1 * Document why putting @FOO@ in _SOURCES doesn't work.
2 This must be done for 1.5
4 * if automake -a, and dangling links exist, automake should
7 * add a way to require C++ support even when no apparent C++ code
8 exists. Likewise for other langs.
10 * in depend2.am, in specialization case, what if @SOURCE@ is found
11 in srcdir? We can't depend on $<! We must search explicitly.
13 * Add a way to say that foo/bar.h should be installed as
14 $(includedir)/foo/bar.h without introducing a new dir var
16 * distcheck should make sure that each file that uses _() is
19 # Verify that all source files using _() are listed in po/POTFILES.in.
21 grep -E -v '^(#|$$)' po/POTFILES.in | sort > $@-1
22 grep -E -l '\b_\(' lib/*.c src/*.c | sort > $@-2
26 * support prog_LIBS as override for LIBS
28 * serious bug. this doesn't work:
31 why did I think it did work?
33 * Sometimes we can warn about a missing `compile' script twice.
35 * A target named `INSTALL' will screw things up.
38 * Scan configure.in using the same trick that autoheader uses.
39 This will be much more reliable.
41 * Test subdir-objects option with yacc, lex, ansi2knr
42 Our locking scheme won't prevent a parallel make from losing
43 if there are two `bar.o' files and the timing is just right
44 This only happens with parallel make and no-`-c -o' compiler,
45 so it probably isn't very important
46 `-c -o' when doing libtool
47 try to find a losing compiler and see if it really works.
48 (actually: hack config.cache and do it)
50 * We're using `$<' in explicit rules when using per-exe flags
51 per-exe flags don't work for CPPFLAGS/YFLAGS/LFLAGS. Fix.
52 LIBOBJS shouldn't be used when there are per-exe flags (?)
54 * Need a way to pass flags to makeinfo
57 * test `make clean' with subdir-objects
59 * Test nodist_SOURCES with lex, yacc, etc.
61 * Support subdir-objects with fortran
63 * Allow creation of Java .zip/.jar files in natural way
64 If you are building a compiled Java library, then the .zip/.jar
65 ought to be made automatically.
67 * Run automake before libtool. It will report an error but
68 still won't put the file into the disty. This is wrong.
69 From Mark H Wilkinson <mhw@kremvax.demon.co.uk>
71 * CFLAGS only defined if C source seen
72 but really it should be a configure variable, shouldn't it?
73 There are other examples of this
75 * in gnu/gnits mode, give error if Makefile.am overrides a user
78 * If we see `foo.o' in LIBOBJS, and we've seen AC_OBJEXT, then complain.
80 * using "include" inside a conditional doesn't currently
83 * examine possibility of using any character in a macro name
84 and rewriting names automatically. this means we must rewrite
85 all references as well.
87 * AM_CONFIG_HEADER might generate the wrong stamp file names
88 when given multiple headers. Write a test.
90 * Currently don't correctly handle multiple inputs to a config header.
92 * header stamp files still in wrong dirs.
93 stamp-h.in must be in dir with h.in file
94 stamp-h must be in dir with output file
100 ... this ought to work. The fix is probably complicated
102 * `distcheck' and `dist' should depend on `all'
104 * Add code to generate foo-config script like gnome, gtk
106 * `DEFS += foo' won't work.
107 That's because DEFS is defined in header-vars.am, which is read
108 after the user's Makefile.am.
109 This will be a problem for any macro defined internally
110 [ fixing this will probably fix the nasty `exeext redefines
111 foo_PROGRAMS' hack that is in there right now ]
112 In some cases this is unfixable. In these cases we should give
113 an error. This can be done by keeping track of whether a macro
114 was assigned with `=' or `+=', and having define_variable give
115 an error if the variable exists and was defined with +=.
117 * document user namespace for macro/target names
118 adopt some conventions and use uniformly
119 [ this is a good thing for the rewrite ]
121 * make distcheck uses directories like `=build'.
122 Some (very rare) POSIX systems don't support `=' in filenames.
123 If this ever becomes a problem, fix it
125 * distclean must remove config.status
126 can't this cause problems for maintainer-clean?
127 shouldn't maintainer-clean print the message before running
128 any part of the make? (just to slow things down long enough
129 for the user to stop it)
130 (maybe doesn't matter since people who even know about
131 maintainer-clean already have a clue)
133 * There are probably more bugs in variable_conditions_sub along
134 the lines of the one that caused cond4.test to fail.
136 * give user more control over -I flags
137 in particular document a way to override the defaults
139 * reintroduce AM_FUNC_FNMATCH which sets LIBOBJS
140 Then have automake know about fnmatch.h.
141 [ probably should wait for autoconf to get right functionality ]
143 * Add a conditional for dependency tracking
144 (what to name it is the biggest problem here)
145 (because we want it to flag dist/no-dist -- not just deps)
146 [ this might not really be correctly doable.
147 instead we need a compile-time conditional for this
150 * Allow per-object cflags:
153 * per-object compiler flags do not apply to libobjs
154 -> give error in this case
155 * At the same time, allow sources in subdirs:
157 This requires `mkdir x' at build time
158 [ both of these require per-file rules, and not pattern rules ]
159 [ use user-written suffix rules to generate the per-file rules in
160 an automatic way -- this would be mucho cool ]
162 * Allow for multiple translations of a texinfo file:
163 LL_info_TEXINFOS = ...
164 will put info files for language LL into $(infodir)/LL.
166 * If you suppress an internal variable by specifying a variable
167 in a Makefile.am, but the variable is conditional, then automake
168 should generate the internal variable conditionally.
169 You have to scan the map of all conditions and fill in the holes here
171 * Every program foo has FOOFLAGS right now.
172 It should also have AM_FOOFLAGS, which can be set in Makefile.am.
173 DONE: but needs to be documented
175 * Should be able to update files that would be installed with -a
178 * "make diff" capability
179 look at gcc's Makefile.in to see what to do
180 or look at maint program
182 * Karl wants to be able to set LIBS and LDFLAGS at build time, like CFLAGS
183 maybe we need something more general?
185 * add $(srcdir)/ before some dependencies?
187 * define LINK if a program is mentioned, even if no C sources appear
189 * BUILT_SOURCES should not be distributed, even when they appear in
190 another _SOURCES line. [? or maybe just leave this up to the
191 to-be-defined generic distribution method ]
192 must completely revisit the entire BUILT_SOURCES idea
194 * in --cygnus, clean-info not generated at top level
196 * what if an element of a scanned variable looks like
198 or some other arbitrary thing?
199 right now we try to cope, but not very well
201 * if `interlock' exists, that should be an error (?)
202 should also warn about using new ylwrap and not old one
203 only do this when looking for ylwrap
205 ** make sure every variable that is used is also defined
207 * make sure `missing' defines are generated
208 * if no AM_INIT_AUTOMAKE, then don't handle `missing' stuff.
210 * missing should handle install -d and rmdir -p (for uninstall)
212 * a couple ways to be smarter:
213 - notice when a .c file is a target somewhere, and auto-add it to
215 - notice a target of the form `.x.y:' and assume it is a suffix rule
218 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
219 [ requires changes to the standard ]
221 * cross-compilation support:
222 programs built and used by the build process need to be
223 built for CC_FOR_BUILD
224 introduce a new variable for this
226 * if foo.y is a source, foo.h isn't auto-distributed?
228 * copyrights on m4 files, aclocal output
230 * is there a way to add a directory and then have "make" do all the
233 * put standards.texi into distribution
236 * should not put texiname_TEXINFOS into distribution
237 should rename this macro anyway, to foo_texi_DEPENDENCIES
239 * *all* installed scripts should support --version, --help
241 For now I guess I'll just have automake give an error if it encounters
242 non-C source in a libtool library specification.
244 * must split $obj into two parts: one for libtool and one for
245 deansification. Otherwise .S files will be deansified!
247 * ansi2knr must currently appear in a directory that has some source
249 * if program has the same name as a target, do something sensible:
250 - if the target is internal, rename it
251 - if the target is mandated (eg, "info"), tell the user
252 consider auto-modifying the program name to work around this
254 * should separate actual options from strictness levels
255 strictness should only cover requirements
256 You should be able to pick and choose options
258 should clean up texinfos.am; one rule is repeated 3 times, but
261 should always use perl -w
263 rewrite in guile (RMS request)
264 at the same time, consider adding a GUI
265 could use the same parsing code for the GUI and the standalone version
266 that means figuring out a better representation of internal state
267 [ that's easy -- anything is better than what we have now ]
269 having just one Makefile for a project would give a big speed increase
270 for a project with many directories, eg glibc. ideally (?) you'd
271 still be able to have a Makefile.am in each directory somehow; this
272 might make editing conceptually easier.
274 * finish up TAGS work
276 * put parser.h into distribution if "yacc -d" is used
278 * only remove libtool at top level?
280 * clean up source directory by moving stuff into subdirs
282 * consider adding pkglibexecdir, maybe others?
283 requests for pkg-dirs with version included
285 Avoid loops when installing; instead unroll them in automake
288 * completely handle multi-":" mode for AC_CONFIG_HEADER
289 * Scan multiple input files when Makefile is generated?
290 This would provide flexibility for large projects; subsumes
291 the "Makefile.tmpl" idea
293 [ can't do this. must explain why in manual.
294 basically, solving all the problems is too hard
295 like: how to remove redundancies between generated .in files
296 instead should implement `include' directive for Makefile.am ]
297 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
298 first input file that has a corresponding .am file.
300 Some long-term projects:
301 * if $(FOO) is used somewhere, ensure FOO is defined, either by
302 user or by automake if possible
304 [ include, += support ]
305 * even better would be allowing targets in different included
306 fragments to be merged. e.g., `install-local'.
308 consider putting all check-* targets onto @check?
309 To support --help/--version checking?
311 take diff-n-query code from libit
314 Per> 1) Being able to build a set of non-source programs
315 Per> from source programs, without necessarily linking them together.
316 Per> I.e. one should be able to say something like:
317 Per> dummy_SOURCES=foo.c bar.c
318 Per> and automake should realize that it needs to build foo.o and bar.o.
319 Per> 2) Being intelligent about new kinds of suffixes.
321 Per> SUFFIXES = .class .java
322 Per> and a suffix rule of the form:
324 Per> then it should be able to realize it can build .class files from
325 Per> .java files, and thus be able to generate a list of
326 Per> .class files from a list of .java source files.
328 !! Must fix require_file stuff. It is really gross, and I don't
329 understand it any more.
331 * error messages should print ``[info blah blah]'' command when a
332 certain part of the standards apply. saw idea in message from
333 Craig Burley. wouldn't it be really cool if compile-mode in Emacs
334 understood this convention, and you could click on such text to
335 go to the appropriate info page? [ I think you can ]
337 Jim's idea: should look for @setfilename and warn if filenames too long
340 ** many requests for a way to omit a file from the distribution.
341 Should be done like `!foo' or `~foo' in _SOURCES, etc.
342 Such files should be removed explicitly after the copy step!
343 Doing this requires rewriting macros before generating Makefile.in.
345 from joerg-martin schwarz:
346 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
347 in an explicitly written rule, you should emit the corresponding
348 Makefile variables automatically.
350 Configuring in the large:
351 * allow hierarchy of dirs to share one aclocal.m4
354 consider printing full file name of Makefile.am or configure.in when
355 giving error. This would help for very large trees with many
358 From the GNU Standards. These things could be checked, and probably
360 * Make sure that the directory into which the distribution unpacks (as
361 well as any subdirectories) are all world-writable (octal mode 777).
362 * Make sure that no file name in the distribution is more than 14
364 * Don't include any symbolic links in the distribution itself.
366 * Make sure that all the files in the distribution are world-readable.
367 ** also, check --help output and --version output. Idea from François
368 * standards no longer prohibit ANSI C. What does this imply
369 for the de-ansi-fication feature? [ must keep it -- some users rely on it ]
371 should be able to determine what is built by looking at rules (and
372 configure.in). Then built man pages (eg) could automatically be
373 omitted from the distribution.
375 Idea from Joerg-Martin Schwarz: allow passing different -D flags to
376 different compiles. This can be done, but with the restriction that a
377 .c cannot appear in 2 different "objects" (programs/libraries)
378 compiled with different -D options (because -c and -o do not always
379 work together and parallel makes must work). This could be
380 implemented by noticing whenever a ".o" target with no rules is being
381 emitted, and adding the appropriate compilation rule as appropriate.
382 This should work with targets from Makefile.am as well as from .P
383 files, which means rewriting so that the Makefile.am contents aren't
384 copied into the output immediately.
385 [ this could be probably done more directly by examining the sources
386 as we scan Makefile.am ]
388 Henrik Frystyk Nielsen says:
389 Henrik> 4) Flags like --include-deps are lost when you make changes to
390 Henrik> Makefile.am files and automake is run automatically. It would
391 Henrik> be nice to keep these flags as I now have to redo everything
393 ... what about other options here too?
395 Think about: maybe "make check" should just bomb if error occurs?
396 Then user must use "make -k check". This is probably more natural.
398 Consider: "cvs" option adds some cvs-specific rules?
400 Right now, targets generated internally (eg "install") are not
401 overridable by user code. This should probably be possible, even
402 though it isn't very important. This could be done by generating all
403 internal rules via a function call instead of just appending to
405 [ this will be harder to implement when scanning a rule like all-recursive
408 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
409 but which could be by running the magic make command.
412 * Must rewrite am_install_var. Should break into multiple functions.
413 This will allow the callers to be a little smarter.
414 * Rewrite clean targets.
415 * Must rewrite error handling code. Right now it is a real mess
416 Should fix up require_file junk at the same time
418 djm wants ``LINKS'' variable; list of things to link together after
419 install. In BSD environment, use:
420 LINKS = from1 to1 from2 to2 ...
422 Need way to say there are no suffixes in a Makefile (Franc,ois'
423 "override" idea suffices here)
425 Check to make sure various scripts are executable (IE when looking for
428 Use recode in dist target when MAINT_CHARSET specified. Read caveats
429 in automake.in before doing this. Note the same problem used to apply
430 to the no-dependencies option; maybe it still should? Note also that
431 each Makefile.am must be rewritten at "make dist" time if
432 MAINT_CHARSET and DIST_CHARSET are not identical. NOTE: gettext must
433 arrange for all .po files not to be recoded. In the long term this
434 might be a problem (consider when some systems use Unicode but the
436 MAINT_CHARSET *must* be local to each Makefile.am, to enable
437 merged distributions.
438 DIST_CHARSET must be passed down to subdir makes during a "make dist"
440 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
441 (why isn't this merged with "pathchk"?) when doing a dist. Do
442 whatever else François says here...
444 Add support for html via an option. Use texi2html. Use
445 "html_TEXINFOS", and htmldir = .../html. Include html files in
446 distribution. Also allow "html_DATA", for raw .html files.
447 [ when will texinfo directly support html? ]
449 uninstall and pkg-dirs should rm -rf the dir.
451 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
452 these files can't be de-ansified. Not a problem?
453 [ fix by using ansi2knr wrapper program ]
455 In general most .am files should be merged into automake. For
456 instance all the "clean" targets could be merged by keeping lists of
457 things to be removed. This would be a lot nicer looking. Note that
458 the install targets probably should not be merged; it is sometimes
459 useful to only install a small part.
462 * Order rules sensibly
463 * Ensure every line has a purpose. Omit unused stuff
464 * Eliminate extraneous rules when possible (eg 'install-am' stuff)
465 * Make sure vertical spacing is correct
466 Omit program transform vars from header if no program installed. This
467 is currently pretty hard to do. (But with beautification code it
468 would probably be easy)
471 * It would be nice to automatically support using bison's better features
472 to rename the output files. This requires autoconf support
473 * Consider supporting syntax from autoconf "derived:source", eg:
475 for yacc and lex source
476 * what if you use flex and the option to avoid -lfl?
479 Multi-language support:
480 * should have mapping of file extensions to languages
481 * should automatically handle the linking issue (special-case C++)
482 * must get compile rules for various languages; FORTRAN probably
483 most important unimplemented language
484 This should be integrated in some way with Per's idea.
485 Eg .f.o rules should be recognized & auto-handled in _SOURCES
486 That way any random language can be treated with C/C++ on a first-class
489 It might be cool to generate .texi dependencies by grepping for
490 @include. (If done, it should be done the same way C dependencies are
493 It would be good to check some parts of GNU standards. Already check
494 for install-sh and mkinstalldirs. What else is required to be in
495 package by GNU standards or by automake?
496 Some things for --strictness=gnits:
497 * "cd $(foo); something" is an error in a rule. Should be:
498 "cd $(foo) && something"
499 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
500 * Look for $(LN) and require AC_PROG_LN_S
502 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
504 Internationalize. [ gettext doesn't have the necessary machinery yet ]
505 am_error should use printf-style arguments (for eventual gettext scheme)
507 François says the ordering of files in a distribution should be as follows:
511 I agree, but I don't see how to implement this yet.
512 It might be easier if "derived files" is limited to those that
513 Automake itself knows about, eg output of yacc.
515 Check all source files to make sure that FSF address is up-to-date.
516 --gnits or --gnu only.
518 Merge each -vars.am file with corresponding ".am" file. Can do this
519 because of changes to &file_contents.
521 Should libexec programs have the name transform done on them?
523 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
524 together and rules are in the usual order.
526 Make the output minimal: only output definitions for variables that
530 David> To avoid comments like the one about subdirs getting buried in
531 David> the middle of a Makefile.in, how about pushing comments that
532 David> start with ### to the top of the Makefile.in (in order)? Sort
533 David> of like how Autoconf uses diversions to force initialization
534 David> code to the top of configure.
537 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
538 Karl> generally uses lowercase. Not that it matters :-).
540 ================================================================
544 probably should put each group of m4 files into a subdir owned by the
545 containing application.
547 ================================================================
553 how to use the generated makefiles
556 - NORMAL_INSTALL junk
558 what goes in AC_CONFIG_AUX_DIR
560 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
561 also a note on how a .am file is found in this case
563 rationale for avoiding
564 make CFLAGS="$CFLAGS" ...
567 a package that installs its own aclocal macros
569 write example of using automake with dejagnu
570 follow calc example in dejagnu docs
572 document which variables are actually scanned and which are not.
574 Document customary ordering of Makefile.am. From François.
576 Should include extended version of diagram from Autoconf (suggested by
579 Make a definition of the term "source"
581 document how to use Automake with CVS. Idea from Mark Galassi. Also
582 include Greg Woods' more sophisticated "cvs-dist" target.
584 document rebuilding configure. CONFIGURE_DEPENDENCIES
585 CONFIG_STATUS_DEPENDENCIES
587 -- must document all variables that are supposed
588 to be public knowledge
590 must document the targets required for integration with
591 non-automake-using subdirs
593 document the "make SHELL='/bin/sh -x'" trick for debugging
595 section on relationship to GNU make. include notes on parallel makes
599 move discussion of cygwin32, etags, mkid under other gnu tools
603 ================================================================
605 Things to do for gcc:
607 Regularize dependency generation. Add new flags:
609 -MH Generate a dummy dependency for each header file mentioned.
613 Set name of output file
615 Then automake can use -MD -MH -MT 'foo.o foo.lo' -MF .deps/...
617 ================================================================
619 Things to do for autoconf:
621 * patch autoreconf to run automake and aclocal. I've done this but it is
622 not really available. It can't be made available until automake
623 is officially released
625 ================================================================
629 * Should support standalone library along with subdir library in same
630 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
631 is not only one specd? [ add an option for this ]
633 ================================================================
637 Would it be useful to integrate in some way with the Debian package
638 building utility? Must check. maybe it would be possible to deal
639 with all the different package utilities somehow. Lately I've been
640 hearing good things about the RedHat packaging utilities. Why are
641 there so many of these? Are they fun to write or something?
642 The RedHat package utility is called RPM; see
643 ftp://ftp.redhat.com/pub/code/rpm
644 It actually has problems, like no configure script and no documentation.
646 For Cygnus it would probably be good to be able to handle the native
647 package utility on each platform. There are probably 3 or 4 of these
648 (sysv, solaris?, aix?)
650 tcl/unix/Makefile.in has some code to generate a Solaris package.
652 Automake probably can't do all of this on its own. A new tool might
655 I have some notes from a Debian developer on how the integration
658 ================================================================
660 A tool to guess what the local Makefile.am should look like:
661 (see Gord's Maint program!)
663 * Probably integrate with autoscan
664 * Use various simple rules to determine what to do:
665 * get name of top directory, sans version info
666 * search for .c files with 'main' in them
667 * if in main.c, use directory name for program
668 * if in more than one, generate multiple programs
669 * if not found, generate a library named after directory
670 * order subdir searches correctly: lib first, src last
671 * assume 'testsuite' dir means we are using dejagnu
672 * maybe be smart about reading existing Makefile.am, so tool
673 can be run for incremental changes? You could imagine:
676 autoproject --incremental
678 ================================================================
680 Stuff NOT to do, and why:
682 consider auto-including any file that matches "*.in".
683 [ no: po/Makefile.in shouldn't be included ]
685 must look at mkid to see how it works (for subdir usage)
686 [ right now, it doesn't. i don't see a simple fix right now ]
688 if configure.in not found, move up a directory and try again? This
689 could eliminate a common source of problems.
690 [ this is just a bad idea ]
692 * scripts are installed in $exec_prefix/bin, not $prefix/bin
694 [ the consensus on Gnits is that this isn't required.
695 doubters can work around it anyway ]
697 * make the auto-dep code crash if GNU make not in use?
698 (doesn't it already?)
700 Looked at a program called 'ezmake', which seems to do something
701 similar. The only idea there that is possibly worth stealing is using
702 globs in definitions. Also has negations. Eg in a directory with
703 files a.c, b.c and c.c, the line:
704 foo_SOURCES = *.c ~c.c
705 would be equivalent to:
706 foo_SOURCES = a.c b.c
707 Is this worth implementing?
708 [ No... it is more reliable to spell everything out. ]
710 Scan source directories and warn about missing files, eg .c/.h files
711 that aren't mentioned?
712 [ distcheck makes this less useful ]
715 - how to install file with a space in its name?
716 [ don't bother with this -- make is just too losing ]