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 Glasgow <Literal>fptools</Literal> suite</title>
35 <para>Building the Glasgow tools <Emphasis>can</Emphasis> be
36 complicated, mostly because there are so many permutations of
37 what/why/how, e.g., ``Build Happy with HBC, everything else with
38 GHC, leave out profiling, and test it all on the `real' NoFib
39 programs.'' Yeeps!</para>
41 <para>Happily, such complications don't apply to most people. A
42 few common ``strategies'' serve most purposes. Pick one and
43 proceed as suggested:</para>
48 <term><indexterm><primary>Binary distribution</primary></indexterm>Binary distribution.</term>
51 If your only purpose is to install some of the
52 <Literal>fptools</Literal> suite then the easiest thing to do is to
53 get a binary distribution. In the binary distribution everything is
54 pre-compiled for your particular machine architecture and operating
55 system, so all you should have to do is install the binaries and
56 libraries in suitable places. The user guide describes how to do this.
60 A binary distribution may not work for you for two reasons. First, we
61 may not have built the suite for the particular architecture/OS
62 platform you want. That may be due to lack of time and energy (in
63 which case you can get a source distribution and build from it; see
64 below). Alternatively, it may be because we haven't yet ported the
65 suite to your architecture, in which case you are considerably worse
70 The second reason a binary distribution may not be what you want is
71 if you want to read or modify the souce code.
73 </listitem></VarListEntry>
75 <term><indexterm><primary>Source distribution</primary></indexterm>Source distribution.</term>
79 platform, but (a) you like the warm fuzzy feeling of compiling things
80 yourself; (b) you want to build something ``extra''—e.g., a set of
81 libraries with strictness-analysis turned off; or (c) you want to hack
86 A source distribution contains complete sources for one or more
87 projects in the <Literal>fptools</Literal> suite. Not only that, but
88 the more awkward machine-independent steps are done for you. For
89 example, if you don't have
90 <Command>happy</Command><indexterm><primary>happy</primary></indexterm>
91 you'll find it convenient that the source distribution contains the
92 result of running <Command>happy</Command> on the parser
93 specifications. If you don't want to alter the parser then this saves
94 you having to find and install <Command>happy</Command>. You will
95 still need a working version of GHC (preferably version 4.08+) on your
96 machine in order to compile (most of) the sources, however.
99 </listitem></VarListEntry>
102 <term>The CVS repository.</term>
103 <indexterm><primary>CVS repository</primary>
106 <para>We make releases infrequently. If you want more
107 up-to-the minute (but less tested) source code then you need
108 to get access to our CVS repository.</para>
110 <para>All the <Literal>fptools</Literal> source code is held
111 in a CVS repository. CVS is a pretty good source-code
112 control system, and best of all it works over the
115 <para>The repository holds source code only. It holds no
116 mechanically generated files at all. So if you check out a
117 source tree from CVS you will need to install every utility
118 so that you can build all the derived files from
121 <para>More information about our CVS repository can be found
122 in <xref linkend="sec-cvs">.</para>
128 <term>Build GHC from intermediate C <Filename>.hc</Filename> files<indexterm><primary>hc files</primary></indexterm>:</term>
130 <para>You need a working GHC to use a source distribution.
131 What if you don't have a working GHC? Then you may be able
132 to bootstrap up from the intermediate C
133 (<filename>.hc</filename>) files that we provide. Building
134 GHC on an unsupported platform falls into this category.
135 Beware: this route is not for the faint hearted! Please see
136 <Xref LinkEnd="sec-booting-from-C">.</para>
138 <para>Once you have built GHC, you can build the other
139 Glasgow tools with it.</para>
141 <para>In theory, you can (could?) build GHC with another
142 Haskell compiler (e.g., HBC). We haven't tried to do this
143 for ages and it almost certainly doesn't work any more (for
144 tedious reasons).</para>
149 <para>If you are going to do any building from sources (either
150 from a source distribution or the CVS repository) then you need to
151 read all of this manual in detail.</para>
155 <title>Using the CVS repository</title>
157 <para>We use <ulink url="http://www.cvshome.org/">CVS</ulink> (Concurrent Version System) to keep track of our
158 sources for various software projects. CVS lets several people
159 work on the same software at the same time, allowing changes to be
160 checked in incrementally. </para>
162 <para>This section is a set of guidelines for how to use our CVS
163 repository, and will probably evolve in time. The main thing to
164 remember is that most mistakes can be undone, but if there's
165 anything you're not sure about feel free to bug the local CVS
166 meister (namely Jeff Lewis
167 <email>jlewis@galconn.com</email>). </para>
169 <sect2 id="cvs-access">
170 <title>Getting access to the CVS Repository</title>
172 <para>You can access the repository in one of two ways:
173 read-only (<xref linkend="cvs-read-only">), or read-write (<xref
174 linkend="cvs-read-write">).</para>
176 <sect3 id="cvs-read-only">
177 <title>Remote Read-only CVS Access</title>
179 <para>Read-only access is available to anyone - there's no
180 need to ask us first. With read-only CVS access you can do
181 anything except commit changes to the repository. You can
182 make changes to your local tree, and still use CVS's merge
183 facility to keep your tree up to date, and you can generate
184 patches using 'cvs diff' in order to send to us for
187 <para>To get read-only access to the repository:</para>
191 <para>Make sure that <application>cvs</application> is
192 installed on your machine.</para>
195 <para>Set your <literal>$CVSROOT</literal> environment variable to
196 <literal>:pserver:anoncvs@glass.cse.ogi.edu:/cvs</literal></para>
199 <para>Run the command</para>
203 <para>The password is simply <literal>cvs</literal>. This
204 sets up a file in your home directory called
205 <literal>.cvspass</literal>, which squirrels away the
206 dummy password, so you only need to do this step once.</para>
210 <para>Now go to <xref linkend="cvs-first">.</para>
215 <sect3 id="cvs-read-write">
216 <title>Remote Read-Write CVS Access</title>
218 <para>We generally supply read-write access to folk doing
219 serious development on some part of the source tree, when
220 going through us would be a pain. If you're developing some
221 feature, or think you have the time and inclination to fix
222 bugs in our sources, feel free to ask for read-write
223 access. There is a certain amount of responsibility that goes
224 with commit privileges; we are more likely to grant you access
225 if you've demonstrated your competence by sending us patches
226 via mail in the past.</para>
228 <para>To get remote read-write CVS access, you need to do the
229 following steps.</para>
233 <para>Make sure that <literal>cvs</literal> and
234 <literal>ssh</literal> are both installed on your
239 <para>Generate a DSA private-key/public-key pair, thus:</para>
243 <para>(<literal>ssh-keygen</literal> comes with
244 <literal>ssh</literal>.) Running <literal>ssh-keygen
245 -d</literal> creates the private and public keys in
246 <literal>$HOME/.ssh/id_dsa</literal> and
247 <literal>$HOME/.ssh/id_dsa.pub</literal> respectively
248 (assuming you accept the standard defaults).</para>
250 <para><literal>ssh-keygen -d</literal> will only work if
251 you have Version 2 <literal>ssh</literal> installed; it
252 will fail harmlessly otherwise. If you only have Version
253 1 you can instead generate an RSA key pair using plain</para>
258 <para>Doing so creates the private and public RSA keys in
259 <literal>$HOME/.ssh/identity</literal> and
260 <literal>$HOME/.ssh/identity.pub</literal>
263 <para>[Deprecated.] Incidentally, you can force a Version
264 2 <literal>ssh</literal> to use the Version 1 protocol by
265 creating <literal>$HOME/config</literal> with the
266 following in it:</para>
274 <para>In both cases, <literal>ssh-keygen</literal> will
275 ask for a <firstterm>passphrase</firstterm>. The
276 passphrase is a password that protects your private key.
277 In response to the 'Enter passphrase' question, you can
281 <para>[Recommended.] Enter a passphrase, which you
282 will quote each time you use CVS.
283 <literal>ssh-agent</literal> makes this entirely
287 <para>[Deprecated.] Just hit return (i.e. use an empty
288 passphrase); then you won't need to quote the
289 passphrase when using CVS. The downside is that
290 anyone who can see into your <literal>.ssh</literal>
291 directory, and thereby get your private key, can mess
292 up the repository. So you must keep the
293 <literal>.ssh</literal> directory with draconian
294 no-access permissions.</para>
300 [Windows users.] The programs <command>ssh-keygen1</command>, <command>ssh1</command>, and <command>cvs</command>,
301 seem to lock up <command>bash</command> entirely if they try to get user input (e.g. if
302 they ask for a password). To solve this, start up <filename>cmd.exe</filename>
303 and run it as follows:
305 c:\tmp> set CYGWIN32=tty
306 c:\tmp> c:/user/local/bin/ssh-keygen1
309 <para>[Windows users.] To protect your
310 <literal>.ssh</literal> from access by anyone else,
311 right-click your <literal>.ssh</literal> directory, and
312 select <literal>Properties</literal>. If you are not on
313 the access control list, add yourself, and give yourself
314 full permissions (the second panel). Remove everyone else
315 from the access control list. Don't leave them there but
316 deny them access, because 'they' may be a list that
321 <para>Send a message to to the CVS repository
322 administrator (currently Jeff Lewis
323 <email>jeff@galconn.com</email>), containing:</para>
326 <para>Your desired user-name.</para>
329 <para>Your <literal>.ssh/id_dsa.pub</literal> (or
330 <literal>.ssh/identity.pub</literal>).</para>
333 <para>He will set up your account.</para>
337 <para>Set the following environment variables:</para>
341 <constant>$HOME</constant>: points to your home directory. This is where CVS
342 will look for its <filename>.cvsrc</filename> file.
348 <constant>$CVS_RSH</constant> to <filename>ssh</filename>
350 <para>[Windows users.] Setting your <literal>CVS_RSH</literal> to
351 <literal>ssh</literal> assumes that your CVS client
352 understands how to execute shell script
353 ("#!"s,really), which is what
354 <literal>ssh</literal> is. This may not be the case on
355 Win32 platforms, so in that case set <literal>CVS_RSH</literal> to
356 <literal>ssh1</literal>.</para>
360 <para><literal>$CVSROOT</literal> to
361 <literal>:ext:</literal><replaceable>your-username</replaceable>
362 <literal>@cvs.haskell.org:/home/cvs/root</literal>
363 where <replaceable>your-username</replaceable> is your user name on
364 <literal>cvs.haskell.org</literal>.
366 <para>The <literal>CVSROOT</literal> environment variable will
367 be recorded in the checked-out tree, so you don't need to set
368 this every time. </para>
374 <constant>$CVSEDITOR</constant>: <filename>bin/gnuclient.exe</filename>
375 if you want to use an Emacs buffer for typing in those long commit messages.
381 <constant>$SHELL</constant>: To use bash as the shell in Emacs, you need to
382 set this to point to <Filename>bash.exe</Filename>.
393 Put the following in <filename>$HOME/.cvsrc</filename>:
404 These are the default options for the specified CVS commands,
405 and represent better defaults than the usual ones. (Feel
406 free to change them.)
410 [Windows users.] Filenames starting with <filename>.</filename> were illegal in
411 the 8.3 DOS filesystem, but that restriction should have
412 been lifted by now (i.e., you're using VFAT or later filesystems.) If
413 you're still having problems creating it, don't worry; <filename>.cvsrc</filename> is entirely
421 <para>[Experts.] Once your account is set up, you can get
422 access from other machines without bothering Jeff, thus:</para>
425 <para>Generate a public/private key pair on the new
429 <para>Use ssh to log in to
430 <literal>cvs.haskell.org</literal>, from your old
434 <para>Add the public key for the new machine to the file
435 <literal>$HOME/ssh/authorized_keys</literal> on
436 <literal>cvs.haskell.org</literal>.
437 (<literal>authorized_keys2</literal>, I think, for Version
441 <para>Make sure that the new version of
442 <literal>authorized_keys</literal> still has 600 file
451 <sect2 id="cvs-first">
452 <title>Checking Out a Source Tree</title>
456 <para>Make sure you set your <literal>CVSROOT</literal>
457 environment variable according to either of the remote
458 methods above. The Approved Way to check out a source tree
459 is as follows:</para>
462 $ cvs checkout fpconfig
465 <para>At this point you have a new directory called
466 <literal>fptools</literal> which contains the basic stuff
467 for the fptools suite, including the configuration files and
468 some other junk. </para>
470 <para>[Windows users.] The following messages appear to be harmless:
472 setsockopt IPTOS_LOWDELAY: Invalid argument
473 setsockopt IPTOS_THROUGHPUT: Invalid argument
478 <para>You can call the fptools directory whatever you like,
479 CVS won't mind: </para>
482 $ mv fptools <replaceable>directory</replaceable>
485 <para> NB: after you've read the CVS manual you might be
486 tempted to try</para>
488 $ cvs checkout -d <replaceable>directory</replaceable> fpconfig
491 <para>instead of checking out <literal>fpconfig</literal>
492 and then renaming it. But this doesn't work, and will
493 result in checking out the entire repository instead of just
494 the <literal>fpconfig</literal> bit.</para>
496 $ cd <replaceable>directory</replaceable>
497 $ cvs checkout ghc hslibs
500 <para>The second command here checks out the relevant
501 modules you want to work on. For a GHC build, for instance,
502 you need at least the <literal>ghc</literal> and
503 <literal>hslibs</literal> modules (for a full list of the
504 projects available, see <xref linkend="projects">).</para>
509 <sect2 id="cvs-committing">
510 <title>Committing Changes</title>
512 <para>This is only if you have read-write access to the
513 repository. For anoncvs users, CVS will issue a "read-only
514 repository" error if you try to commit changes.</para>
518 <para>Build the software, if necessary. Unless you're just
519 working on documentation, you'll probably want to build the
520 software in order to test any changes you make.</para>
524 <para>Make changes. Preferably small ones first.</para>
528 <para>Test them. You can see exactly what changes you've
529 made by using the <literal>cvs diff</literal> command:</para>
533 <para>lists all the changes (using the
534 <literal>diff</literal> command) in and below the current
535 directory. In emacs, <literal>C-c C-v =</literal> runs
536 <literal>cvs diff</literal> on the current buffer and shows
537 you the results.</para>
541 <para>Before checking in a change, you need to update your
548 <para>This pulls in any changes that other people have made,
549 and merges them with yours. If there are any conflicts, CVS
550 will tell you, and you'll have to resolve them before you
551 can check your changes in. The documentation describes what
552 to do in the event of a conflict.</para>
554 <para>It's not always necessary to do a full cvs update
555 before checking in a change, since CVS will always tell you
556 if you try to check in a file that someone else has changed.
557 However, you should still update at regular intervals to
558 avoid making changes that don't work in conjuction with
559 changes that someone else made. Keeping an eye on what goes
560 by on the mailing list can help here.</para>
564 <para>When you're happy that your change isn't going to
565 break anything, check it in. For a one-file change:</para>
568 $ cvs commit <replaceable>filename</replaceable>
571 <para>CVS will then pop up an editor for you to enter a
572 "commit message", this is just a short description
573 of what your change does, and will be kept in the history of
576 <para>If you're using emacs, simply load up the file into a
577 buffer and type <literal>C-x C-q</literal>, and emacs will
578 prompt for a commit message and then check in the file for
581 <para>For a multiple-file change, things are a bit
582 trickier. There are several ways to do this, but this is the
583 way I find easiest. First type the commit message into a
584 temporary file. Then either</para>
587 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>file_1</replaceable> .... <replaceable>file_n</replaceable>
590 <para>or, if nothing else has changed in this part of the
594 $ cvs commit -F <replaceable>commit-message</replaceable> <replaceable>directory</replaceable>
597 <para>where <replaceable>directory</replaceable> is a common
598 parent directory for all your changes, and
599 <replaceable>commit-message</replaceable> is the name of the
600 file containing the commit message.</para>
602 <para>Shortly afterwards, you'll get some mail from the
603 relevant mailing list saying which files changed, and giving
604 the commit message. For a multiple-file change, you should
605 still get only <emphasis>one</emphasis> message.</para>
610 <sect2 id="cvs-update">
611 <title>Updating Your Source Tree</title>
613 <para>It can be tempting to cvs update just part of a source
614 tree to bring in some changes that someone else has made, or
615 before committing your own changes. This is NOT RECOMMENDED!
616 Quite often changes in one part of the tree are dependent on
617 changes in another part of the tree (the
618 <literal>mk/*.mk</literal> files are a good example where
619 problems crop up quite often). Having an inconsistent tree is a
620 major cause of headaches. </para>
622 <para>So, to avoid a lot of hassle, follow this recipe for
623 updating your tree: </para>
627 $ cvs update -Pd 2>&1 | tee log</screen>
629 <para>Look at the log file, and fix any conflicts (denoted by a
630 <quote>C</quote> in the first column). If you're using multiple
631 build trees, then for every build tree you have pointing at this
632 source tree, you need to update the links in case any new files
633 have appeared: </para>
636 $ cd <replaceable>build-tree</replaceable>
637 $ lndir <replaceable>source-tree</replaceable>
640 <para>Some files might have been removed, so you need to remove
641 the links pointing to these non-existent files:</para>
644 $ find . -xtype l -exec rm '{}' \;
647 <para>To be <emphasis>really</emphasis> safe, you should do
650 <screen>$ gmake all</screen>
652 <para>from the top-level, to update the dependencies and build
653 any changed files. </para>
656 <sect2 id="cvs-tags">
657 <title>GHC Tag Policy</title>
659 <para>If you want to check out a particular version of GHC,
660 you'll need to know how we tag versions in the repository. The
661 policy (as of 4.04) is:</para>
665 <para>The tree is branched before every major release. The
666 branch tag is <literal>ghc-x-xx-branch</literal>, where
667 <literal>x-xx</literal> is the version number of the release
668 with the <literal>'.'</literal> replaced by a
669 <literal>'-'</literal>. For example, the 4.04 release lives
670 on <literal>ghc-4-04-branch</literal>.</para>
674 <para>The release itself is tagged with
675 <literal>ghc-x-xx</literal> (on the branch). eg. 4.06 is
676 called <literal>ghc-4-06</literal>.</para>
680 <para>We didn't always follow these guidelines, so to see
681 what tags there are for previous versions, do <literal>cvs
682 log</literal> on a file that's been around for a while (like
683 <literal>fptools/ghc/README</literal>).</para>
687 <para>So, to check out a fresh GHC 4.06 tree you would
691 $ cvs co -r ghc-4-06 fpconfig
693 $ cvs co -r ghc-4-06 ghc hslibs
697 <sect2 id="cvs-hints">
698 <title>General Hints</title>
702 <para>As a general rule: commit changes in small units,
703 preferably addressing one issue or implementing a single
704 feature. Provide a descriptive log message so that the
705 repository records exactly which changes were required to
706 implement a given feature/fix a bug. I've found this
707 <emphasis>very</emphasis> useful in the past for finding out
708 when a particular bug was introduced: you can just wind back
709 the CVS tree until the bug disappears.</para>
713 <para>Keep the sources at least *buildable* at any given
714 time. No doubt bugs will creep in, but it's quite easy to
715 ensure that any change made at least leaves the tree in a
716 buildable state. We do nightly builds of GHC to keep an eye
717 on what things work/don't work each day and how we're doing
718 in relation to previous verions. This idea is truely wrecked
719 if the compiler won't build in the first place!</para>
723 <para>To check out extra bits into an already-checked-out
724 tree, use the following procedure. Suppose you have a
725 checked-out fptools tree containing just ghc, and you want
726 to add nofib to it:</para>
737 $ cvs update -d nofib
740 <para>(the -d flag tells update to create a new
741 directory). If you just want part of the nofib suite, you
746 $ cvs checkout nofib/spectral
749 <para>This works because <literal>nofib</literal> is a
750 module in its own right, and spectral is a subdirectory of
751 the nofib module. The path argument to checkout must always
752 start with a module name. There's no equivalent form of this
753 command using <literal>update</literal>.</para>
759 <sect1 id="projects">
760 <title>What projects are there?</title>
762 <para>The <literal>fptools</literal> suite consists of several
763 <firstterm>projects</firstterm>, most of which can be downloaded,
764 built and installed individually. Each project corresponds to a
765 subdirectory in the source tree, and if checking out from CVS then
766 each project can be checked out individually by sitting in the top
767 level of your source tree and typing <command>cvs checkout
768 <replaceable>project</replaceable></command>.</para>
770 <para>Here is a list of the projects currently available:</para>
774 <term><literal>ghc</literal></term>
775 <indexterm><primary><literal>ghc</literal></primary>
776 <secondary>project</secondary></indexterm>
778 <para>The <ulink url="http://www.haskell.org/ghc/">Glasgow
779 Haskell Compiler</ulink> (minus libraries). Absolutely
780 required for building GHC.</para>
785 <term><literal>glafp-utils</literal></term>
786 <indexterm><primary><literal>glafp-utils</literal></primary><secondary>project</secondary></indexterm>
788 <para>Utility programs, some of which are used by the
789 build/installation system. Required for pretty much
795 <term><literal>green-card</literal></term>
796 <indexterm><primary><literal>green-card</literal></primary><secondary>project</secondary></indexterm>
799 url="http://www.haskell.org/greencard/">Green Card</ulink>
800 system for generating Haskell foreign function
806 <term><literal>haggis</literal></term>
807 <indexterm><primary><literal>haggis</literal></primary><secondary>project</secondary></indexterm>
810 url="http://www.dcs.gla.ac.uk/fp/software/haggis/">Haggis</ulink>
811 Haskell GUI framework.</para>
816 <term><literal>happy</literal></term>
817 <indexterm><primary><literal>happy</literal></primary><secondary>project</secondary></indexterm>
820 url="http://www.haskell.org/happy/">Happy</ulink> Parser
826 <term><literal>hdirect</literal></term>
827 <indexterm><primary><literal>hdirect</literal></primary><secondary>project</secondary></indexterm>
830 url="http://www.haskell.org/hdirect/">H/Direct</ulink>
831 Haskell interoperability tool.</para>
836 <term><literal>hood</literal></term>
837 <indexterm><primary><literal>hood</literal></primary><secondary>project</secondary></indexterm>
839 <para>The <ulink url="http://www.haskell.org/hood/">Haskell
840 Object Observation Debugger</ulink>.</para>
845 <term><literal>hslibs</literal></term>
846 <indexterm><primary><literal>hslibs</literal></primary><secondary>project</secondary></indexterm>
848 <para>GHC's libraries. Required for building GHC.</para>
853 <term><literal>libraries</literal></term>
854 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
856 <para>Hierarchical Haskell library suite
857 (experimental).</para>
862 <term><literal>mhms</literal></term>
863 <indexterm><primary><literal></literal></primary><secondary>project</secondary></indexterm>
865 <para>The Modular Haskell Metric System.</para>
870 <term><literal>nofib</literal></term>
871 <indexterm><primary><literal>nofib</literal></primary><secondary>project</secondary></indexterm>
873 <para>The NoFib suite: A collection of Haskell programs used
874 primarily for benchmarking.</para>
879 <term><literal>testsuite</literal></term>
880 <indexterm><primary><literal>testsuite</literal></primary><secondary>project</secondary></indexterm>
882 <para>A testing framework, including GHC's regression test
888 <para>So, to build GHC you need at least the
889 <literal>ghc</literal> and <literal>hslibs</literal> projects (a
890 GHC source distribution will already include the bits you
894 <sect1 id="sec-build-checks">
895 <title>Things to check before you start</title>
897 <para>Here's a list of things to check before you get
903 <indexterm><primary>Disk space needed</primary></indexterm>
904 <para>Disk space needed: from about 100Mb for a basic GHC
905 build, up to probably 500Mb for a GHC build with everything
906 included (libraries built several different ways,
911 <para>Use an appropriate machine, compilers, and things.
912 SPARC boxes, PCs running Linux or FreeBSD, and Alphas running
913 OSF/1 are all fully supported. Win32 and HP boxes are in
914 pretty good shape. PCs running Solaris, DEC Alphas running
915 Linux or some BSD variant, MIPS and AIX boxes will need some
916 minimal porting effort before they work (as of 4.06). <xref
917 linkend="sec-port-info"> gives the full run-down on ports or
922 <para>Be sure that the ``pre-supposed'' utilities are
923 installed. <Xref LinkEnd="sec-pre-supposed">
928 <para>If you have any problem when building or installing the
929 Glasgow tools, please check the ``known pitfalls'' (<Xref
930 LinkEnd="sec-build-pitfalls">). Also check the FAQ for the
931 version you're building, which should be available from the
932 relevant download page on the <ULink
933 URL="http://www.haskell.org/ghc/" >GHC web
936 <indexterm><primary>known bugs</primary></indexterm>
937 <indexterm><primary>bugs, known</primary></indexterm>
939 <para>If you feel there is still some shortcoming in our
940 procedure or instructions, please report it.</para>
942 <para>For GHC, please see the bug-reporting section of the GHC
943 Users' Guide (separate document), to maximise the usefulness
944 of your report.</para>
946 <indexterm><primary>bugs</primary><secondary>seporting</secondary></indexterm>
948 <para>If in doubt, please send a message to
949 <email>glasgow-haskell-bugs@haskell.org</email>.
950 <indexterm><primary>bugs</primary><secondary>mailing
951 list</secondary></indexterm></para>
957 <Sect1 id="sec-port-info">
958 <Title>What machines the Glasgow tools run on
962 <indexterm><primary>ports, GHC</primary></indexterm>
963 <indexterm><primary>GHC ports</primary></indexterm>
964 <indexterm><primary>supported platforms</primary></indexterm>
965 <indexterm><primary>platforms, supported</primary></indexterm>
966 The main question is whether or not the Haskell compiler (GHC) runs on
971 A ``platform'' is a architecture/manufacturer/operating-system
972 combination, such as <Literal>sparc-sun-solaris2</Literal>. Other common ones are
973 <Literal>alpha-dec-osf2</Literal>, <Literal>hppa1.1-hp-hpux9</Literal>, <Literal>i386-unknown-linux</Literal>,
974 <Literal>i386-unknown-solaris2</Literal>, <Literal>i386-unknown-freebsd</Literal>,
975 <Literal>i386-unknown-cygwin32</Literal>, <Literal>m68k-sun-sunos4</Literal>, <Literal>mips-sgi-irix5</Literal>,
976 <Literal>sparc-sun-sunos4</Literal>, <Literal>sparc-sun-solaris2</Literal>, <Literal>powerpc-ibm-aix</Literal>.
980 Bear in mind that certain ``bundles'', e.g. parallel Haskell, may not
981 work on all machines for which basic Haskell compiling is supported.
985 Some libraries may only work on a limited number of platforms; for
986 example, a sockets library is of no use unless the operating system
987 supports the underlying BSDisms.
991 <Title>What platforms the Haskell compiler (GHC) runs on</Title>
994 <indexterm><primary>fully-supported platforms</primary></indexterm>
995 <indexterm><primary>native-code generator</primary></indexterm>
996 <indexterm><primary>registerised ports</primary></indexterm>
997 <indexterm><primary>unregisterised ports</primary></indexterm>
998 The GHC hierarchy of Porting Goodness: (a) Best is a native-code
999 generator; (b) next best is a ``registerised''
1000 port; (c) the bare minimum is an ``unregisterised'' port.
1001 (``Unregisterised'' is so terrible that we won't say more about it).
1005 We use Sparcs running Solaris 2.7 and x86 boxes running FreeBSD and
1006 Linux, so those are the best supported platforms, unsurprisingly.
1010 Here's everything that's known about GHC ports. We identify platforms
1011 by their ``canonical'' CPU/Manufacturer/OS triple.
1017 <term>alpha-dec-{osf,linux,freebsd,openbsd,netbsd}:</term>
1018 <indexterm><primary>alpha-dec-osf</primary></indexterm>
1019 <indexterm><primary>alpha-dec-linux</primary></indexterm>
1020 <indexterm><primary>alpha-dec-freebsd</primary></indexterm>
1021 <indexterm><primary>alpha-dec-openbsd</primary></indexterm>
1022 <indexterm><primary>alpha-dec-netbsd</primary></indexterm>
1025 <para>The OSF port is currently working (as of GHC version
1026 5.02.1) and well supported. The native code generator is
1027 currently non-working. Other operating systems will
1028 require some minor porting.</para>
1033 <term>sparc-sun-sunos4</term>
1034 <indexterm><primary>sparc-sun-sunos4</primary></indexterm>
1036 <para>Probably works with minor tweaks, hasn't been tested
1042 <term>sparc-sun-solaris2</term>
1043 <indexterm><primary>sparc-sun-solaris2</primary></indexterm>
1045 <para>Fully supported, including native-code
1051 <term>hppa1.1-hp-hpux (HP-PA boxes running HPUX 9.x)</term>
1052 <indexterm><primary>hppa1.1-hp-hpux</primary></indexterm>
1054 <para>Works registerised. No native-code
1060 <term>i386-unknown-linux (PCs running Linux, ELF binary format)</term>
1061 <indexterm><primary>i386-*-linux</primary></indexterm>
1063 <para>GHC works registerised and has a native code
1064 generator. You <Emphasis>must</Emphasis> have GCC 2.7.x
1065 or later. NOTE about <literal>glibc</literal> versions:
1066 GHC binaries built on a system running <literal>glibc
1067 2.0</literal> won't work on a system running
1068 <literal>glibc 2.1</literal>, and vice versa. In general,
1069 don't expect compatibility between
1070 <literal>glibc</literal> versions, even if the shared
1071 library version hasn't changed.</para>
1076 <term>i386-unknown-freebsd (PCs running FreeBSD 2.2
1078 <indexterm><primary>i386-unknown-freebsd</primary></indexterm>
1080 <para>GHC works registerised. Pre-built packages are
1081 available in the native package format, so if you just
1082 need binaries you're better off just installing the
1088 <term>i386-unknown-{netbsd,openbsd) (PCs running NetBSD
1090 <indexterm><primary>i386-unknown-netbsd</primary></indexterm>
1091 <indexterm><primary>i386-unknown-openbsd</primary></indexterm>
1093 <para>Will require some minor porting effort, but should
1094 work registerised.</para>
1099 <term>i386-unknown-cygwin32:</term>
1100 <indexterm><primary>i386-unknown-cygwin32</primary></indexterm>
1102 <para>Fully supported under Win9x/NT, including a native
1103 code generator. Requires the <Literal>cygwin32</Literal>
1104 compatibility library and a healthy collection of GNU
1105 tools (i.e., gcc, GNU ld, bash etc.).</para>
1110 <term>mips-sgi-irix5</term>
1111 <indexterm><primary>mips-sgi-irix[5-6]</primary></indexterm>
1113 <para>Port currently doesn't work, needs some minimal
1114 porting effort. As usual, we don't have access to
1115 machines and there hasn't been an overwhelming demand for
1116 this port, but feel free to get in touch.</para>
1121 <term>powerpc-ibm-aix</term>
1122 <indexterm><primary>powerpc-ibm-aix</primary></indexterm>
1124 <para>Port currently doesn't work, needs some minimal
1125 porting effort. As usual, we don't have access to
1126 machines and there hasn't been an overwhelming demand for
1127 this port, but feel free to get in touch.</para>
1132 <term>powerpc-apple-darwin</term>
1133 <indexterm><primary>powerpc-apple-darwin</primary></indexterm>
1135 <para>Works, unregisterised only at the moment.</para>
1140 <para>Various other systems have had GHC ported to them in the
1141 distant past, including various Motorola 68k boxes. The 68k
1142 support still remains, but porting to one of these systems will
1143 certainly be a non-trivial task.</para>
1147 <title>What machines the other tools run on</title>
1149 <para>Unless you hear otherwise, the other tools work if GHC
1155 <Sect1 id="sec-pre-supposed">
1156 <Title>Installing pre-supposed utilities
1158 <indexterm><primary>pre-supposed utilities</primary></indexterm>
1159 <indexterm><primary>utilities, pre-supposed</primary></indexterm></Title>
1162 Here are the gory details about some utility programs you may need;
1163 <Command>perl</Command>, <Command>gcc</Command> and
1164 <command>happy</command> are the only important
1165 ones. (PVM<indexterm><primary>PVM</primary></indexterm> is important
1166 if you're going for Parallel Haskell.) The
1167 <Command>configure</Command><indexterm><primary>configure</primary></indexterm>
1168 script will tell you if you are missing something.
1176 <indexterm><primary>pre-supposed: Perl</primary></indexterm>
1177 <indexterm><primary>Perl, pre-supposed</primary></indexterm>
1180 <Emphasis>You have to have Perl to proceed!</Emphasis>
1181 It is pretty easy to install.
1185 Perl 5 is required. For Win32 platforms, you should use the binary
1186 supplied in the InstallShield (copy it to <filename>/bin</filename>).
1187 The Cygwin-supplied Perl seems not to work.
1191 Perl should be put somewhere so that it can be invoked by the
1192 <Literal>#!</Literal> script-invoking mechanism. The full
1193 pathname may need to be less than 32 characters long on some
1197 </listitem></VarListEntry>
1199 <term>GNU C (<Command>gcc</Command>):</term>
1200 <indexterm><primary>pre-supposed: GCC (GNU C compiler)</primary></indexterm>
1201 <indexterm><primary>GCC (GNU C compiler), pre-supposed</primary></indexterm>
1205 We recommend using GCC version 2.95.2 on all platforms. Failing that,
1206 version 2.7.2 is stable on most platforms. Earlier versions of GCC
1207 can be assumed not to work, and versions in between 2.7.2 and 2.95.2
1208 (including <command>egcs</command>) have varying degrees of stability
1209 depending on the platform.
1213 If your GCC dies with ``internal error'' on some GHC source file,
1214 please let us know, so we can report it and get things improved.
1215 (Exception: on iX86 boxes—you may need to fiddle with GHC's
1216 <Option>-monly-N-regs</Option> option; see the User's Guide)
1218 </listitem></VarListEntry>
1222 <indexterm><primary>Happy</primary></indexterm>
1224 <para>Happy is a parser generator tool for Haskell, and is used to
1225 generate GHC's parsers. Happy is written in Haskell, and is a project
1226 in the CVS repository (<literal>fptools/happy</literal>). It can be
1227 built from source, but bear in mind that you'll need GHC installed in
1228 order to build it. To avoid the chicken/egg problem, install a binary
1229 distribtion of either Happy or GHC to get started. Happy
1230 distributions are available from <ulink
1231 url="http://www.haskell.org/happy/">Happy's Web Page</ulink>.
1237 <term>Autoconf:</term>
1238 <indexterm><primary>pre-supposed: Autoconf</primary></indexterm>
1239 <indexterm><primary>Autoconf, pre-supposed</primary></indexterm>
1242 GNU Autoconf is needed if you intend to build from the CVS sources, it
1243 is <Emphasis>not</Emphasis> needed if you just intend to build a
1244 standard source distribution.
1248 Autoconf builds the <Command>configure</Command> script from
1249 <Filename>configure.in</Filename> and <Filename>aclocal.m4</Filename>.
1250 If you modify either of these files, you'll need
1251 <command>autoconf</command> to rebuild <Filename>configure</Filename>.
1254 </listitem></VarListEntry>
1256 <term><Command>sed</Command></term>
1257 <indexterm><primary>pre-supposed: sed</primary></indexterm>
1258 <indexterm><primary>sed, pre-supposed</primary></indexterm>
1261 You need a working <Command>sed</Command> if you are going to build
1262 from sources. The build-configuration stuff needs it. GNU sed
1263 version 2.0.4 is no good! It has a bug in it that is tickled by the
1264 build-configuration. 2.0.5 is OK. Others are probably OK too
1265 (assuming we don't create too elaborate configure scripts.)
1267 </listitem></VarListEntry>
1272 One <Literal>fptools</Literal> project is worth a quick note at this
1273 point, because it is useful for all the others:
1274 <Literal>glafp-utils</Literal> contains several utilities which aren't
1275 particularly Glasgow-ish, but Occasionally Indispensable. Like
1276 <Command>lndir</Command> for creating symbolic link trees.
1279 <Sect2 id="pre-supposed-gph-tools">
1280 <Title>Tools for building parallel GHC (GPH)
1287 <term>PVM version 3:</term>
1288 <indexterm><primary>pre-supposed: PVM3 (Parallel Virtual Machine)</primary></indexterm>
1289 <indexterm><primary>PVM3 (Parallel Virtual Machine), pre-supposed</primary></indexterm>
1293 PVM is the Parallel Virtual Machine on which Parallel Haskell programs
1294 run. (You only need this if you plan to run Parallel Haskell.
1295 Concurent Haskell, which runs concurrent threads on a uniprocessor
1296 doesn't need it.) Underneath PVM, you can have (for example) a
1297 network of workstations (slow) or a multiprocessor box (faster).
1301 The current version of PVM is 3.3.11; we use 3.3.7. It is readily
1302 available on the net; I think I got it from
1303 <Literal>research.att.com</Literal>, in <Filename>netlib</Filename>.
1307 A PVM installation is slightly quirky, but easy to do. Just follow
1308 the <Filename>Readme</Filename> instructions.
1310 </listitem></VarListEntry>
1312 <term><Command>bash</Command>:</term>
1313 <indexterm><primary>bash, presupposed (Parallel Haskell only)</primary></indexterm>
1316 Sadly, the <Command>gr2ps</Command> script, used to convert ``parallelism profiles''
1317 to PostScript, is written in Bash (GNU's Bourne Again shell).
1318 This bug will be fixed (someday).
1320 </listitem></VarListEntry>
1326 <Sect2 id="pre-supposed-doc-tools">
1327 <Title>Tools for building the Documentation
1331 The following additional tools are required if you want to format the
1332 documentation that comes with the <Literal>fptools</Literal> projects:
1339 <term>DocBook:</term>
1340 <indexterm><primary>pre-supposed: DocBook</primary></indexterm>
1341 <indexterm><primary>DocBook, pre-supposed</primary></indexterm>
1344 All our documentation is written in SGML, using the DocBook DTD.
1345 Instructions on installing and configuring the DocBook tools are in the
1346 installation guide (in the GHC user guide).
1349 </listitem></VarListEntry>
1352 <indexterm><primary>pre-supposed: TeX</primary></indexterm>
1353 <indexterm><primary>TeX, pre-supposed</primary></indexterm>
1356 A decent TeX distribution is required if you want to produce printable
1357 documentation. We recomment teTeX, which includes just about
1358 everything you need.
1360 </listitem></VarListEntry>
1365 In order to actually build any documentation, you need to set
1366 <constant>SGMLDocWays</constant> in your
1367 <filename>build.mk</filename>. Valid values to add to this
1368 list are: <literal>dvi</literal>, <literal>ps</literal>,
1369 <literal>pdf</literal>, <literal>html</literal>, and
1370 <literal>rtf</literal>.
1375 <Sect2 id="pre-supposed-other-tools">
1376 <Title>Other useful tools
1382 <indexterm><primary>pre-supposed: flex</primary></indexterm>
1383 <indexterm><primary>flex, pre-supposed</primary></indexterm>
1387 This is a quite-a-bit-better-than-Lex lexer. Used to build a couple
1388 of utilities in <Literal>glafp-utils</Literal>. Depending on your
1389 operating system, the supplied <Command>lex</Command> may or may not
1390 work; you should get the GNU version.
1392 </listitem></VarListEntry>
1399 <Sect1 id="sec-building-from-source">
1400 <Title>Building from source
1402 <indexterm><primary>Building from source</primary></indexterm>
1403 <indexterm><primary>Source, building from</primary></indexterm></Title>
1406 You've been rash enough to want to build some of
1407 the Glasgow Functional Programming tools (GHC, Happy,
1408 nofib, etc.) from source. You've slurped the source,
1409 from the CVS repository or from a source distribution, and
1410 now you're sitting looking at a huge mound of bits, wondering
1415 Gingerly, you type <Command>make</Command>. Wrong already!
1419 This rest of this guide is intended for duffers like me, who aren't
1420 really interested in Makefiles and systems configurations, but who
1421 need a mental model of the interlocking pieces so that they can make
1422 them work, extend them consistently when adding new software, and lay
1423 hands on them gently when they don't work.
1426 <Sect2 id="sec-source-tree">
1427 <Title>Your source tree
1431 The source code is held in your <Emphasis>source tree</Emphasis>.
1432 The root directory of your source tree <Emphasis>must</Emphasis>
1433 contain the following directories and files:
1442 <Filename>Makefile</Filename>: the root Makefile.
1448 <Filename>mk/</Filename>: the directory that contains the
1449 main Makefile code, shared by all the
1450 <Literal>fptools</Literal> software.
1456 <Filename>configure.in</Filename>, <Filename>config.sub</Filename>, <Filename>config.guess</Filename>:
1457 these files support the configuration process.
1463 <Filename>install-sh</Filename>.
1472 All the other directories are individual <Emphasis>projects</Emphasis> of the
1473 <Literal>fptools</Literal> system—for example, the Glasgow Haskell Compiler
1474 (<Literal>ghc</Literal>), the Happy parser generator (<Literal>happy</Literal>), the <Literal>nofib</Literal> benchmark
1475 suite, and so on. You can have zero or more of these. Needless to
1476 say, some of them are needed to build others.
1480 The important thing to remember is that even if you want only one
1481 project (<Literal>happy</Literal>, say), you must have a source tree whose root
1482 directory contains <Filename>Makefile</Filename>, <Filename>mk/</Filename>, <Filename>configure.in</Filename>, and the
1483 project(s) you want (<Filename>happy/</Filename> in this case). You cannot get by with
1484 just the <Filename>happy/</Filename> directory.
1491 <indexterm><primary>build trees</primary></indexterm>
1492 <indexterm><primary>link trees, for building</primary></indexterm></Title>
1495 While you can build a system in the source tree, we don't recommend it.
1496 We often want to build multiple versions of our software
1497 for different architectures, or with different options (e.g. profiling).
1498 It's very desirable to share a single copy of the source code among
1503 So for every source tree we have zero or more <Emphasis>build trees</Emphasis>. Each
1504 build tree is initially an exact copy of the source tree, except that
1505 each file is a symbolic link to the source file, rather than being a
1506 copy of the source file. There are ``standard'' Unix utilities that
1507 make such copies, so standard that they go by different names:
1508 <Command>lndir</Command><indexterm><primary>lndir</primary></indexterm>, <Command>mkshadowdir</Command><indexterm><primary>mkshadowdir</primary></indexterm> are two (If you
1509 don't have either, the source distribution includes sources for the
1510 X11 <Command>lndir</Command>—check out <Filename>fptools/glafp-utils/lndir</Filename>). See <Xref LinkEnd="sec-storysofar"> for a typical invocation.
1514 The build tree does not need to be anywhere near the source tree in
1515 the file system. Indeed, one advantage of separating the build tree
1516 from the source is that the build tree can be placed in a
1517 non-backed-up partition, saving your systems support people from
1518 backing up untold megabytes of easily-regenerated, and
1519 rapidly-changing, gubbins. The golden rule is that (with a single
1520 exception—<XRef LinkEnd="sec-build-config">)
1521 <Emphasis>absolutely everything in the build tree is either a symbolic
1522 link to the source tree, or else is mechanically generated</Emphasis>.
1523 It should be perfectly OK for your build tree to vanish overnight; an
1524 hour or two compiling and you're on the road again.
1528 You need to be a bit careful, though, that any new files you create
1529 (if you do any development work) are in the source tree, not a build tree!
1533 Remember, that the source files in the build tree are <Emphasis>symbolic
1534 links</Emphasis> to the files in the source tree. (The build tree soon
1535 accumulates lots of built files like <Filename>Foo.o</Filename>, as well.) You
1536 can <Emphasis>delete</Emphasis> a source file from the build tree without affecting
1537 the source tree (though it's an odd thing to do). On the other hand,
1538 if you <Emphasis>edit</Emphasis> a source file from the build tree, you'll edit the
1539 source-tree file directly. (You can set up Emacs so that if you edit
1540 a source file from the build tree, Emacs will silently create an
1541 edited copy of the source file in the build tree, leaving the source
1542 file unchanged; but the danger is that you think you've edited the
1543 source file whereas actually all you've done is edit the build-tree
1544 copy. More commonly you do want to edit the source file.)
1548 Like the source tree, the top level of your build tree must be (a
1549 linked copy of) the root directory of the <Literal>fptools</Literal> suite. Inside
1550 Makefiles, the root of your build tree is called
1551 <constant>$(FPTOOLS_TOP)</constant><indexterm><primary>FPTOOLS_TOP</primary></indexterm>. In the rest of this document path
1552 names are relative to <constant>$(FPTOOLS_TOP)</constant> unless otherwise stated. For
1553 example, the file <Filename>ghc/mk/target.mk</Filename> is actually
1554 <Filename><constant>$(FPTOOLS_TOP)</constant>/ghc/mk/target.mk</Filename>.
1559 <Sect2 id="sec-build-config">
1560 <Title>Getting the build you want
1564 When you build <Literal>fptools</Literal> you will be compiling code on a particular
1565 <Emphasis>host platform</Emphasis>, to run on a particular <Emphasis>target platform</Emphasis>
1566 (usually the same as the host platform)<indexterm><primary>platform</primary></indexterm>. The
1567 difficulty is that there are minor differences between different
1568 platforms; minor, but enough that the code needs to be a bit different
1569 for each. There are some big differences too: for a different
1570 architecture we need to build GHC with a different native-code
1575 There are also knobs you can turn to control how the <Literal>fptools</Literal>
1576 software is built. For example, you might want to build GHC optimised
1577 (so that it runs fast) or unoptimised (so that you can compile it fast
1578 after you've modified it. Or, you might want to compile it with
1579 debugging on (so that extra consistency-checking code gets included)
1584 All of this stuff is called the <Emphasis>configuration</Emphasis> of your build.
1585 You set the configuration using a three-step process.
1589 <term>Step 1: get ready for configuration.</term>
1591 <para>Change directory to
1592 <constant>$(FPTOOLS_TOP)</constant> and
1594 <Command>autoconf</Command><indexterm><primary>autoconf</primary></indexterm>
1595 (with no arguments). This GNU program converts
1596 <Filename><constant>$(FPTOOLS_TOP)</constant>/configure.in</Filename>
1597 to a shell script called
1598 <Filename><constant>$(FPTOOLS_TOP)</constant>/configure</Filename>.
1601 <para>Some projects, including GHC, have their own
1602 configure script. If there's an
1603 <constant>$(FPTOOLS_TOP)/<project>/configure.in</constant>,
1604 then you need to run <command>autoconf</command> in that
1605 directory too.</para>
1607 <para>Both these steps are completely
1608 platform-independent; they just mean that the
1609 human-written file (<Filename>configure.in</Filename>)
1610 can be short, although the resulting shell script,
1611 <Command>configure</Command>, and
1612 <Filename>mk/config.h.in</Filename>, are long.</para>
1614 <para>In case you don't have <Command>autoconf</Command>
1615 we distribute the results, <Command>configure</Command>,
1616 and <Filename>mk/config.h.in</Filename>, with the source
1617 distribution. They aren't kept in the repository,
1623 <term>Step 2: system configuration.</term>
1625 <para>Runs the newly-created
1626 <Command>configure</Command> script, thus:</para>
1629 ./configure <optional><parameter>args</parameter></optional>
1632 <para><Command>configure</Command>'s mission is to
1633 scurry round your computer working out what architecture
1634 it has, what operating system, whether it has the
1635 <Function>vfork</Function> system call, where
1636 <Command>yacc</Command> is kept, whether
1637 <Command>gcc</Command> is available, where various
1638 obscure <Literal>#include</Literal> files are,
1639 whether it's a leap year, and what the systems manager
1640 had for lunch. It communicates these snippets of
1641 information in two ways:</para>
1647 <Filename>mk/config.mk.in</Filename><indexterm><primary>config.mk.in</primary></indexterm>
1649 <Filename>mk/config.mk</Filename><indexterm><primary>config.mk</primary></indexterm>,
1650 substituting for things between
1651 ``<Literal>@</Literal>'' brackets. So,
1652 ``<Literal>@HaveGcc@</Literal>'' will be replaced by
1653 ``<Literal>YES</Literal>'' or
1654 ``<Literal>NO</Literal>'' depending on what
1655 <Command>configure</Command> finds.
1656 <Filename>mk/config.mk</Filename> is included by
1657 every Makefile (directly or indirectly), so the
1658 configuration information is thereby communicated to
1659 all Makefiles.</para>
1663 <para> It translates
1664 <Filename>mk/config.h.in</Filename><indexterm><primary>config.h.in</primary></indexterm>
1666 <Filename>mk/config.h</Filename><indexterm><primary>config.h</primary></indexterm>.
1667 The latter is <Literal>#include</Literal>d by
1668 various C programs, which can thereby make use of
1669 configuration information.</para>
1673 <para><command>configure</command> takes some optional
1674 arguments. Use <literal>./configure --help</literal> to
1675 get a list of the available arguments. Here are some of
1676 the ones you might need:</para>
1680 <term><literal>--with-ghc=<parameter>path</parameter></literal></term>
1681 <indexterm><primary><literal>--with-ghc</literal></primary>
1684 <para>Specifies the path to an installed GHC which
1685 you would like to use. This compiler will be used
1686 for compiling GHC-specific code (eg. GHC itself).
1687 This option <emphasis>cannot</emphasis> be
1688 specified using <filename>build.mk</filename> (see
1689 later), because <command>configure</command> needs
1690 to auto-detect the version of GHC you're using.
1691 The default is to look for a compiler named
1692 <literal>ghc</literal> in your path.</para>
1697 <term><literal>--with-hc=<parameter>path</parameter></literal></term>
1698 <indexterm><primary><literal>--with-hc</literal></primary>
1701 <para>Specifies the path to any installed Haskell
1702 compiler. This compiler will be used for
1703 compiling generic Haskell code. The default is to
1704 use <literal>ghc</literal>.</para>
1709 <term><literal>--with-gcc=<parameter>path</parameter></literal></term>
1710 <indexterm><primary><literal>--with-gcc</literal></primary>
1713 <para>Specifies the path to the installed
1714 GCC. This compiler will be used to compile all C
1715 files, <emphasis>except</emphasis> any generated
1716 by the installed Haskell compiler, which will have
1717 its own idea of which C compiler (if any) to use.
1718 The default is to use <literal>gcc</literal>.</para>
1723 <para><command>configure</command> caches the results of
1724 its run in <Filename>config.cache</Filename>. Quite
1725 often you don't want that; you're running
1726 <Command>configure</Command> a second time because
1727 something has changed. In that case, simply delete
1728 <Filename>config.cache</Filename>.</para>
1733 <term>Step 3: build configuration.</term>
1736 Next, you say how this build of <Literal>fptools</Literal> is to differ from the
1737 standard defaults by creating a new file <Filename>mk/build.mk</Filename><indexterm><primary>build.mk</primary></indexterm>
1738 <Emphasis>in the build tree</Emphasis>. This file is the one and only file you edit
1739 in the build tree, precisely because it says how this build differs
1740 from the source. (Just in case your build tree does die, you might
1741 want to keep a private directory of <Filename>build.mk</Filename> files, and use a
1742 symbolic link in each build tree to point to the appropriate one.) So
1743 <Filename>mk/build.mk</Filename> never exists in the source tree—you create one in
1744 each build tree from the template. We'll discuss what to put in it
1747 </listitem></VarListEntry>
1752 And that's it for configuration. Simple, eh?
1755 <para>What do you put in your build-specific configuration file
1756 <filename>mk/build.mk</filename>? <Emphasis>For almost all
1757 purposes all you will do is put make variable definitions that
1758 override those in</Emphasis>
1759 <filename>mk/config.mk.in</filename>. The whole point of
1760 <filename>mk/config.mk.in</filename>—and its derived
1761 counterpart <filename>mk/config.mk</filename>—is to define
1762 the build configuration. It is heavily commented, as you will
1763 see if you look at it. So generally, what you do is look at
1764 <filename>mk/config.mk.in</filename>, and add definitions in
1765 <filename>mk/build.mk</filename> that override any of the
1766 <filename>config.mk</filename> definitions that you want to
1767 change. (The override occurs because the main boilerplate file,
1768 <filename>mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>,
1769 includes <filename>build.mk</filename> after
1770 <filename>config.mk</filename>.)</para>
1772 <para>For example, <filename>config.mk.in</filename> contains
1773 the definition:</para>
1776 GhcHcOpts=-O -Rghc-timing
1779 <para>The accompanying comment explains that this is the list of
1780 flags passed to GHC when building GHC itself. For doing
1781 development, it is wise to add <literal>-DDEBUG</literal>, to
1782 enable debugging code. So you would add the following to
1783 <filename>build.mk</filename>:</para>
1785 <para>or, if you prefer,</para>
1788 GhcHcOpts += -DDEBUG
1791 <para>GNU <Command>make</Command> allows existing definitions to
1792 have new text appended using the ``<Literal>+=</Literal>''
1793 operator, which is quite a convenient feature.)</para>
1795 <para>If you want to remove the <literal>-O</literal> as well (a
1796 good idea when developing, because the turn-around cycle gets a
1797 lot quicker), you can just override
1798 <literal>GhcLibHcOpts</literal> altogether:</para>
1801 GhcHcOpts=-DDEBUG -Rghc-timing
1804 <para>When reading <filename>config.mk.in</filename>, remember
1805 that anything between ``@...@'' signs is going to be substituted
1806 by <Command>configure</Command> later. You
1807 <Emphasis>can</Emphasis> override the resulting definition if
1808 you want, but you need to be a bit surer what you are doing.
1809 For example, there's a line that says:</para>
1815 <para>This defines the Make variables <constant>YACC</constant>
1816 to the pathname for a <Command>yacc</Command> that
1817 <Command>configure</Command> finds somewhere. If you have your
1818 own pet <Command>yacc</Command> you want to use instead, that's
1819 fine. Just add this line to <filename>mk/build.mk</filename>:</para>
1825 <para>You do not <Emphasis>have</Emphasis> to have a
1826 <filename>mk/build.mk</filename> file at all; if you don't,
1827 you'll get all the default settings from
1828 <filename>mk/config.mk.in</filename>.</para>
1830 <para>You can also use <filename>build.mk</filename> to override
1831 anything that <Command>configure</Command> got wrong. One place
1832 where this happens often is with the definition of
1833 <constant>FPTOOLS_TOP_ABS</constant>: this
1834 variable is supposed to be the canonical path to the top of your
1835 source tree, but if your system uses an automounter then the
1836 correct directory is hard to find automatically. If you find
1837 that <Command>configure</Command> has got it wrong, just put the
1838 correct definition in <filename>build.mk</filename>.</para>
1842 <sect2 id="sec-storysofar">
1843 <title>The story so far</title>
1845 <para>Let's summarise the steps you need to carry to get
1846 yourself a fully-configured build tree from scratch.</para>
1850 <para> Get your source tree from somewhere (CVS repository
1851 or source distribution). Say you call the root directory
1852 <filename>myfptools</filename> (it does not have to be
1853 called <filename>fptools</filename>). Make sure that you
1854 have the essential files (see <XRef
1855 LinkEnd="sec-source-tree">).</para>
1860 <para>(Optional) Use <Command>lndir</Command> or
1861 <Command>mkshadowdir</Command> to create a build tree.</para>
1865 $ mkshadowdir . /scratch/joe-bloggs/myfptools-sun4
1868 <para>(N.B. <Command>mkshadowdir</Command>'s first argument
1869 is taken relative to its second.) You probably want to give
1870 the build tree a name that suggests its main defining
1871 characteristic (in your mind at least), in case you later
1876 <para>Change directory to the build tree. Everything is
1877 going to happen there now.</para>
1880 $ cd /scratch/joe-bloggs/myfptools-sun4
1886 <para>Prepare for system configuration:</para>
1892 <para>(You can skip this step if you are starting from a
1893 source distribution, and you already have
1894 <filename>configure</filename> and
1895 <filename>mk/config.h.in</filename>.)</para>
1897 <para>Some projects, including GHC itself, have their own
1898 configure scripts, so it is necessary to run autoconf again
1899 in the appropriate subdirectories. eg:</para>
1902 $ (cd ghc; autoconf)
1907 <para>Do system configuration:</para>
1913 <para>Don't forget to check whether you need to add any
1914 arguments to <literal>configure</literal>; for example, a
1915 common requirement is to specify which GHC to use with
1916 <option>--with-ghc=<replaceable>ghc</replaceable></option>.</para>
1920 <para>Create the file <filename>mk/build.mk</filename>,
1921 adding definitions for your desired configuration
1930 <para>You can make subsequent changes to
1931 <filename>mk/build.mk</filename> as often as you like. You do
1932 not have to run any further configuration programs to make these
1933 changes take effect. In theory you should, however, say
1934 <Command>gmake clean</Command>, <Command>gmake all</Command>,
1935 because configuration option changes could affect
1936 anything—but in practice you are likely to know what's
1941 <Title>Making things</Title>
1943 <para>At this point you have made yourself a fully-configured
1944 build tree, so you are ready to start building real
1947 <para>The first thing you need to know is that <Emphasis>you
1948 must use GNU <Command>make</Command>, usually called
1949 <Command>gmake</Command>, not standard Unix
1950 <Command>make</Command></Emphasis>. If you use standard Unix
1951 <Command>make</Command> you will get all sorts of error messages
1952 (but no damage) because the <Literal>fptools</Literal>
1953 <Command>Makefiles</Command> use GNU <Command>make</Command>'s
1954 facilities extensively.</para>
1956 <para>To just build the whole thing, <command>cd</command> to
1957 the top of your <literal>fptools</literal> tree and type
1958 <command>gmake</command>. This will prepare the tree and build
1959 the various projects in the correct order.</para>
1963 <Sect2 id="sec-standard-targets">
1964 <Title>Standard Targets</title>
1965 <indexterm><primary>targets, standard makefile</primary></indexterm>
1966 <indexterm><primary>makefile targets</primary></indexterm>
1968 <para>In any directory you should be able to make the following:
1973 <term><Literal>boot</Literal>:</term>
1975 <para>does the one-off preparation required to get ready for the real
1976 work. Notably, it does <Command>gmake depend</Command> in all
1977 directories that contain programs. It also builds the necessary tools
1978 for compilation to proceed.</para>
1980 <para>Invoking the <literal>boot</literal> target explicitly is not
1981 normally necessary. From the top-level <literal>fptools</literal>
1982 directory, invoking <literal>gmake</literal> causes <literal>gmake
1983 boot all</literal> to be invoked in each of the project
1984 subdirectories, in the order specified by
1985 <literal>$(AllTargets)</literal> in
1986 <literal>config.mk</literal>.</para>
1988 <para>If you're working in a subdirectory somewhere and need to update
1989 the dependencies, <literal>gmake boot</literal> is a good way to do it.</para>
1991 </listitem></VarListEntry>
1993 <term><Literal>all</Literal>:</term>
1996 makes all the final target(s) for this Makefile.
1997 Depending on which directory you are in a ``final target'' may be an
1998 executable program, a library archive, a shell script, or a Postscript
1999 file. Typing <Command>gmake</Command> alone is generally the same as typing <Command>gmake all</Command>.
2001 </listitem></VarListEntry>
2003 <term><Literal>install</Literal>:</term>
2006 installs the things built by <Literal>all</Literal> (except for the documentation). Where does it
2007 install them? That is specified by
2008 <filename>mk/config.mk.in</filename>; you can override it in
2009 <filename>mk/build.mk</filename>, or by running
2010 <command>configure</command> with command-line arguments like
2011 <literal>--bindir=/home/simonpj/bin</literal>; see <literal>./configure
2012 --help</literal> for the full details.
2014 </listitem></VarListEntry>
2016 <term><Literal>install-docs</Literal>:</term>
2019 installs the documentation. Otherwise behaves just like <literal>install</literal>.
2021 </listitem></VarListEntry>
2023 <term><Literal>uninstall</Literal>:</term>
2026 reverses the effect of <Literal>install</Literal>.
2028 </listitem></VarListEntry>
2031 <term><Literal>clean</Literal>:</term>
2034 Delete all files from the current directory that are normally created
2035 by building the program. Don't delete the files that record the
2036 configuration, or files generated by <Command>gmake boot</Command>.
2037 Also preserve files that could be made by building, but normally
2038 aren't because the distribution comes with them.</para>
2039 </listitem></VarListEntry>
2042 <term><literal>distclean</literal>:</term>
2044 <para>Delete all files from the current directory that are created by
2045 configuring or building the program. If you have unpacked the source
2046 and built the program without creating any other files, <literal>make
2047 distclean</literal> should leave only the files that were in the
2048 distribution.</para>
2053 <term><literal>mostlyclean</literal>:</term>
2055 <para>Like <literal>clean</literal>, but may refrain from deleting a
2056 few files that people normally don't want to recompile.</para>
2061 <term><Literal>maintainer-clean</Literal>:</term>
2064 Delete everything from the current directory that can be reconstructed
2065 with this Makefile. This typically includes everything deleted by
2066 <literal>distclean</literal>, plus more: C source files produced by
2067 Bison, tags tables, Info files, and so on.</para>
2069 <para>One exception, however: <literal>make maintainer-clean</literal>
2070 should not delete <filename>configure</filename> even if
2071 <filename>configure</filename> can be remade using a rule in the
2072 <filename>Makefile</filename>. More generally, <literal>make
2073 maintainer-clean</literal> should not delete anything that needs to
2074 exist in order to run <filename>configure</filename> and then begin to
2075 build the program.</para>
2080 <term><Literal>check</Literal>:</term>
2085 </listitem></VarListEntry>
2090 All of these standard targets automatically recurse into
2091 sub-directories. Certain other standard targets do not:
2098 <term><Literal>configure</Literal>:</term>
2101 is only available in the root directory
2102 <constant>$(FPTOOLS_TOP)</constant>; it has been discussed in <XRef LinkEnd="sec-build-config">.
2104 </listitem></VarListEntry>
2106 <term><Literal>depend</Literal>:</term>
2109 make a <filename>.depend</filename> file in each directory that needs
2110 it. This <filename>.depend</filename> file contains mechanically-generated dependency
2111 information; for example, suppose a directory contains a Haskell
2112 source module <filename>Foo.lhs</filename> which imports another module <Literal>Baz</Literal>.
2113 Then the generated <filename>.depend</filename> file will contain the dependency:
2125 which says that the object file <filename>Foo.o</filename> depends on the interface file
2126 <filename>Baz.hi</filename> generated by compiling module <Literal>Baz</Literal>. The <filename>.depend</filename> file is
2127 automatically included by every Makefile.
2129 </listitem></VarListEntry>
2131 <term><Literal>binary-dist</Literal>:</term>
2134 make a binary distribution. This is the
2135 target we use to build the binary distributions of GHC and Happy.
2137 </listitem></VarListEntry>
2139 <term><Literal>dist</Literal>:</term>
2142 make a source distribution. Note that this target does “make
2143 distclean” as part of its work; don't use it if you want to keep
2146 </listitem></VarListEntry>
2151 Most <filename>Makefile</filename>s have targets other than these. You can discover them by looking in the <filename>Makefile</filename> itself.
2157 <title>Using a project from the build tree</title>
2159 If you want to build GHC (say) and just use it direct from the build
2160 tree without doing <literal>make install</literal> first, you can run
2161 the in-place driver script:
2162 <filename>ghc/compiler/ghc-inplace</filename>.
2165 <para> Do <emphasis>NOT</emphasis> use
2166 <filename>ghc/compiler/ghc</filename>, or
2167 <filename>ghc/compiler/ghc-5.xx</filename>, as these are the scripts
2168 intended for installation, and contain hard-wired paths to the
2169 installed libraries, rather than the libraries in the build tree.
2173 Happy can similarly be run from the build tree, using
2174 <filename>happy/src/happy-inplace</filename>.
2179 <Title>Fast Making <indexterm><primary>fastmake</primary></indexterm>
2180 <indexterm><primary>dependencies, omitting</primary></indexterm>
2181 <indexterm><primary>FAST, makefile
2182 variable</primary></indexterm></Title>
2185 Sometimes the dependencies get in the way: if you've made a small
2186 change to one file, and you're absolutely sure that it won't affect
2187 anything else, but you know that <Command>make</Command> is going to rebuild everything
2188 anyway, the following hack may be useful:
2200 This tells the make system to ignore dependencies and just build what
2201 you tell it to. In other words, it's equivalent to temporarily
2202 removing the <filename>.depend</filename> file in the current directory (where
2203 <Command>mkdependHS</Command> and friends store their dependency information).
2207 A bit of history: GHC used to come with a <Command>fastmake</Command> script that did
2208 the above job, but GNU make provides the features we need to do it
2209 without resorting to a script. Also, we've found that fastmaking is
2210 less useful since the advent of GHC's recompilation checker (see the
2211 User's Guide section on "Separate Compilation").
2218 <Sect1 id="sec-makefile-arch">
2219 <Title>The <filename>Makefile</filename> architecture
2220 <indexterm><primary>makefile architecture</primary></indexterm></Title>
2223 <Command>make</Command> is great if everything works—you type <Command>gmake install</Command> and
2224 lo! the right things get compiled and installed in the right places.
2225 Our goal is to make this happen often, but somehow it often doesn't;
2226 instead some weird error message eventually emerges from the bowels of
2227 a directory you didn't know existed.
2231 The purpose of this section is to give you a road-map to help you figure
2232 out what is going right and what is going wrong.
2236 <title>Debugging</title>
2238 <para>Debugging <filename>Makefile</filename>s is something of a
2239 black art, but here's a couple of tricks that we find
2240 particularly useful. The following command allows you to see
2241 the contents of any make variable in the context of the current
2242 <filename>Makefile</filename>:</para>
2244 <screen>$ make show VALUE=HS_SRCS</screen>
2246 <para>where you can replace <literal>HS_SRCS</literal> with the
2247 name of any variable you wish to see the value of.</para>
2249 <para>GNU make has a <option>-d</option> option which generates
2250 a dump of the decision procedure used to arrive at a conclusion
2251 about which files should be recompiled. Sometimes useful for
2252 tracking down problems with superfluous or missing
2253 recompilations.</para>
2257 <Title>A small project</Title>
2260 To get started, let us look at the <filename>Makefile</filename> for an imaginary small
2261 <Literal>fptools</Literal> project, <Literal>small</Literal>. Each project in <Literal>fptools</Literal> has its own
2262 directory in <constant>FPTOOLS_TOP</constant>, so the <Literal>small</Literal> project will have its own
2263 directory <constant>FPOOLS_TOP/small/</constant>. Inside the <filename>small/</filename> directory there
2264 will be a <filename>Makefile</filename>, looking something like this:
2268 <indexterm><primary>Makefile, minimal</primary></indexterm>
2271 # Makefile for fptools project "small"
2274 include $(TOP)/mk/boilerplate.mk
2276 SRCS = $(wildcard *.lhs) $(wildcard *.c)
2279 include $(TOP)/target.mk
2285 This <filename>Makefile</filename> has three sections:
2294 The first section includes
2298 One of the most important
2299 features of GNU <Command>make</Command> that we use is the ability for a <filename>Makefile</filename> to
2300 include another named file, very like <Command>cpp</Command>'s <Literal>#include</Literal>
2305 a file of ``boilerplate'' code from the level
2306 above (which in this case will be
2307 <filename><constant>FPTOOLS_TOP</constant>/mk/boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm>). As its name
2308 suggests, <filename>boilerplate.mk</filename> consists of a large quantity of standard
2309 <filename>Makefile</filename> code. We discuss this boilerplate in more detail in
2310 <XRef LinkEnd="sec-boiler">.
2311 <indexterm><primary>include, directive in Makefiles</primary></indexterm>
2312 <indexterm><primary>Makefile inclusion</primary></indexterm>
2314 Before the <Literal>include</Literal> statement, you must define the <Command>make</Command> variable
2315 <constant>TOP</constant><indexterm><primary>TOP</primary></indexterm> to be the directory containing the <filename>mk</filename> directory in
2316 which the <filename>boilerplate.mk</filename> file is. It is <Emphasis>not</Emphasis> OK to simply say
2320 include ../mk/boilerplate.mk # NO NO NO
2324 Why? Because the <filename>boilerplate.mk</filename> file needs to know where it is, so
2325 that it can, in turn, <Literal>include</Literal> other files. (Unfortunately, when an
2326 <Literal>include</Literal>d file does an <Literal>include</Literal>, the filename is treated relative to
2327 the directory in which <Command>gmake</Command> is being run, not the directory in
2328 which the <Literal>include</Literal>d sits.) In general, <Emphasis>every file <filename>foo.mk</filename>
2329 assumes that <filename><constant>$(TOP)</constant>/mk/foo.mk</filename> refers to itself.</Emphasis> It is up to the
2330 <filename>Makefile</filename> doing the <Literal>include</Literal> to ensure this is the case.
2332 Files intended for inclusion in other <filename>Makefile</filename>s are written to have
2333 the following property: <Emphasis>after <filename>foo.mk</filename> is <Literal>include</Literal>d, it leaves
2334 <constant>TOP</constant> containing the same value as it had just before the <Literal>include</Literal>
2335 statement</Emphasis>. In our example, this invariant guarantees that the
2336 <Literal>include</Literal> for <filename>target.mk</filename> will look in the same directory as that for
2337 <filename>boilerplate.mk</filename>.
2344 The second section defines the following standard <Command>make</Command>
2345 variables: <constant>SRCS</constant><indexterm><primary>SRCS</primary></indexterm> (the source files from which is to be
2346 built), and <constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm> (the executable binary to be
2347 built). We will discuss in more detail what the ``standard
2348 variables'' are, and how they affect what happens, in <XRef LinkEnd="sec-targets">.
2350 The definition for <constant>SRCS</constant> uses the useful GNU <Command>make</Command> construct
2351 <Literal>$(wildcard $pat$)</Literal><indexterm><primary>wildcard</primary></indexterm>, which expands to a list of all
2352 the files matching the pattern <Literal>pat</Literal> in the current directory. In
2353 this example, <constant>SRCS</constant> is set to the list of all the <filename>.lhs</filename> and <filename>.c</filename>
2354 files in the directory. (Let's suppose there is one of each,
2355 <filename>Foo.lhs</filename> and <filename>Baz.c</filename>.)
2362 The last section includes a second file of standard code,
2363 called <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>. It contains the rules that tell
2364 <Command>gmake</Command> how to make the standard targets (<Xref LinkEnd="sec-standard-targets">). Why, you ask,
2365 can't this standard code be part of <filename>boilerplate.mk</filename>? Good question.
2366 We discuss the reason later, in <Xref LinkEnd="sec-boiler-arch">.
2368 You do not <Emphasis>have</Emphasis> to <Literal>include</Literal> the <filename>target.mk</filename> file. Instead, you
2369 can write rules of your own for all the standard targets. Usually,
2370 though, you will find quite a big payoff from using the canned rules
2371 in <filename>target.mk</filename>; the price tag is that you have to understand what
2372 canned rules get enabled, and what they do (<Xref LinkEnd="sec-targets">).
2382 In our example <filename>Makefile</filename>, most of the work is done by the two
2383 <Literal>include</Literal>d files. When you say <Command>gmake all</Command>, the following things
2393 <Command>gmake</Command> figures out that the object files are <filename>Foo.o</filename> and
2394 <filename>Baz.o</filename>.
2401 It uses a boilerplate pattern rule to compile <filename>Foo.lhs</filename> to
2402 <filename>Foo.o</filename> using a Haskell compiler. (Which one? That is set in the
2403 build configuration.)
2410 It uses another standard pattern rule to compile <filename>Baz.c</filename> to
2411 <filename>Baz.o</filename>, using a C compiler. (Ditto.)
2418 It links the resulting <filename>.o</filename> files together to make <Literal>small</Literal>,
2419 using the Haskell compiler to do the link step. (Why not use <Command>ld</Command>?
2420 Because the Haskell compiler knows what standard libraries to link in.
2421 How did <Command>gmake</Command> know to use the Haskell compiler to do the link,
2422 rather than the C compiler? Because we set the variable <constant>HS_PROG</constant>
2423 rather than <constant>C_PROG</constant>.)
2433 All <filename>Makefile</filename>s should follow the above three-section format.
2439 <Title>A larger project</Title>
2442 Larger projects are usually structured into a number of sub-directories,
2443 each of which has its own <filename>Makefile</filename>. (In very large projects, this
2444 sub-structure might be iterated recursively, though that is rare.)
2445 To give you the idea, here's part of the directory structure for
2446 the (rather large) GHC project:
2459 ...source files for documentation...
2462 ...source files for driver...
2465 parser/...source files for parser...
2466 renamer/...source files for renamer...
2473 The sub-directories <filename>docs</filename>, <filename>driver</filename>, <filename>compiler</filename>, and so on, each
2474 contains a sub-component of GHC, and each has its own <filename>Makefile</filename>.
2475 There must also be a <filename>Makefile</filename> in <filename><constant>$(FPTOOLS_TOP)</constant>/ghc</filename>. It does most
2476 of its work by recursively invoking <Command>gmake</Command> on the <filename>Makefile</filename>s in the
2477 sub-directories. We say that <filename>ghc/Makefile</filename> is a <Emphasis>non-leaf
2478 <filename>Makefile</filename></Emphasis>, because it does little except organise its children,
2479 while the <filename>Makefile</filename>s in the sub-directories are all <Emphasis>leaf
2480 <filename>Makefile</filename>s</Emphasis>. (In principle the sub-directories might themselves
2481 contain a non-leaf <filename>Makefile</filename> and several sub-sub-directories, but
2482 that does not happen in GHC.)
2486 The <filename>Makefile</filename> in <filename>ghc/compiler</filename> is considered a leaf <filename>Makefile</filename> even
2487 though the <filename>ghc/compiler</filename> has sub-directories, because these sub-directories
2488 do not themselves have <filename>Makefile</filename>s in them. They are just used to structure
2489 the collection of modules that make up GHC, but all are managed by the
2490 single <filename>Makefile</filename> in <filename>ghc/compiler</filename>.
2494 You will notice that <filename>ghc/</filename> also contains a directory <filename>ghc/mk/</filename>. It
2495 contains GHC-specific <filename>Makefile</filename> boilerplate code. More precisely:
2504 <filename>ghc/mk/boilerplate.mk</filename> is included at the top of
2505 <filename>ghc/Makefile</filename>, and of all the leaf <filename>Makefile</filename>s in the
2506 sub-directories. It in turn <Literal>include</Literal>s the main boilerplate file
2507 <filename>mk/boilerplate.mk</filename>.
2515 <filename>ghc/mk/target.mk</filename> is <Literal>include</Literal>d at the bottom of
2516 <filename>ghc/Makefile</filename>, and of all the leaf <filename>Makefile</filename>s in the
2517 sub-directories. It in turn <Literal>include</Literal>s the file <filename>mk/target.mk</filename>.
2527 So these two files are the place to look for GHC-wide customisation
2528 of the standard boilerplate.
2533 <Sect2 id="sec-boiler-arch">
2534 <Title>Boilerplate architecture
2535 <indexterm><primary>boilerplate architecture</primary></indexterm>
2539 Every <filename>Makefile</filename> includes a <filename>boilerplate.mk</filename><indexterm><primary>boilerplate.mk</primary></indexterm> file
2540 at the top, and <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm> file at the bottom. In
2541 this section we discuss what is in these files, and why there have to
2542 be two of them. In general:
2551 <filename>boilerplate.mk</filename> consists of:
2557 <Emphasis>Definitions of millions of <Command>make</Command> variables</Emphasis> that
2558 collectively specify the build configuration. Examples:
2559 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>, the options to feed to the Haskell compiler;
2560 <constant>NoFibSubDirs</constant><indexterm><primary>NoFibSubDirs</primary></indexterm>, the sub-directories to enable within the
2561 <Literal>nofib</Literal> project; <constant>GhcWithHc</constant><indexterm><primary>GhcWithHc</primary></indexterm>, the name of the Haskell
2562 compiler to use when compiling GHC in the <Literal>ghc</Literal> project.
2568 <Emphasis>Standard pattern rules</Emphasis> that tell <Command>gmake</Command> how to construct one
2576 <filename>boilerplate.mk</filename> needs to be <Literal>include</Literal>d at the <Emphasis>top</Emphasis>
2577 of each <filename>Makefile</filename>, so that the user can replace the
2578 boilerplate definitions or pattern rules by simply giving a new
2579 definition or pattern rule in the <filename>Makefile</filename>. <Command>gmake</Command>
2580 simply takes the last definition as the definitive one.
2582 Instead of <Emphasis>replacing</Emphasis> boilerplate definitions, it is also quite
2583 common to <Emphasis>augment</Emphasis> them. For example, a <filename>Makefile</filename> might say:
2591 thereby adding ``<Option>-O</Option>'' to the end of <constant>SRC_HC_OPTS</constant><indexterm><primary>SRC_HC_OPTS</primary></indexterm>.
2598 <filename>target.mk</filename> contains <Command>make</Command> rules for the standard
2599 targets described in <Xref LinkEnd="sec-standard-targets">. These rules are selectively included,
2600 depending on the setting of certain <Command>make</Command> variables. These
2601 variables are usually set in the middle section of the
2602 <filename>Makefile</filename> between the two <Literal>include</Literal>s.
2604 <filename>target.mk</filename> must be included at the end (rather than being part of
2605 <filename>boilerplate.mk</filename>) for several tiresome reasons:
2612 <Command>gmake</Command> commits target and dependency lists earlier than
2613 it should. For example, <FIlename>target.mk</FIlename> has a rule that looks like
2618 $(HS_PROG) : $(OBJS)
2619 $(HC) $(LD_OPTS) $< -o $@
2623 If this rule was in <filename>boilerplate.mk</filename> then <constant>$(HS_PROG)</constant><indexterm><primary>HS_PROG</primary></indexterm>
2624 and <constant>$(OBJS)</constant><indexterm><primary>OBJS</primary></indexterm> would not have their final values at the
2625 moment <Command>gmake</Command> encountered the rule. Alas, <Command>gmake</Command> takes a snapshot
2626 of their current values, and wires that snapshot into the rule. (In
2627 contrast, the commands executed when the rule ``fires'' are only
2628 substituted at the moment of firing.) So, the rule must follow the
2629 definitions given in the <filename>Makefile</filename> itself.
2636 Unlike pattern rules, ordinary rules cannot be overriden or
2637 replaced by subsequent rules for the same target (at least, not without an
2638 error message). Including ordinary rules in <filename>boilerplate.mk</filename> would
2639 prevent the user from writing rules for specific targets in specific cases.
2646 There are a couple of other reasons I've forgotten, but it doesn't
2662 <Sect2 id="sec-boiler">
2663 <Title>The main <filename>mk/boilerplate.mk</filename> file
2665 <indexterm><primary>boilerplate.mk</primary></indexterm></Title>
2668 If you look at <filename><constant>$(FPTOOLS_TOP)</constant>/mk/boilerplate.mk</filename> you will find
2669 that it consists of the following sections, each held in a separate
2676 <term><filename>config.mk</filename></term>
2677 <indexterm><primary>config.mk</primary></indexterm>
2679 <para>is the build configuration file we discussed at
2680 length in <Xref LinkEnd="sec-build-config">.</para>
2685 <term><filename>paths.mk</filename></term>
2686 <indexterm><primary>paths.mk</primary></indexterm>
2688 <para>defines <Command>make</Command> variables for
2689 pathnames and file lists. This file contains code for
2690 automatically compiling lists of source files and deriving
2691 lists of object files from those. The results can be
2692 overriden in the <filename>Makefile</filename>, but in
2693 most cases the automatic setup should do the right
2696 <para>The following variables may be set in the
2697 <filename>Makefile</filename> to affect how the automatic
2698 source file search is done:</para>
2702 <term><literal>ALL_DIRS</literal></term>
2703 <indexterm><primary><literal>ALL_DIRS</literal></primary>
2706 <para>Set to a list of directories to search in
2707 addition to the current directory for source
2713 <term><literal>EXCLUDE_SRCS</literal></term>
2714 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2717 <para>Set to a list of source files (relative to the
2718 current directory) to omit from the automatic
2719 search. The source searching machinery is clever
2720 enough to know that if you exclude a source file
2721 from which other sources are derived, then the
2722 derived sources should also be excluded. For
2723 example, if you set <literal>EXCLUDED_SRCS</literal>
2724 to include <filename>Foo.y</filename>, then
2725 <filename>Foo.hs</filename> will also be
2731 <term><literal>EXTRA_SRCS</literal></term>
2732 <indexterm><primary><literal>EXCLUDE_SRCS</literal></primary>
2735 <para>Set to a list of extra source files (perhaps
2736 in directories not listed in
2737 <literal>ALL_DIRS</literal>) that should be
2743 <para>The results of the automatic source file search are
2744 placed in the following make variables:</para>
2748 <term><literal>SRCS</literal></term>
2749 <indexterm><primary><literal>SRCS</literal></primary></indexterm>
2751 <para>All source files found, sorted and without
2752 duplicates, including those which might not exist
2753 yet but will be derived from other existing sources.
2754 <literal>SRCS</literal> <emphasis>can</emphasis> be
2755 overriden if necessary, in which case the variables
2756 below will follow suit.</para>
2761 <term><literal>HS_SRCS</literal></term>
2762 <indexterm><primary><literal>HS_SRCS</literal></primary></indexterm>
2764 <para>all Haskell source files in the current
2765 directory, including those derived from other source
2766 files (eg. Happy sources also give rise to Haskell
2772 <term><literal>HS_OBJS</literal></term>
2773 <indexterm><primary><literal>HS_OBJS</literal></primary></indexterm>
2775 <para>Object files derived from
2776 <literal>HS_SRCS</literal>.</para>
2781 <term><literal>HS_IFACES</literal></term>
2782 <indexterm><primary><literal>HS_IFACES</literal></primary></indexterm>
2784 <para>Interface files (<literal>.hi</literal> files)
2785 derived from <literal>HS_SRCS</literal>.</para>
2790 <term><literal>C_SRCS</literal></term>
2791 <indexterm><primary><literal>C_SRCS</literal></primary></indexterm>
2793 <para>All C source files found.</para>
2798 <term><literal>C_OBJS</literal></term>
2799 <indexterm><primary><literal>C_OBJS</literal></primary></indexterm>
2801 <para>Object files derived from
2802 <literal>C_SRCS</literal>.</para>
2807 <term><literal>SCRIPT_SRCS</literal></term>
2808 <indexterm><primary><literal>SCRIPT_SRCS</literal></primary></indexterm>
2810 <para>All script source files found
2811 (<literal>.lprl</literal> files).</para>
2816 <term><literal>SCRIPT_OBJS</literal></term>
2817 <indexterm><primary><literal>SCRIPT_OBJS</literal></primary></indexterm>
2819 <para><quote>object</quote> files derived from
2820 <literal>SCRIPT_SRCS</literal>
2821 (<literal>.prl</literal> files).</para>
2826 <term><literal>HSC_SRCS</literal></term>
2827 <indexterm><primary><literal>HSC_SRCS</literal></primary></indexterm>
2829 <para>All <literal>hsc2hs</literal> source files
2830 (<literal>.hsc</literal> files).</para>
2835 <term><literal>HAPPY_SRCS</literal></term>
2836 <indexterm><primary><literal>HAPPY_SRCS</literal></primary></indexterm>
2838 <para>All <literal>happy</literal> source files
2839 (<literal>.y</literal> or <literal>.hy</literal> files).</para>
2844 <term><literal>OBJS</literal></term>
2845 <indexterm><primary>OBJS</primary></indexterm>
2847 <para>the concatenation of
2848 <literal>$(HS_OBJS)</literal>,
2849 <literal>$(C_OBJS)</literal>, and
2850 <literal>$(SCRIPT_OBJS)</literal>.</para>
2855 <para>Any or all of these definitions can easily be
2856 overriden by giving new definitions in your
2857 <filename>Makefile</filename>.</para>
2859 <para>What, exactly, does <filename>paths.mk</filename>
2860 consider a <quote>source file</quote> to be? It's based
2861 on the file's suffix (e.g. <filename>.hs</filename>,
2862 <filename>.lhs</filename>, <filename>.c</filename>,
2863 <filename>.hy</filename>, etc), but this is the kind of
2864 detail that changes, so rather than enumerate the source
2865 suffices here the best thing to do is to look in
2866 <filename>paths.mk</filename>.</para>
2871 <term><filename>opts.mk</filename></term>
2872 <indexterm><primary>opts.mk</primary></indexterm>
2874 <para>defines <Command>make</Command> variables for option
2875 strings to pass to each program. For example, it defines
2876 <constant>HC_OPTS</constant><indexterm><primary>HC_OPTS</primary></indexterm>,
2877 the option strings to pass to the Haskell compiler. See
2878 <Xref LinkEnd="sec-suffix">.</para>
2883 <term><filename>suffix.mk</filename></term>
2884 <indexterm><primary>suffix.mk</primary></indexterm>
2886 <para>defines standard pattern rules—see <Xref
2887 LinkEnd="sec-suffix">.</para>
2893 Any of the variables and pattern rules defined by the boilerplate file
2894 can easily be overridden in any particular <filename>Makefile</filename>, because the
2895 boilerplate <Literal>include</Literal> comes first. Definitions after this <Literal>include</Literal>
2896 directive simply override the default ones in <filename>boilerplate.mk</filename>.
2901 <Sect2 id="sec-suffix">
2902 <Title>Pattern rules and options
2904 <indexterm><primary>Pattern rules</primary></indexterm></Title>
2907 The file <filename>suffix.mk</filename><indexterm><primary>suffix.mk</primary></indexterm> defines standard <Emphasis>pattern
2908 rules</Emphasis> that say how to build one kind of file from another, for
2909 example, how to build a <filename>.o</filename> file from a <filename>.c</filename> file. (GNU <Command>make</Command>'s
2910 <Emphasis>pattern rules</Emphasis> are more powerful and easier to use than Unix
2911 <Command>make</Command>'s <Emphasis>suffix rules</Emphasis>.)
2915 Almost all the rules look something like this:
2923 $(CC) $(CC_OPTS) -c $< -o $@
2929 Here's how to understand the rule. It says that
2930 <Emphasis>something</Emphasis><filename>.o</filename> (say <filename>Foo.o</filename>) can be built from
2931 <Emphasis>something</Emphasis><filename>.c</filename> (<filename>Foo.c</filename>), by invoking the C compiler
2932 (path name held in <constant>$(CC)</constant>), passing to it the options
2933 <constant>$(CC_OPTS)</constant> and the rule's dependent file of the rule
2934 <Literal>$<</Literal> (<filename>Foo.c</filename> in this case), and putting the result in
2935 the rule's target <Literal>$@</Literal> (<filename>Foo.o</filename> in this case).
2939 Every program is held in a <Command>make</Command> variable defined in
2940 <filename>mk/config.mk</filename>—look in <filename>mk/config.mk</filename> for the
2941 complete list. One important one is the Haskell compiler, which is
2942 called <constant>$(HC)</constant>.
2946 Every program's options are are held in a <Command>make</Command> variables called
2947 <constant><prog>_OPTS</constant>. the <constant><prog>_OPTS</constant> variables are defined in
2948 <filename>mk/opts.mk</filename>. Almost all of them are defined like this:
2954 CC_OPTS = $(SRC_CC_OPTS) $(WAY$(_way)_CC_OPTS) $($*_CC_OPTS) $(EXTRA_CC_OPTS)
2960 The four variables from which <constant>CC_OPTS</constant> is built have the following meaning:
2967 <term><constant>SRC_CC_OPTS</constant><indexterm><primary>SRC_CC_OPTS</primary></indexterm>:</term>
2970 options passed to all C
2973 </listitem></varlistentry>
2975 <term><constant>WAY_<way>_CC_OPTS</constant>:</term>
2979 compilations for way <Literal><way></Literal>. For example,
2980 <constant>WAY_mp_CC_OPTS</constant> gives options to pass to the C compiler when
2981 compiling way <Literal>mp</Literal>. The variable <constant>WAY_CC_OPTS</constant> holds
2982 options to pass to the C compiler when compiling the standard way.
2983 (<Xref LinkEnd="sec-ways"> dicusses multi-way
2986 </listitem></varlistentry>
2988 <term><constant><module>_CC_OPTS</constant>:</term>
2992 pass to the C compiler that are specific to module <Literal><module></Literal>. For example, <constant>SMap_CC_OPTS</constant> gives the specific options
2993 to pass to the C compiler when compiling <filename>SMap.c</filename>.
2995 </listitem></varlistentry>
2997 <term><constant>EXTRA_CC_OPTS</constant><indexterm><primary>EXTRA_CC_OPTS</primary></indexterm>:</term>
3000 extra options to pass to all
3001 C compilations. This is intended for command line use, thus:
3007 gmake libHS.a EXTRA_CC_OPTS="-v"
3011 </listitem></varlistentry>
3017 <Sect2 id="sec-targets">
3018 <Title>The main <filename>mk/target.mk</filename> file
3020 <indexterm><primary>target.mk</primary></indexterm></Title>
3023 <filename>target.mk</filename> contains canned rules for all the standard targets
3024 described in <Xref LinkEnd="sec-standard-targets">. It is complicated by the fact that you don't want all of
3025 these rules to be active in every <filename>Makefile</filename>. Rather than have a
3026 plethora of tiny files which you can include selectively, there is a
3027 single file, <filename>target.mk</filename>, which selectively includes rules based on
3028 whether you have defined certain variables in your <filename>Makefile</filename>. This
3029 section explains what rules you get, what variables control them, and
3030 what the rules do. Hopefully, you will also get enough of an idea of
3031 what is supposed to happen that you can read and understand any weird
3032 special cases yourself.
3039 <term><constant>HS_PROG</constant><indexterm><primary>HS_PROG</primary></indexterm>.</term>
3042 If <constant>HS_PROG</constant> is defined, you get
3043 rules with the following targets:
3047 <term><filename>HS_PROG</filename><indexterm><primary>HS_PROG</primary></indexterm></term>
3050 itself. This rule links <constant>$(OBJS)</constant>
3051 with the Haskell runtime system to get an executable called
3052 <constant>$(HS_PROG)</constant>.
3054 </listitem></varlistentry>
3056 <term><Literal>install</Literal><indexterm><primary>install</primary></indexterm></term>
3059 installs <constant>$(HS_PROG)</constant>
3060 in <constant>$(bindir)</constant>.
3062 </listitem></varlistentry>
3065 </listitem></varlistentry>
3067 <term><constant>C_PROG</constant><indexterm><primary>C_PROG</primary></indexterm></term>
3070 is similar to <constant>HS_PROG</constant>, except that
3071 the link step links <constant>$(C_OBJS)</constant> with the C runtime system.
3073 </listitem></varlistentry>
3075 <term><constant>LIBRARY</constant><indexterm><primary>LIBRARY</primary></indexterm></term>
3078 is similar to <constant>HS_PROG</constant>, except that
3079 it links <constant>$(LIB_OBJS)</constant> to make the library archive <constant>$(LIBRARY)</constant>, and
3080 <Literal>install</Literal> installs it in <constant>$(libdir)</constant>.
3082 </listitem></varlistentry>
3084 <term><constant>LIB_DATA</constant><indexterm><primary>LIB_DATA</primary></indexterm></term>
3089 </listitem></varlistentry>
3091 <term><constant>LIB_EXEC</constant><indexterm><primary>LIB_EXEC</primary></indexterm></term>
3096 </listitem></varlistentry>
3098 <term><constant>HS_SRCS</constant><indexterm><primary>HS_SRCS</primary></indexterm>, <constant>C_SRCS</constant><indexterm><primary>C_SRCS</primary></indexterm>.</term>
3101 If <constant>HS_SRCS</constant>
3102 is defined and non-empty, a rule for the target <Literal>depend</Literal> is included,
3103 which generates dependency information for Haskell programs.
3104 Similarly for <constant>C_SRCS</constant>.
3106 </listitem></varlistentry>
3111 All of these rules are ``double-colon'' rules, thus
3117 install :: $(HS_PROG)
3118 ...how to install it...
3124 GNU <Command>make</Command> treats double-colon rules as separate entities. If there
3125 are several double-colon rules for the same target it takes each in
3126 turn and fires it if its dependencies say to do so. This means that
3127 you can, for example, define both <constant>HS_PROG</constant> and <constant>LIBRARY</constant>, which will
3128 generate two rules for <Literal>install</Literal>. When you type <Command>gmake install</Command> both
3129 rules will be fired, and both the program and the library will be
3130 installed, just as you wanted.
3135 <Sect2 id="sec-subdirs">
3138 <indexterm><primary>recursion, in makefiles</primary></indexterm>
3139 <indexterm><primary>Makefile, recursing into subdirectories</primary></indexterm></Title>
3142 In leaf <filename>Makefile</filename>s the variable <constant>SUBDIRS</constant><indexterm><primary>SUBDIRS</primary></indexterm> is undefined.
3143 In non-leaf <filename>Makefile</filename>s, <constant>SUBDIRS</constant> is set to the list of
3144 sub-directories that contain subordinate <filename>Makefile</filename>s. <Emphasis>It is up to
3145 you to set <constant>SUBDIRS</constant> in the <filename>Makefile</filename>.</Emphasis> There is no automation here—<constant>SUBDIRS</constant> is too important to automate.
3149 When <constant>SUBDIRS</constant> is defined, <filename>target.mk</filename> includes a rather
3150 neat rule for the standard targets (<Xref LinkEnd="sec-standard-targets"> that simply invokes
3151 <Command>make</Command> recursively in each of the sub-directories.
3155 <Emphasis>These recursive invocations are guaranteed to occur in the order
3156 in which the list of directories is specified in <constant>SUBDIRS</constant>. </Emphasis>This
3157 guarantee can be important. For example, when you say <Command>gmake boot</Command> it
3158 can be important that the recursive invocation of <Command>make boot</Command> is done
3159 in one sub-directory (the include files, say) before another (the
3160 source files). Generally, put the most independent sub-directory
3161 first, and the most dependent last.
3166 <sect2 id="sec-ways">
3167 <title>Way management</title>
3168 <indexterm><primary>way management</primary></indexterm>
3170 <para>We sometimes want to build essentially the same system in
3171 several different ``ways''. For example, we want to build GHC's
3172 <literal>Prelude</literal> libraries with and without profiling,
3173 so that there is an appropriately-built library archive to link
3174 with when the user compiles his program. It would be possible
3175 to have a completely separate build tree for each such ``way'',
3176 but it would be horribly bureaucratic, especially since often
3177 only parts of the build tree need to be constructed in multiple
3181 <filename>target.mk</filename><indexterm><primary>target.mk</primary></indexterm>
3182 contains some clever magic to allow you to build several
3183 versions of a system; and to control locally how many versions
3184 are built and how they differ. This section explains the
3187 <para>The files for a particular way are distinguished by
3188 munging the suffix. The <quote>normal way</quote> is always
3189 built, and its files have the standard suffices
3190 <filename>.o</filename>, <filename>.hi</filename>, and so on.
3191 In addition, you can build one or more extra ways, each
3192 distinguished by a <emphasis>way tag</emphasis>. The object
3193 files and interface files for one of these extra ways are
3194 distinguished by their suffix. For example, way
3195 <literal>mp</literal> has files
3196 <filename>.mp_o</filename> and
3197 <filename>.mp_hi</filename>. Library archives have their
3198 way tag the other side of the dot, for boring reasons; thus,
3199 <filename>libHS_mp.a</filename>.</para>
3201 <para>A <Command>make</Command> variable called
3202 <constant>way</constant> holds the current way tag.
3203 <emphasis><constant>way</constant> is only ever set on the
3204 command line of <Command>gmake</Command></emphasis> (usually in
3205 a recursive invocation of <command>gmake</command> by the
3206 system). It is never set inside a
3207 <filename>Makefile</filename>. So it is a global constant for
3208 any one invocation of <Command>gmake</Command>. Two other
3209 <Command>make</Command> variables,
3210 <constant>way_</constant> and
3211 <constant>_way</constant> are immediately derived from
3212 <constant>$(way)</constant> and never altered. If
3213 <constant>way</constant> is not set, then neither are
3214 <constant>way_</constant> and
3215 <constant>_way</constant>, and the invocation of
3216 <Command>make</Command> will build the <quote>normal
3217 way</quote>. If <constant>way</constant> is set, then the other
3218 two variables are set in sympathy. For example, if
3219 <constant>$(way)</constant> is ``<Literal>mp</Literal>'',
3220 then <constant>way_</constant> is set to
3221 ``<Literal>mp_</Literal>'' and
3222 <constant>_way</constant> is set to
3223 ``<Literal>_mp</Literal>''. These three variables are
3224 then used when constructing file names.</para>
3226 <para>So how does <Command>make</Command> ever get recursively
3227 invoked with <constant>way</constant> set? There are two ways
3228 in which this happens:</para>
3232 <para>For some (but not all) of the standard targets, when
3233 in a leaf sub-directory, <Command>make</Command> is
3234 recursively invoked for each way tag in
3235 <constant>$(WAYS)</constant>. You set
3236 <constant>WAYS</constant> in the
3237 <filename>Makefile</filename> to the list of way tags you
3238 want these targets built for. The mechanism here is very
3239 much like the recursive invocation of
3240 <Command>make</Command> in sub-directories (<Xref
3241 LinkEnd="sec-subdirs">). It is up to you to set
3242 <constant>WAYS</constant> in your
3243 <filename>Makefile</filename>; this is how you control what
3244 ways will get built.</para>
3248 <para>For a useful collection of targets (such as
3249 <filename>libHS_mp.a</filename>,
3250 <filename>Foo.mp_o</filename>) there is a rule which
3251 recursively invokes <Command>make</Command> to make the
3252 specified target, setting the <constant>way</constant>
3253 variable. So if you say <Command>gmake
3254 Foo.mp_o</Command> you should see a recursive
3255 invocation <Command>gmake Foo.mp_o way=mp</Command>,
3256 and <Emphasis>in this recursive invocation the pattern rule
3257 for compiling a Haskell file into a <filename>.o</filename>
3258 file will match</Emphasis>. The key pattern rules (in
3259 <filename>suffix.mk</filename>) look like this:
3263 $(HC) $(HC_OPTS) $< -o $@
3270 <para>You can invoke <command>make</command> with a
3271 particular <literal>way</literal> setting yourself, in order
3272 to build files related to a particular
3273 <literal>way</literal> in the current directory. eg.
3279 will build files for the profiling way only in the current
3287 <Title>When the canned rule isn't right</Title>
3290 Sometimes the canned rule just doesn't do the right thing. For
3291 example, in the <Literal>nofib</Literal> suite we want the link step to print out
3292 timing information. The thing to do here is <Emphasis>not</Emphasis> to define
3293 <constant>HS_PROG</constant> or <constant>C_PROG</constant>, and instead define a special purpose rule in
3294 your own <filename>Makefile</filename>. By using different variable names you will avoid
3295 the canned rules being included, and conflicting with yours.
3302 <Sect1 id="sec-booting-from-C">
3303 <Title>Booting/porting from C (<filename>.hc</filename>) files
3305 <indexterm><primary>building GHC from .hc files</primary></indexterm>
3306 <indexterm><primary>booting GHC from .hc files</primary></indexterm>
3307 <indexterm><primary>porting GHC</primary></indexterm></Title>
3310 This section is for people trying to get GHC going by using the supplied
3311 intermediate C (<filename>.hc</filename>) files. This would probably be
3312 because no binaries have been provided, or because the machine is not ``fully
3317 The intermediate C files are normally made available together with a source
3318 release, please check the announce message for exact directions of where to
3319 find them. If we haven't made them available or you can't find them, please
3324 Assuming you've got them, unpack them on top of a fresh source tree. This
3325 will place matching <filename>.hc</filename> files next to the corresponding
3326 Haskell source in the compiler subdirectory <filename>ghc</filename> and in
3327 the language package of hslibs (i.e., in <filename>hslibs/lang</filename>).
3328 Then follow the `normal' instructions in <Xref
3329 LinkEnd="sec-building-from-source"> for setting up a build tree.
3333 The actual build process is fully automated by the
3334 <filename>hc-build</filename> script located in the
3335 <filename>distrib</filename> directory. If you eventually want to install GHC
3336 into the directory <filename>INSTALL_DIRECTORY</filename>, the following
3337 command will execute the whole build process (it won't install yet):
3340 foo% distrib/hc-build --prefix=INSTALL_DIRECTORY
3342 <indexterm><primary>--hc-build</primary></indexterm>
3344 By default, the installation directory is <filename>/usr/local</filename>. If
3345 that is what you want, you may omit the argument to
3346 <filename>hc-build</filename>. Generally, any option given to
3347 <filename>hc-build</filename> is passed through to the configuration script
3348 <filename>configure</filename>. If <filename>hc-build</filename>
3349 successfully completes the build process, you can install the resulting
3350 system, as normal, with
3357 That's the mechanics of the boot process, but, of course, if you're
3358 trying to boot on a platform that is not supported and significantly
3359 `different' from any of the supported ones, this is only the start of
3360 the adventure…(ToDo: porting tips—stuff to look out for, etc.)
3365 <Sect1 id="sec-build-pitfalls">
3366 <Title>Known pitfalls in building Glasgow Haskell
3368 <indexterm><primary>problems, building</primary></indexterm>
3369 <indexterm><primary>pitfalls, in building</primary></indexterm>
3370 <indexterm><primary>building pitfalls</primary></indexterm></Title>
3373 WARNINGS about pitfalls and known ``problems'':
3382 One difficulty that comes up from time to time is running out of space
3383 in <literal>TMPDIR</literal>. (It is impossible for the configuration stuff to
3384 compensate for the vagaries of different sysadmin approaches to temp
3386 <indexterm><primary>tmp, running out of space in</primary></indexterm>
3388 The quickest way around it is <Command>setenv TMPDIR /usr/tmp</Command><indexterm><primary>TMPDIR</primary></indexterm> or
3389 even <Command>setenv TMPDIR .</Command> (or the equivalent incantation with your shell
3392 The best way around it is to say
3395 export TMPDIR=<dir>
3398 in your <filename>build.mk</filename> file.
3399 Then GHC and the other <Literal>fptools</Literal> programs will use the appropriate directory
3408 In compiling some support-code bits, e.g., in <filename>ghc/rts/gmp</filename> and even
3409 in <filename>ghc/lib</filename>, you may get a few C-compiler warnings. We think these
3417 When compiling via C, you'll sometimes get ``warning: assignment from
3418 incompatible pointer type'' out of GCC. Harmless.
3425 Similarly, <Command>ar</Command>chiving warning messages like the following are not
3429 ar: filename GlaIOMonad__1_2s.o truncated to GlaIOMonad_
3430 ar: filename GlaIOMonad__2_2s.o truncated to GlaIOMonad_
3440 In compiling the compiler proper (in <filename>compiler/</filename>), you <Emphasis>may</Emphasis>
3441 get an ``Out of heap space'' error message. These can vary with the
3442 vagaries of different systems, it seems. The solution is simple:
3449 If you're compiling with GHC 4.00 or later, then the
3450 <Emphasis>maximum</Emphasis> heap size must have been reached. This
3451 is somewhat unlikely, since the maximum is set to 64M by default.
3452 Anyway, you can raise it with the
3453 <Option>-optCrts-M<size></Option> flag (add this flag to
3454 <constant><module>_HC_OPTS</constant>
3455 <Command>make</Command> variable in the appropriate
3456 <filename>Makefile</filename>).
3463 For GHC < 4.00, add a suitable <Option>-H</Option> flag to the <filename>Makefile</filename>, as
3472 and try again: <Command>gmake</Command>. (see <Xref LinkEnd="sec-suffix"> for information about
3473 <constant><module>_HC_OPTS</constant>.)
3475 Alternatively, just cut to the chase:
3479 % make EXTRA_HC_OPTS=-optCrts-M128M
3488 If you try to compile some Haskell, and you get errors from GCC about
3489 lots of things from <filename>/usr/include/math.h</filename>, then your GCC was
3490 mis-installed. <Command>fixincludes</Command> wasn't run when it should've been.
3492 As <Command>fixincludes</Command> is now automagically run as part of GCC installation,
3493 this bug also suggests that you have an old GCC.
3501 You <Emphasis>may</Emphasis> need to re-<Command>ranlib</Command><indexterm><primary>ranlib</primary></indexterm> your libraries (on Sun4s).
3505 % cd $(libdir)/ghc-x.xx/sparc-sun-sunos4
3506 % foreach i ( `find . -name '*.a' -print` ) # or other-shell equiv...
3508 ? # or, on some machines: ar s $i
3513 We'd be interested to know if this is still necessary.
3521 GHC's sources go through <Command>cpp</Command> before being compiled, and <Command>cpp</Command> varies
3522 a bit from one Unix to another. One particular gotcha is macro calls
3527 SLIT("Hello, world")
3531 Some <Command>cpp</Command>s treat the comma inside the string as separating two macro
3532 arguments, so you get
3536 :731: macro `SLIT' used with too many (2) args
3540 Alas, <Command>cpp</Command> doesn't tell you the offending file!
3542 Workaround: don't put weird things in string args to <Command>cpp</Command> macros.
3553 <Sect1 id="winbuild"><Title>Notes for building under Windows</Title>
3556 This section summarises how to get the utilities you need on your
3557 Win95/98/NT/2000 machine to use CVS and build GHC. Similar notes for
3558 installing and running GHC may be found in the user guide. In general,
3559 Win95/Win98 behave the same, and WinNT/Win2k behave the same.
3560 You should read the GHC installation guide sections on Windows (in the user
3561 guide) before continuing to read these notes.
3565 <Sect2><Title>Before you start</Title>
3570 Make sure that the user environment variable
3571 <constant>MAKE_MODE</constant> is set to <Literal>UNIX</Literal>. If you
3572 don't do this you get very weird messages when you type
3573 <Command>make</Command>, such as:
3575 /c: /c: No such file or directory
3581 <para>GHC uses the <emphasis>mingw</emphasis> C compiler to
3582 generate code, so you have to install that. Just pick up a mingw bundle at
3583 <ulink url="http://www.mingw.org/">http://www.mingw.org/</ulink>.
3584 We install it in <filename>c:/mingw</filename>.
3590 Install a version of GHC, and put it in your
3591 <constant>PATH</constant> (the installer tells you the path element
3592 you need to add upon completion.)
3599 Because of various hard-wired infelicities, you need to copy
3600 <Filename>bash.exe</Filename>, <Filename>perl.exe</Filename> and
3601 <Filename>cat.exe</Filename> (from Cygwin's <Filename>bin</Filename>
3602 directory), to <Filename>/bin</Filename> (discover where your Cygwin
3603 root directory is by typing <Command>mount</Command>). If
3604 <Command>/bin</Command> points to the Cygwin <Filename>bin</Filename>
3605 directory, there's no need to copy anything.
3611 By default, cygwin provides the command shell <filename>ash</filename>
3612 as <filename>sh.exe</filename>. It has a couple of 'issues', so
3613 in your <filename>/bin</filename> directory, make sure that <filename>
3614 bash.exe</filename> is also provided as <filename>sh.exe</filename>.
3620 <para> Both <command>cvs</command> and <command>ssh</command>
3621 come with Cygwin, but make sure you select them when running
3622 the Cygwin installer.
3626 <para> Check out a copy of GHC sources from
3627 the CVS repository, following the instructions above (<xref linkend="cvs-access">).
3633 <Sect2><Title>Building GHC</Title>
3639 Run <Command>autoconf</Command> both in <filename>fptools</filename>
3640 and in <filename>fptools/ghc</filename>. If you omit the latter step you'll
3641 get an error when you run <filename>./configure</filename>:
3644 creating mk/config.h
3645 mk/config.h is unchanged
3647 running /bin/sh ./configure --cache-file=.././config.cache --srcdir=.
3648 ./configure: ./configure: No such file or directory
3649 configure: error: ./configure failed for ghc
3656 You either need to add <filename>ghc</filename> to your
3657 <constant>PATH</constant> before you invoke
3658 <Command>configure</Command>, or use the <Command>configure</Command>
3659 option <option>--with-ghc=c:/ghc/ghc-some-version/bin/ghc</option>.
3663 The Windows installer for GHC tells you at the end what
3664 additions you need to make to your <constant>PATH</constant>.
3670 After <command>autoconf</command> run <command>./configure</command> in
3671 <filename>fptools/</filename> thus:
3674 ./configure --host=i386-unknown-mingw32 --with-gcc=/mingw/bin/gcc
3677 Both these options are important! It's possible to get into
3678 trouble using the wrong C compiler!
3688 <title>Building the Windows InstallShield® Installer</title>
3691 This section is intended for GHC developers only; no-one else
3692 should need to build an InstallShield.
3696 Having built a second-stage tree and done <command>make
3697 install</command> on it, open the InstallShield
3698 (<filename>.ism</filename>) file. Open the Project screen, and
3699 then the Project subfolder of the Path variables folder, and
3700 set <literal>SourceFiles</literal> to the top of your
3701 tree. You might also need to set <literal>GHCBITS</literal> to
3702 point to the tree of various external bits that are added into
3703 the IS mix. You should then be able to build an InstallShield.
3707 <title>Extra features of the InstallShield</title>
3710 The InstallShield has some IS-specific twiddles:
3715 Two registry entries are set under
3716 <literal>HKEY_LOCAL_MACHINE\SOFTWARE\GHC</literal>:
3717 <literal>Path</literal> and
3718 <literal>Version</literal>, which record respectively
3719 the directory in which GHC was installed, and the
3725 The InstallShield adds some entries to the Program
3726 menu, for GHCi and for the documentation. See under
3727 Setup Design and the individual components (each
3728 component can add entries to the menu).
3736 <title>External add-ins</title>
3739 The external add-ins consist of Mingwin gcc and Mingwin
3740 Perl. The layout of the add-ins tree is as follows:
3745 perl.exe (Mingwin perl)
3748 Mingwin gcc binaries, libraries and headers