3862 definitions for other architectures and use these as a
3863 starting point.</para>
3867 <title>The native code generator</title>
3869 <para>The native code generator isn't essential to getting a
3870 registerised build going, but it's a desirable thing to have
3871 because it can cut compilation times in half. The native code
3872 generator is described in some detail in the <ulink
3873 url="http://www.cse.unsw.edu.au/~chak/haskell/ghc/comm/">GHC
3874 commentary</ulink>.</para>
3880 <para>To support GHCi, you need to port the dynamic linker
3881 (<filename>fptools/ghc/rts/Linker.c</filename>). The linker
3882 currently supports the ELF and PEi386 object file formats - if
3883 your platform uses one of these then you probably don't have
3884 to do anything except fiddle with the
3885 <literal>#ifdef</literal>s at the top of
3886 <filename>Linker.c</filename> to tell it about your OS.</para>
3888 <para>If your system uses a different object file format, then
3889 you have to write a linker — good luck!</para>
3895 <sect1 id="sec-build-pitfalls">
3896 <title>Known pitfalls in building Glasgow Haskell
3898 <indexterm><primary>problems, building</primary></indexterm>
3899 <indexterm><primary>pitfalls, in building</primary></indexterm>
3900 <indexterm><primary>building pitfalls</primary></indexterm></title>
3903 WARNINGS about pitfalls and known “problems”:
3912 One difficulty that comes up from time to time is running out of space
3913 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
3914 compensate for the vagaries of different sysadmin approaches to temp
3916 <indexterm><primary>tmp, running out of space in</primary></indexterm>
3918 The quickest way around it is <command>setenv TMPDIR /usr/tmp</command><indexterm><primary>TMPDIR</primary></indexterm> or
3919 even <command>setenv TMPDIR .</command> (or the equivalent incantation with your shell
3922 The best way around it is to say
3925 export TMPDIR=<dir>
3928 in your <filename>build.mk</filename> file.
3929 Then GHC and the other <literal>fptools</literal> programs will use the appropriate directory
3938 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
3939 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
3947 When compiling via C, you'll sometimes get “warning: assignment from
3948 incompatible pointer type” out of GCC. Harmless.
3955 Similarly, <command>ar</command>chiving warning messages like the following are not
3959 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
3960 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
3970 In compiling the compiler proper (in <filename>compiler/</filename>), you <emphasis>may</emphasis>
3971 get an “Out of heap space” error message. These can vary with the
3972 vagaries of different systems, it seems. The solution is simple:
3979 If you're compiling with GHC 4.00 or later, then the
3980 <emphasis>maximum</emphasis> heap size must have been reached. This
3981 is somewhat unlikely, since the maximum is set to 64M by default.
3982 Anyway, you can raise it with the
3983 <option>-optCrts-M<size></option> flag (add this flag to
3984 <constant><module>_HC_OPTS</constant>
3985 <command>make</command> variable in the appropriate
3986 <filename>Makefile</filename>).
3993 For GHC < 4.00, add a suitable <option>-H</option> flag to the <filename>Makefile</filename>, as
4002 and try again: <command>gmake</command>. (see <Xref LinkEnd="sec-suffix"> for information about
4003 <constant><module>_HC_OPTS</constant>.)
4005 Alternatively, just cut to the chase:
4009 % make EXTRA_HC_OPTS=-optCrts-M128M
4018 If you try to compile some Haskell, and you get errors from GCC about
4019 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
4020 mis-installed. <command>fixincludes</command> wasn't run when it should've been.
4022 As <command>fixincludes</command> is now automagically run as part of GCC installation,
4023 this bug also suggests that you have an old GCC.
4031 You <emphasis>may</emphasis> need to re-<command>ranlib</command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
4035 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
4036 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
4038 ? # or, on some machines: ar s $i
4043 We'd be interested to know if this is still necessary.
4051 GHC's sources go through <command>cpp</command> before being compiled, and <command>cpp</command> varies
4052 a bit from one Unix to another. One particular gotcha is macro calls
4057 SLIT("Hello, world")
4061 Some <command>cpp</command>s treat the comma inside the string as separating two macro
4062 arguments, so you get
4066 :731: macro `SLIT' used with too many (2) args
4070 Alas, <command>cpp</command> doesn't tell you the offending file!
4072 Workaround: don't put weird things in string args to <command>cpp</command> macros.
4083 <Sect1 id="winbuild"><Title>Notes for building under Windows</Title>
4086 This section summarises how to get the utilities you need on your
4087 Win95/98/NT/2000 machine to use CVS and build GHC. Similar notes for
4088 installing and running GHC may be found in the user guide. In general,
4089 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
4090 You should read the GHC installation guide sections on Windows (in the user
4091 guide) before continuing to read these notes.
4095 <sect2 id="cygwin-and-mingw"><Title>Cygwin and MinGW</Title>
4097 <para> The Windows situation for building GHC is rather confusing. This section
4098 tries to clarify, and to establish terminology.</para>
4100 <sect3 id="ghc-mingw"><title>GHC-mingw</title>
4102 <para> <ulink url="http://www.mingw.org">MinGW (Minimalist GNU for Windows)</ulink>
4103 is a collection of header
4104 files and import libraries that allow one to use <command>gcc</command> and produce
4105 native Win32 programs that do not rely on any third-party DLLs. The
4106 current set of tools include GNU Compiler Collection (<command>gcc</command>), GNU Binary
4107 Utilities (Binutils), GNU debugger (Gdb), GNU make, and a assorted
4110 <para>The GHC that we distribute includes, inside the distribution itself, the MinGW <command>gcc</command>,
4111 <command>as</command>, <command>ld</command>, and a bunch of input/output libraries.
4112 GHC compiles Haskell to C (or to
4113 assembly code), and then invokes these MinGW tools to generate an executable binary.
4114 The resulting binaries can run on any Win32 system.
4116 <para> We will call a GHC that targets MinGW in this way <emphasis>GHC-mingw</emphasis>.</para>
4118 <para> The down-side of GHC-mingw is that the MinGW libraries do not support anything like the full
4119 Posix interface. So programs compiled with GHC-mingw cannot import the (Haskell) Posix
4120 library; they have to do
4121 their input output using standard Haskell I/O libraries, or native Win32 bindings.
4125 <sect3 id="ghc-cygwin"><title>GHC-cygwin</title>
4127 <para>There <emphasis>is</emphasis> a way to get the full Posix interface, which is to use Cygwin.
4128 <ulink url="http://www.cygwin.com">Cygwin</ulink> is a complete Unix simulation that runs on Win32.
4129 Cygwin comes with a shell, and all the usual Unix commands: <command>mv</command>, <command>rm</command>,
4130 <command>ls</command>, plus of course <command>gcc</command>, <command>ld</command> and so on.
4131 A C program compiled with the Cygwin <command>gcc</command> certainly can use all of Posix.
4133 <para>So why doesn't GHC use the Cygwin <command>gcc</command> and libraries? Because
4134 Cygwin comes with a DLL <emphasis>that must be linked with every runnable Cygwin-compiled program</emphasis>.
4135 A program compiled by the Cygwin tools cannot run at all unless Cygwin is installed.
4136 If GHC targeted Cygwin, users would have to install Cygwin just to run the Haskell programs
4137 that GHC compiled; and the Cygwin DLL would have to be in the DLL load path.
4138 Worse, Cygwin is a moving target. The name of the main DLL, <literal>cygwin1.dll</literal>
4139 does not change, but the implementation certainly does. Even the interfaces to functions
4140 it exports seem to change occasionally. So programs compiled by GHC might only run with
4141 particular versions of Cygwin. All of this seems very undesirable.
4144 Nevertheless, it is certainly possible to build a version of GHC that targets Cygwin;
4145 we will call that <emphasis>GHC-cygwin</emphasis>. The up-side of GHC-cygwin is
4146 that Haskell programs compiled by GHC-cygwin can import the (Haskell) Posix library.
4150 <sect3><title>HOST_OS vs TARGET_OS</title>
4153 In the source code you'll find various ifdefs looking like:
4155 #ifdef mingw32_HOST_OS
4161 #ifdef mingw32_TARGET_OS
4165 These macros are set by the configure script (via the file config.h).
4166 Which is which? The criterion is this. In the ifdefs in GHC's source code:
4169 The "host" system is the one on which GHC itself will be run.
4172 The "target" system is the one for which the program compiled by GHC will be run.
4175 For a stage-2 compiler, in which GHCi is available, the "host" and "target" systems must be the same.
4176 So then it doesn't really matter whether you use the HOST_OS or TARGET_OS cpp macros.
4181 <sect3><title>Summary</title>
4183 <para>Notice that "GHC-mingw" means "GHC that <emphasis>targets</emphasis> MinGW". It says nothing about
4184 how that GHC was <emphasis>built</emphasis>. It is entirely possible to have a GHC-mingw that was built
4185 by compiling GHC's Haskell sources with a GHC-cygwin, or vice versa.</para>
4187 <para>We distribute only a GHC-mingw built by a GHC-mingw; supporting
4188 GHC-cygwin too is beyond our resources. The GHC we distribute
4189 therefore does not require Cygwin to run, nor do the programs it
4190 compiles require Cygwin.</para>
4192 <para>The instructions that follow describe how to build GHC-mingw. It is
4193 possible to build GHC-cygwin, but it's not a supported route, and the build system might
4196 <para>In your build tree, you build a compiler called <Command>ghc-inplace</Command>. It
4197 uses the <Command>gcc</Command> that you specify using the
4198 <option>--with-gcc</option> flag when you run
4199 <Command>configure</Command> (see below).
4200 The makefiles are careful to use <Command>ghc-inplace</Command> (not <Command>gcc</Command>)
4201 to compile any C files, so that it will in turn invoke the right <Command>gcc</Command> rather that
4202 whatever one happens to be in your path. However, the makefiles do use whatever <Command>ld</Command>
4203 and <Command>ar</Command> happen to be in your path. This is a bit naughty, but (a) they are only
4204 used to glom together .o files into a bigger .o file, or a .a file,
4205 so they don't ever get libraries (which would be bogus; they might be the wrong libraries), and (b)
4206 Cygwin and Mingw use the same .o file format. So its ok.
4211 <Sect2><Title>Installing and configuring Cygwin</Title>
4213 <para>You don't need Cygwin to <emphasis>use</emphasis> GHC,
4214 but you do need it to <emphasis>build</emphasis> GHC.</para>
4216 <para> Install Cygwin from <ulink url="http://www.cygwin.com/">http://www.cygwin.com/</ulink>.
4217 The installation process is straightforward; we install it in <Filename>c:/cygwin</Filename>.
4218 During the installation dialogue, make sure that you select:
4219 <command>cvs</command>, <command>openssh</command>,
4220 <command>autoconf</command>,
4221 <command>binutils</command> (includes ld and (I think) ar),
4222 <command>gcc</command>,
4223 <command>flex</command>,
4224 <command>make</command>.
4227 <para> Now set the following user environment variables:
4230 <listitem><para> Add <filename>c:/cygwin/bin</filename> and <filename>c:/cygwin/usr/bin</filename> to your
4231 <constant>PATH</constant></para></listitem>
4235 Set <constant>MAKE_MODE</constant> to <Literal>UNIX</Literal>. If you
4236 don't do this you get very weird messages when you type
4237 <Command>make</Command>, such as:
4239 /c: /c: No such file or directory
4244 <listitem><para> Set <constant>SHELL</constant> to
4245 <Filename>c:/cygwin/bin/sh</Filename>. When you invoke a shell in Emacs, this
4246 <constant>SHELL</constant> is what you get.
4249 <listitem><para> Set <constant>HOME</constant> to point to your
4250 home directory. This is where, for example,
4251 <command>bash</command> will look for your <filename>.bashrc</filename>
4252 file. Ditto <command>emacs</command> looking for <filename>.emacsrc</filename>
4258 There are a few other things to do:
4262 By default, cygwin provides the command shell <filename>ash</filename>
4263 as <filename>sh.exe</filename>. We have often seen build-system problems that
4264 turn out to be due to bugs in <filename>ash</filename>
4266 and length of command lines). On the other hand <filename>bash</filename> seems
4268 So, in <filename>cygwin/bin</filename>
4269 remove the supplied <filename>sh.exe</filename> (or rename it as <filename>ash.exe</filename>),
4270 and copy <filename>bash.exe</filename> to <filename>sh.exe</filename>.
4271 You'll need to do this in Windows Explorer or the Windows <command>cmd</command> shell, because
4272 you can't rename a running program!
4278 Some script files used in the make system start with "<Command>#!/bin/perl</Command>",
4279 (and similarly for <Command>sh</Command>). Notice the hardwired path!
4280 So you need to ensure that your <Filename>/bin</Filename> directory has the following
4283 <listitem> <para><Command>sh</Command></para></listitem>
4284 <listitem> <para><Command>perl</Command></para></listitem>
4285 <listitem> <para><Command>cat</Command></para></listitem>
4287 All these come in Cygwin's <Filename>bin</Filename> directory, which you probably have
4288 installed as <Filename>c:/cygwin/bin</Filename>. By default Cygwin mounts "<Filename>/</Filename>" as
4289 <Filename>c:/cygwin</Filename>, so if you just take the defaults it'll all work ok.
4290 (You can discover where your Cygwin
4291 root directory <Filename>/</Filename> is by typing <Command>mount</Command>.)
4292 Provided <Filename>/bin</Filename> points to the Cygwin <Filename>bin</Filename>
4293 directory, there's no need to copy anything. If not, copy these binaries from the <filename>cygwin/bin</filename>
4294 directory (after fixing the <filename>sh.exe</filename> stuff mentioned in the previous bullet).
4300 <para>Finally, here are some things to be aware of when using Cygwin:
4302 <listitem> <para>Cygwin doesn't deal well with filenames that include
4303 spaces. "<filename>Program Files</filename>" and "<filename>Local files</filename>" are
4307 <listitem> <para> Cygwin implements a symbolic link as a text file with some
4308 magical text in it. So other programs that don't use Cygwin's
4309 I/O libraries won't recognise such files as symlinks.
4310 In particular, programs compiled by GHC are meant to be runnable
4311 without having Cygwin, so they don't use the Cygwin library, so
4312 they don't recognise symlinks.
4316 Win32 has a <command>find</command> command which is not the same as Cygwin's find.
4317 You will probably discover that the Win32 <command>find</command> appears in your <constant>PATH</constant>
4318 before the Cygwin one, because it's in the <emphasis>system</emphasis> <constant>PATH</constant>
4319 environment variable, whereas you have probably modified the <emphasis>user</emphasis> <constant>PATH</constant>
4320 variable. You can always invoke <command>find</command> with an absolute path, or rename it.
4327 <Sect2 id="configure-ssh"><Title>Configuring SSH</Title>
4329 <para><command>ssh</command> comes with Cygwin, provided you remember to ask for it when
4330 you install Cygwin. (If not, the installer lets you update easily.) Look for <command>openssh</command>
4331 (not ssh) in the Cygwin list of applications!</para>
4333 <para>There are several strange things about <command>ssh</command> on Windows that you need to know.
4337 The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
4338 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
4339 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
4340 and run it as follows:
4342 c:\tmp> set CYGWIN32=tty
4343 c:\tmp> c:/user/local/bin/ssh-keygen1
4348 <command>ssh</command> needs to access your directory <filename>.ssh</filename>, in your home directory.
4349 To determine your home directory <command>ssh</command> first looks in
4350 <filename>c:/cygwin/etc/passwd</filename> (or wherever you have Cygwin installed). If there's an entry
4351 there with your userid, it'll use that entry to determine your home directory, <emphasis>ignoring
4352 the setting of the environment variable $HOME</emphasis>. If the home directory is
4353 bogus, <command>ssh</command> fails horribly. The best way to see what is going on is to say
4355 ssh -v cvs.haskell.org
4357 which makes <command>ssh</command> print out information about its activity.
4359 <para> You can fix this problem, either by correcting the home-directory field in
4360 <filename>c:/cygwin/etc/passwd</filename>, or by simply deleting the entire entry for your userid. If
4361 you do that, <command>ssh</command> uses the $HOME environment variable instead.
4367 <para>To protect your
4368 <literal>.ssh</literal> from access by anyone else,
4369 right-click your <literal>.ssh</literal> directory, and
4370 select <literal>Properties</literal>. If you are not on
4371 the access control list, add yourself, and give yourself
4372 full permissions (the second panel). Remove everyone else
4373 from the access control list. Don't leave them there but
4374 deny them access, because 'they' may be a list that
4375 includes you!</para>
4379 <para>In fact <command>ssh</command> 3.6.1 now seems to <emphasis>require</emphasis>
4380 you to have Unix permissions 600 (read/write for owner only)
4381 on the <literal>.ssh/identity</literal> file, else it
4382 bombs out. For your local C drive, it seems that <literal>chmod 600 identity</literal> works,
4383 but on Windows NT/XP, it doesn't work on a network drive (exact dteails obscure).
4384 The solution seems to be to set the $CYGWIN environment
4385 variable to "<literal>ntsec neta</literal>". The $CYGWIN environment variable is discussed
4386 in <ulink url="http://cygwin.com/cygwin-ug-net/using-cygwinenv.html">the Cygwin User's Guide</ulink>,
4387 and there are more details in <ulink url="http://cygwin.com/faq/faq_4.html#SEC44">the Cygwin FAQ</ulink>.
4394 <Sect2><Title>Other things you need to install</Title>
4396 <para>You have to install the following other things to build GHC:
4400 Install an executable GHC, from <ulink url="http://www.haskell.org/ghc">http://www.haskell.org/ghc</ulink>.
4401 This is what you will use to compile GHC. Add it in your
4402 <constant>PATH</constant>: the installer tells you the path element
4403 you need to add upon completion.
4409 Install an executable Happy, from <ulink url="http://www.haskell.org/happy">http://www.haskell.org/happy</ulink>.
4410 Happy is a parser generator used to compile the Haskell grammar. Add it in your
4411 <constant>PATH</constant>.
4417 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
4418 generate code, so you have to install that (see <xref linkend="cygwin-and-mingw">).
4419 Just pick up a mingw bundle at
4420 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
4421 We install it in <filename>c:/mingw</filename>.
4423 <para>Do <emphasis>not</emphasis> add any of the <emphasis>mingw</emphasis> binaries to your path.
4424 They are only going to get used by explicit access (via the --with-gcc flag you
4425 give to <Command>configure</Command> later). If you do add them to your path
4426 you are likely to get into a mess because their names overlap with Cygwin binaries.
4432 <para>We use <command>emacs</command> a lot, so we install that too.
4433 When you are in <filename>fptools/ghc/compiler</filename>, you can use
4434 "<literal>make tags</literal>" to make a TAGS file for emacs. That uses the utility
4435 <filename>fptools/ghc/utils/hasktags/hasktags</filename>, so you need to make that first.
4436 The most convenient way to do this is by going <literal>make boot</literal> in <filename>fptools/ghc</filename>.
4437 The <literal>make tags</literal> command also uses <command>etags</command>, which comes with <command>emacs</command>,
4438 so you will need to add <filename>emacs/bin</filename> to your <literal>PATH</literal>.
4444 <para> Finally, check out a copy of GHC sources from
4445 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
4452 <Sect2><Title>Building GHC</Title>
4455 Now go read the documentation above on building from source (<xref linkend="sec-building-from-source">);
4456 the bullets below only tell
4457 you about Windows-specific wrinkles.</para>
4461 Run <Command>autoconf</Command> both in <filename>fptools</filename>
4462 and in <filename>fptools/ghc</filename>. If you omit the latter step you'll
4463 get an error when you run <filename>./configure</filename>:
4466 creating mk/config.h
4467 mk/config.h is unchanged
4469 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
4470 ./configure: ./configure: No such file or directory
4471 configure: error: ./configure failed for ghc
4476 <listitem> <para><command>autoconf</command> seems to create the file <filename>configure</filename>
4477 read-only. So if you need to run autoconf again (which I sometimes do for safety's sake),
4480 /usr/bin/autoconf: cannot create configure: permission denied
4482 Solution: delete <filename>configure</filename> first.
4487 You either need to add <filename>ghc</filename> to your
4488 <constant>PATH</constant> before you invoke
4489 <Command>configure</Command>, or use the <Command>configure</Command>
4490 option <option>--with-ghc=c:/ghc/ghc-some-version/bin/ghc</option>.
4495 If you are paranoid, delete <filename>config.cache</filename> if it exists.
4496 This file occasionally remembers out-of-date configuration information, which
4497 can be really confusing.
4503 After <command>autoconf</command> run <command>./configure</command> in
4504 <filename>fptools/</filename> thus:
4507 ./configure --host=i386-unknown-mingw32 --with-gcc=c:/mingw/bin/gcc
4509 This is the point at which you specify that you are building GHC-mingw
4510 (see <xref linkend="ghc-mingw">). </para>
4512 <para> Both these options are important! It's possible to get into
4513 trouble using the wrong C compiler!</para>
4515 Furthermore, it's <emphasis>very important</emphasis> that you specify a
4516 full MinGW path for <command>gcc</command>, not a Cygwin path, because GHC (which
4517 uses this path to invoke <command>gcc</command>) is a MinGW program and won't
4518 understand a Cygwin path. For example, if you
4519 say <literal>--with-gcc=/mingw/bin/gcc</literal>, it'll be interpreted as
4520 <filename>/cygdrive/c/mingw/bin/gcc</filename>, and GHC will fail the first
4521 time it tries to invoke it. Worse, the failure comes with
4522 no error message whatsoever. GHC simply fails silently when first invoked,
4523 typically leaving you with this:
4525 make[4]: Leaving directory `/cygdrive/e/fptools-stage1/ghc/rts/gmp'
4526 ../../ghc/compiler/ghc-inplace -optc-mno-cygwin -optc-O
4527 -optc-Wall -optc-W -optc-Wstrict-prototypes -optc-Wmissing-prototypes
4528 -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
4529 -optc-Wbad-function-cast -optc-Wcast-align -optc-I../includes
4530 -optc-I. -optc-Iparallel -optc-DCOMPILING_RTS
4531 -optc-fomit-frame-pointer -O2 -static
4532 -package-name rts -O -dcore-lint -c Adjustor.c -o Adjustor.o
4533 make[2]: *** [Adjustor.o] Error 1
4534 make[1]: *** [all] Error 1
4535 make[1]: Leaving directory `/cygdrive/e/fptools-stage1/ghc'
4536 make: *** [all] Error 1
4542 If you want to build GHC-cygwin (<xref linkend="ghc-cygwin">)
4543 you'll have to do something more like:
4545 ./configure --with-gcc=...the Cygwin gcc...
4550 <listitem><para> You almost certainly want to set
4554 in your <filename>build.mk</filename> configuration file (see <xref linkend="sec-build-config">).
4555 This tells the build system not to split each library into a myriad of little object files, one
4556 for each function. Doing so reduces binary sizes for statically-linked binaries, but on Windows
4557 it dramatically increases the time taken to build the libraries in the first place.
4561 <listitem><para> Do not attempt to build the documentation.
4562 It needs all kinds of wierd Jade stuff that we haven't worked out for
4563 Win32.</para></listitem>