1 the new YFLAGS code doesn't correctly handle srcdir
3 allow foo_NAME to rename an object (library or program)
6 remove _LTLIBRARIES and just use _LIBRARIES
7 then use this for zip/jar as well
9 consider `sub am_unique_error', which only prints a given error
10 message once. Use eg in lang_c_rewrite for ansi2knr error
13 investigate problems with conditionally defined libraries
15 add an error if the user makefile.am violates our
18 have 'make check' print tests which are skipped
20 we need a document describing automake from the end user's point of view
21 eg describe INSTALL_HEADER there, among other things
27 > DISTCLEAN -test -z "$(DISTCLEANFILES)" || rm -f $(DISTCLEANFILES)
29 > maintainer-clean-generic:
30 > +## FIXME: shouldn't we really print these messages before running
31 > +## the dependencies?
32 > + @echo "This command is intended for maintainers to use"
33 > + @echo "it deletes files that may require special tools to rebuild."
37 > I'd like to eventually fix the FIXME comment by having
38 > maintainer-clean look like:
44 > We're left with the question of whether we should repeat them in every
49 > Hmm... Interesting. It must have been a side effect of the enabling
50 > of forced `relink' on GNU/Linux/x86. Anyway, on platforms that
51 > actually require relinking, this problem remains, and I see no way to
52 > overcome it other than arranging for automake to install libraries
53 > before executables, as you suggest. This shouldn't be a big problem,
56 > A bigger problem could show up if two libraries in the same directory,
57 > one dependent on the other, are installed concurrently. If relinking
58 > is needed for the dependent library, we have a problem. It appears to
59 > me that user will have to live without `make -j install', in this
63 > Here's an Automake patch and changelog entry allow make -j install on
64 > such degenerate systems (and Linux with buggy libtool <g>)
66 > If you install to locations other that bin_ and lib_ then a larger fix
67 > is necessary, but this should fix the 90% case.
69 * think about how per-object flags should work. in particular:
70 * how should they be specified?
71 using the object name is confusing when .lo/.obj in use
72 however, the object name provides a nice interaction with
74 * how should they interact with per-executable flags?
75 [ this is probably a feature in search of a problem ]
77 * cross-compilation support:
78 programs built and used by the build process need to be
79 built for CC_FOR_BUILD
80 introduce a new prefxi for this, e.g. `build_PROGRAMS'
81 [ we can do this in an automatic way I think.
82 unfortunately it isn't that useful until autoconf has support
83 for this sort of thing as well ]
85 * distcheck should make sure that each file that uses _() is
88 # Verify that all source files using _() are listed in po/POTFILES.in.
90 grep -E -v '^(#|$$)' po/POTFILES.in | sort > $@-1
91 grep -E -l '\b_\(' lib/*.c src/*.c | sort > $@-2
95 * one performance enhancement would be to have autoconf write
96 a single file containing all the macro assignments.
97 then read this file via `include'
98 unfortunately this can't be done because of conditionals
100 * support prog_LIBS as override for LIBS
102 * Scan configure.in using traces as autoheader does.
103 This will be much more reliable.
105 * Test subdir-objects option with yacc, lex, ansi2knr
106 Our locking scheme won't prevent a parallel make from losing
107 if there are two `bar.o' files and the timing is just right
108 This only happens with parallel make and no-`-c -o' compiler,
109 so it probably isn't very important
110 `-c -o' when doing libtool
111 try to find a losing compiler and see if it really works.
112 (actually: hack config.cache and do it)
115 ** We're using `$<' in explicit rules when using per-exe flags
116 ** per-exe flags don't work for CPPFLAGS/YFLAGS/LFLAGS. Fix.
117 ** LIBOBJS shouldn't be used when there are per-exe flags (?)
119 * Test nodist_SOURCES with lex, yacc, etc.
121 * Support subdir-objects with fortran
123 * Allow creation of Java .zip/.jar files in natural way
124 If you are building a compiled Java library, then the .zip/.jar
125 ought to be made automatically.
127 * Run automake before libtool. It will report an error but
128 still won't put the file into the disty. This is wrong.
129 From Mark H Wilkinson <mhw@kremvax.demon.co.uk>
131 * CFLAGS only defined if C source seen
132 but really it should be a configure variable, shouldn't it?
133 There are other examples of this
134 [ moving to autoconf --trace ought to fix this ]
136 * in gnu/gnits mode, give error if Makefile.am overrides a user
137 variable like CFLAGS.
138 [ this is low priority because the package author can always
139 circumvent our check by redefining in configure.in
140 plus it is probably better to encourage good behavior than to
143 * If we see `foo.o' in LIBOBJS, and we've seen AC_OBJEXT, then complain.
144 [ how will we know that? it is better to handle this automatically
145 via an autoconf hook ]
147 * examine possibility of using any character in a macro name
148 and rewriting names automatically. this means we must rewrite
149 all references as well.
150 [ this is a 2.0-style feature ]
152 * AM_CONFIG_HEADER might generate the wrong stamp file names
153 when given multiple headers. Write a test.
155 * Currently don't correctly handle multiple inputs to a config header.
156 [ this should no matter in the future as acconfig.h and so on are
157 obsoleted by the AH series of macros.]
159 * header stamp files still in wrong dirs.
160 stamp-h.in must be in dir with h.in file
161 stamp-h must be in dir with output file
163 * conditionals and macros
164 Our current scheme cause combinatoric explosion.
166 In fact, to be honest, I no longer understand very well why we perform
167 such a closure. I mean, as is, Automake transforms (this is
170 | bin_PROGRAMS = targ
190 | targ_SOURCES = $(SONE) $(STWO) $(STHREE)
194 | @ONE_FALSE@@THREE_FALSE@@TWO_TRUE@am_targ_OBJECTS = two.$(OBJEXT)
195 | @ONE_FALSE@@THREE_FALSE@@TWO_FALSE@am_targ_OBJECTS =
196 | @ONE_FALSE@@THREE_TRUE@@TWO_TRUE@am_targ_OBJECTS = two.$(OBJEXT) \
197 | @ONE_FALSE@@THREE_TRUE@@TWO_TRUE@ three.$(OBJEXT)
198 | @ONE_FALSE@@THREE_TRUE@@TWO_FALSE@am_targ_OBJECTS = three.$(OBJEXT)
199 | @ONE_TRUE@@THREE_FALSE@@TWO_TRUE@am_targ_OBJECTS = one.$(OBJEXT) \
200 | @ONE_TRUE@@THREE_FALSE@@TWO_TRUE@ two.$(OBJEXT)
201 | @ONE_TRUE@@THREE_FALSE@@TWO_FALSE@am_targ_OBJECTS = one.$(OBJEXT)
202 | @ONE_TRUE@@THREE_TRUE@@TWO_TRUE@am_targ_OBJECTS = one.$(OBJEXT) \
203 | @ONE_TRUE@@THREE_TRUE@@TWO_TRUE@ two.$(OBJEXT) three.$(OBJEXT)
204 | @ONE_TRUE@@THREE_TRUE@@TWO_FALSE@am_targ_OBJECTS = one.$(OBJEXT) \
205 | @ONE_TRUE@@THREE_TRUE@@TWO_FALSE@ three.$(OBJEXT)
207 why don't we just output
209 | @ONE_TRUE@am_SONE_OBJECTS = one.$(OBJEXT)
210 | @ONE_FALSE@am_SONE_OBJECTS =
212 | @TWO_TRUE@am_STWO_OBJECTS = two.$(OBJEXT)
213 | @TWO_FALSE@am_STWO_OBJECTS =
215 | @THREE_TRUE@am_STHREE_OBJECTS = three.$(OBJEXT)
216 | @THREE_FALSE@am_STHREE_OBJECTS =
218 | am_targ_OBJECTS = $(am_SONE_OBJECTS) $(am_STWO_OBJECTS) $(am_STHREE_OBJECTS)
220 which means also, why do we look for the closure of PROGRAMS, instead
221 of just adding $(EXEEXT) to all its components and sub components
222 (i.e., inside sub vars such as $(SONE) above being a sub var of
226 Aaaaaaaaaaah! I think I know... Must be because of `+='.
228 Hm... No. Indeed we transform
237 | @BAR_TRUE@FOO = foo bar
238 | @BAR_FALSE@FOO = foo
240 but this seems good to me too?
242 | FOO = foo $(BAR_FOO)
243 | @BAR_TRUE@BAR_FOO = bar
244 | @BAR_FALSE@BAR_FOO =
251 ... this ought to work. The fix is probably complicated, but might
252 come for free when we rewrite the handling of conditionals.
254 * `distcheck' and `dist' should depend on `all'
256 * Add code to generate foo-config script like gnome, gtk
258 * `DEFS += foo' won't work.
259 That's because DEFS is defined in header-vars.am, which is read
260 after the user's Makefile.am.
261 This will be a problem for any macro defined internally
262 [ fixing this will probably fix the nasty `exeext redefines
263 foo_PROGRAMS' hack that is in there right now ]
264 [ we currently give an error when this occurs, so this is very low
267 * document user namespace for macro/target names
268 adopt some conventions and use uniformly
269 [ this is a good thing for the rewrite ]
271 * make distcheck uses directories like `=build'.
272 Some (very rare) POSIX systems don't support `=' in filenames.
273 If this ever becomes a problem, fix it
275 * distclean must remove config.status
276 can't this cause problems for maintainer-clean?
277 shouldn't maintainer-clean print the message before running
278 any part of the make? (just to slow things down long enough
279 for the user to stop it)
280 (maybe doesn't matter since people who even know about
281 maintainer-clean already have a clue)
283 * reintroduce AM_FUNC_FNMATCH which sets LIBOBJS
284 Then have automake know about fnmatch.h.
285 [ probably should wait for autoconf to get right functionality ]
287 * "make diff" capability
288 look at gcc's Makefile.in to see what to do
289 or look at maint program
291 * in --cygnus, clean-info not generated at top level
293 * what if an element of a scanned variable looks like
295 or some other arbitrary thing?
296 right now we try to cope, but not very well
297 [ this is only of theoretical interest for now ]
299 * make sure every variable that is used is also defined
300 [ we don't really look at variable uses in detail.
303 * make sure `missing' defines are generated
305 * missing should handle install -d and rmdir -p (for uninstall)
307 * notice when a .c file is a target somewhere, and auto-add it to
310 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
311 [ requires changes to the standard ]
313 * copyrights on m4 files, aclocal output
315 * should not put texiname_TEXINFOS into distribution
316 should rename this macro anyway, to foo_texi_DEPENDENCIES
318 * *all* installed scripts should support --version, --help
320 * For now I guess I'll just have automake give an error if it encounters
321 non-C source in a libtool library specification.
323 * must split $obj into two parts: one for libtool and one for
324 deansification. Otherwise .S files will be deansified!
326 * ansi2knr must currently appear in a directory that has some source
328 * if program has the same name as a target, do something sensible:
329 - if the target is internal, rename it
330 - if the target is mandated (eg, "info"), tell the user
331 consider auto-modifying the program name to work around this
333 * should separate actual options from strictness levels
334 strictness should only cover requirements
335 You should be able to pick and choose options
337 * rewrite in guile (RMS request)
338 at the same time, consider adding a GUI
339 could use the same parsing code for the GUI and the standalone version
340 that means figuring out a better representation of internal state
341 [ that's easy -- anything is better than what we have now ]
343 having just one Makefile for a project would give a big speed increase
344 for a project with many directories, eg glibc. ideally (?) you'd
345 still be able to have a Makefile.am in each directory somehow; this
346 might make editing conceptually easier.
348 * finish up TAGS work
350 * only remove libtool at top level?
352 * clean up source directory by moving stuff into subdirs
354 * consider adding pkglibexecdir, maybe others?
355 requests for pkg-dirs with version included
357 Avoid loops when installing; instead unroll them in automake
360 * completely handle multi-":" mode for AC_CONFIG_HEADER
361 * Scan multiple input files when Makefile is generated?
362 This would provide flexibility for large projects; subsumes
363 the "Makefile.tmpl" idea
365 [ can't do this. must explain why in manual.
366 basically, solving all the problems is too hard
367 like: how to remove redundancies between generated .in files
368 instead should implement `include' directive for Makefile.am ]
369 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
370 first input file that has a corresponding .am file.
372 Some long-term projects:
373 * if $(FOO) is used somewhere, ensure FOO is defined, either by
374 user or by automake if possible
376 [ include, += support ]
377 * even better would be allowing targets in different included
378 fragments to be merged. e.g., `install-local'.
380 consider putting all check-* targets onto @check?
381 To support --help/--version checking?
383 take diff-n-query code from libit
386 Per> 1) Being able to build a set of non-source programs
387 Per> from source programs, without necessarily linking them together.
388 Per> I.e. one should be able to say something like:
389 Per> dummy_SOURCES=foo.c bar.c
390 Per> and automake should realize that it needs to build foo.o and bar.o.
391 Per> 2) Being intelligent about new kinds of suffixes.
393 Per> SUFFIXES = .class .java
394 Per> and a suffix rule of the form:
396 Per> then it should be able to realize it can build .class files from
397 Per> .java files, and thus be able to generate a list of
398 Per> .class files from a list of .java source files.
400 !! Must fix require_file stuff. It is really gross, and I don't
401 understand it any more.
403 Jim's idea: should look for @setfilename and warn if filenames too long
406 ** many requests for a way to omit a file from the distribution.
407 Should be done like `!foo' or `~foo' in _SOURCES, etc.
408 Such files should be removed explicitly after the copy step!
409 Doing this requires rewriting macros before generating Makefile.in.
411 from joerg-martin schwarz:
412 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
413 in an explicitly written rule, you should emit the corresponding
414 Makefile variables automatically.
416 Configuring in the large:
417 * allow hierarchy of dirs to share one aclocal.m4
420 consider printing full file name of Makefile.am or configure.in when
421 giving error. This would help for very large trees with many
424 From the GNU Standards. These things could be checked, and probably
426 * Make sure that the directory into which the distribution unpacks (as
427 well as any subdirectories) are all world-writable (octal mode 777).
428 * Make sure that no file name in the distribution is more than 14
430 * Don't include any symbolic links in the distribution itself.
432 * Make sure that all the files in the distribution are world-readable.
433 ** also, check --help output and --version output. Idea from François
434 * standards no longer prohibit ANSI C. What does this imply
435 for the de-ansi-fication feature? [ must keep it -- some users rely on it ]
437 should be able to determine what is built by looking at rules (and
438 configure.in). Then built man pages (eg) could automatically be
439 omitted from the distribution.
441 Henrik Frystyk Nielsen says:
442 Henrik> 4) Flags like --include-deps are lost when you make changes to
443 Henrik> Makefile.am files and automake is run automatically. It would
444 Henrik> be nice to keep these flags as I now have to redo everything
446 ... what about other options here too?
448 Think about: maybe "make check" should just bomb if error occurs?
449 Then user must use "make -k check". This is probably more natural.
451 Consider: "cvs" option adds some cvs-specific rules?
453 Right now, targets generated internally (eg "install") are not
454 overridable by user code. This should probably be possible, even
455 though it isn't very important. This could be done by generating all
456 internal rules via a function call instead of just appending to
458 [ this will be harder to implement when scanning a rule like all-recursive
461 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
462 but which could be by running the magic make command.
465 * Must rewrite am_install_var. Should break into multiple functions.
466 This will allow the callers to be a little smarter.
467 * Rewrite clean targets.
468 * Must rewrite error handling code. Right now it is a real mess
469 Should fix up require_file junk at the same time
471 djm wants ``LINKS'' variable; list of things to link together after
472 install. In BSD environment, use:
473 LINKS = from1 to1 from2 to2 ...
475 Need way to say there are no suffixes in a Makefile (Franc,ois'
476 "override" idea suffices here)
478 Check to make sure various scripts are executable (IE when looking for
481 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
482 (why isn't this merged with "pathchk"?) when doing a dist. Do
483 whatever else François says here...
485 Add support for html via an option. Use texi2html. Use
486 "html_TEXINFOS", and htmldir = .../html. Include html files in
487 distribution. Also allow "html_DATA", for raw .html files.
488 [ when will texinfo directly support html? ]
490 uninstall and pkg-dirs should rm -rf the dir.
492 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
493 these files can't be de-ansified. Not a problem?
494 [ fix by using ansi2knr wrapper program ]
496 In general most .am files should be merged into automake. For
497 instance all the "clean" targets could be merged by keeping lists of
498 things to be removed. This would be a lot nicer looking. Note that
499 the install targets probably should not be merged; it is sometimes
500 useful to only install a small part.
502 * Clean up the output:
503 ** Order rules sensibly
504 ** Ensure every line has a purpose. Omit unused stuff
505 ** Eliminate extraneous rules when possible (eg 'install-am' stuff)
506 ** Make sure vertical spacing is correct
507 Omit program transform vars from header if no program installed. This
508 is currently pretty hard to do. (But with beautification code it
509 would probably be easy)
512 ** It would be nice to automatically support using bison's better features
513 to rename the output files. This requires autoconf support
514 ** Consider supporting syntax from autoconf "derived:source", eg:
516 for yacc and lex source
517 ** what if you use flex and the option to avoid -lfl?
520 * Multi-language support:
521 ** should have mapping of file extensions to languages
522 ** should automatically handle the linking issue (special-case C++)
523 ** must get compile rules for various languages; FORTRAN probably
524 most important unimplemented language
525 This should be integrated in some way with Per's idea.
526 Eg .f.o rules should be recognized & auto-handled in _SOURCES
527 That way any random language can be treated with C/C++ on a first-class
530 It might be cool to generate .texi dependencies by grepping for
531 @include. (If done, it should be done the same way C dependencies are
534 It would be good to check some parts of GNU standards. Already check
535 for install-sh and mkinstalldirs. What else is required to be in
536 package by GNU standards or by automake?
537 Some things for --strictness=gnits:
538 * "cd $(foo); something" is an error in a rule. Should be:
539 "cd $(foo) && something"
540 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
541 * Look for $(LN) and require AC_PROG_LN_S
543 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
545 Internationalize. [ gettext doesn't have the necessary machinery yet ]
546 am_error should use printf-style arguments (for eventual gettext scheme)
548 François says the ordering of files in a distribution should be as follows:
552 I agree, but I don't see how to implement this yet.
553 It might be easier if "derived files" is limited to those that
554 Automake itself knows about, eg output of yacc.
556 Check all source files to make sure that FSF address is up-to-date.
557 --gnits or --gnu only.
559 Merge each -vars.am file with corresponding ".am" file. Can do this
560 because of changes to &file_contents.
562 Should libexec programs have the name transform done on them?
564 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
565 together and rules are in the usual order.
567 Make the output minimal: only output definitions for variables that
571 David> To avoid comments like the one about subdirs getting buried in
572 David> the middle of a Makefile.in, how about pushing comments that
573 David> start with ### to the top of the Makefile.in (in order)? Sort
574 David> of like how Autoconf uses diversions to force initialization
575 David> code to the top of configure.
578 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
579 Karl> generally uses lowercase. Not that it matters :-).
581 ================================================================
585 probably should put each group of m4 files into a subdir owned by the
586 containing application.
588 ================================================================
594 how to use the generated makefiles
597 - NORMAL_INSTALL junk
599 what goes in AC_CONFIG_AUX_DIR
601 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
602 also a note on how a .am file is found in this case
604 rationale for avoiding
605 make CFLAGS="$CFLAGS" ...
608 a package that installs its own aclocal macros
610 write example of using automake with dejagnu
611 follow calc example in dejagnu docs
613 document which variables are actually scanned and which are not.
615 Document customary ordering of Makefile.am. From François.
617 Should include extended version of diagram from Autoconf (suggested by
620 Make a definition of the term "source"
622 document how to use Automake with CVS. Idea from Mark Galassi. Also
623 include Greg Woods' more sophisticated "cvs-dist" target.
625 document rebuilding configure. CONFIGURE_DEPENDENCIES
626 CONFIG_STATUS_DEPENDENCIES
628 -- must document all variables that are supposed
629 to be public knowledge
631 must document the targets required for integration with
632 non-automake-using subdirs
634 document the "make SHELL='/bin/sh -x'" trick for debugging
636 section on relationship to GNU make. include notes on parallel makes
640 move discussion of cygwin32, etags, mkid under other gnu tools
644 ================================================================
646 Things to do for gcc:
648 Regularize dependency generation. Add new flags:
650 -MH Generate a dummy dependency for each header file mentioned.
654 Set name of output file
656 Then automake can use -MD -MH -MT 'foo.o foo.lo' -MF .deps/...
658 ================================================================
662 * Should support standalone library along with subdir library in same
663 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
664 is not only one specd? [ add an option for this ]
666 ================================================================
670 Would it be useful to integrate in some way with the Debian package
671 building utility? Must check. maybe it would be possible to deal
672 with all the different package utilities somehow. Lately I've been
673 hearing good things about the RedHat packaging utilities. Why are
674 there so many of these? Are they fun to write or something?
675 The RedHat package utility is called RPM; see
676 ftp://ftp.redhat.com/pub/code/rpm
677 It actually has problems, like no configure script and no documentation.
679 For Cygnus it would probably be good to be able to handle the native
680 package utility on each platform. There are probably 3 or 4 of these
681 (sysv, solaris?, aix?)
683 tcl/unix/Makefile.in has some code to generate a Solaris package.
685 Automake probably can't do all of this on its own. A new tool might
688 I have some notes from a Debian developer on how the integration
691 ================================================================
693 A tool to guess what the local Makefile.am should look like:
694 (see Gord's Maint program!)
696 * Probably integrate with autoscan
697 * Use various simple rules to determine what to do:
698 * get name of top directory, sans version info
699 * search for .c files with 'main' in them
700 * if in main.c, use directory name for program
701 * if in more than one, generate multiple programs
702 * if not found, generate a library named after directory
703 * order subdir searches correctly: lib first, src last
704 * assume 'testsuite' dir means we are using dejagnu
705 * maybe be smart about reading existing Makefile.am, so tool
706 can be run for incremental changes? You could imagine:
709 autoproject --incremental
711 ================================================================
713 Stuff NOT to do, and why:
715 consider auto-including any file that matches "*.in".
716 [ no: po/Makefile.in shouldn't be included ]
718 must look at mkid to see how it works (for subdir usage)
719 [ right now, it doesn't. i don't see a simple fix right now ]
721 if configure.in not found, move up a directory and try again? This
722 could eliminate a common source of problems.
723 [ this is just a bad idea ]
725 * scripts are installed in $exec_prefix/bin, not $prefix/bin
727 [ the consensus on Gnits is that this isn't required.
728 doubters can work around it anyway ]
730 Scan source directories and warn about missing files, eg .c/.h files
731 that aren't mentioned?
732 [ distcheck makes this less useful ]
735 - how to install file with a space in its name?
736 [ don't bother with this -- make is just too losing ]