1 * Run automake before libtool. It will report an error but
2 still won't put the file into the disty. This is wrong.
3 From Mark H Wilkinson <mhw@kremvax.demon.co.uk>
5 * AM_CONFIG_HEADER might generate the wrong stamp file names
6 when given multiple headers. Write a test.
8 * Currently don't correctly handle multiple inputs to a config header.
10 * header stamp files still in wrong dirs.
11 stamp-h.in must be in dir with h.in file
12 stamp-h must be in dir with output file
18 ... this ought to work. The fix is probably complicated
20 * `distcheck' and `dist' should depend on `all'
22 * Document why putting @FOO@ in _SOURCES doesn't work.
24 * Add code to generate foo-config script like gnome, gtk
26 * right now automake sets `TAR' when automake is configured
27 this loses in some situations.
28 however, checking for it in every configure.in also seems lame.
29 probably should just give in on this; meanwhile people can
30 override TAR themselves.
32 * `DEFS += foo' won't work.
33 That's because DEFS is defined in header-vars.am, which is read
34 after the user's Makefile.am.
35 This will be a problem for any macro defined internally
36 [ fixing this will probabl fix the nasty `exeext redefines
37 foo_PROGRAMS' hack that is in there right now ]
39 * document user namespace for macro/target names
40 adopt some conventions and use uniformly
41 [ this is a good thing for the rewrite ]
43 * make distcheck uses directories like `=build'.
44 Some (very rare) POSIX systems don't support `=' in filenames.
45 If this ever becomes a problem, fix it
47 * distclean must remove config.status
48 can't this cause problems for maintainer-clean?
49 shouldn't maintainer-clean print the message before running
50 any part of the make? (just to slow things down long enough
51 for the user to stop it)
52 (maybe doesn't matter since people who even know about
53 maintainer-clean already have a clue)
55 * There are probably more bugs in variable_conditions_sub along
56 the lines of the one that caused cond4.test to fail.
58 * give user more control over -I flags
59 in particular document a way to override the defaults
61 * reintroduce AM_FUNC_FNMATCH which sets LIBOBJS
62 Then have automake know about fnmatch.h.
63 [ probably should wait for autoconf to get right functionality ]
65 * Add a conditional for dependency tracking
66 (what to name it is the biggest problem here)
67 (because we want it to flag dist/no-dist -- not just deps)
68 [ this might not really be correctly doable.
69 instead we need a compile-time conditional for this
72 * Allow per-object cflags:
75 * At the same time, allow sources in subdirs:
77 This requires `mkdir x' at build time
78 [ both of these require per-file rules, and not pattern rules ]
79 [ use user-written suffix rules to generate the per-file rules in
80 an automatic way -- this would be mucho cool ]
82 * Allow for multiple translations of a texinfo file:
83 LL_info_TEXINFOS = ...
84 will put info files for language LL into $(infodir)/LL.
86 * If you suppress an internal variable by specifying a variable
87 in a Makefile.am, but the variable is conditional, then automake
88 should generate the internal variable conditionally.
89 You have to scan the map of all conditions and fill in the holes here
91 * Every program foo has FOOFLAGS right now.
92 It should also have AM_FOOFLAGS, which can be set in Makefile.am.
93 DONE: but needs to be documented
95 * Should be able to update files that would be installed with -a
98 * "make diff" capability
99 look at gcc's Makefile.in to see what to do
100 or look at maint program
102 * Karl wants to be able to set LIBS and LDFLAGS at build time, like CFLAGS
103 maybe we need something more general?
105 * add $(srcdir)/ before some dependencies?
107 * define LINK if a program is mentioned, even if no C sources appear
109 * BUILT_SOURCES should not be distributed, even when they appear in
110 another _SOURCES line. [? or maybe just leave this up to the
111 to-be-defined generic distribution method ]
112 must completely revisit the entire BUILT_SOURCES idea
114 * in --cygnus, clean-info not generated at top level
116 * what if an element of a scanned variable looks like
118 or some other arbitrary thing?
119 right now we try to cope, but not very well
121 * if `interlock' exists, that should be an error (?)
122 should also warn about using new ylwrap and not old one
123 only do this when looking for ylwrap
125 ** make sure every variable that is used is also defined
127 * make sure `missing' defines are generated
128 * if no AM_INIT_AUTOMAKE, then don't handle `missing' stuff.
130 * missing should handle install -d and rmdir -p (for uninstall)
132 * a couple ways to be smarter:
133 - notice when a .c file is a target somewhere, and auto-add it to
135 - notice a target of the form `.x.y:' and assume it is a suffix rule
138 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
139 [ requires changes to the standard ]
141 * cross-compilation support:
142 programs built and used by the build process need to be
143 built for CC_FOR_BUILD
144 introduce a new variable for this
146 * if foo.y is a source, foo.h isn't auto-distributed?
148 * copyrights on m4 files, aclocal output
150 * is there a way to add a directory and then have "make" do all the
153 * put standards.texi into distribution
156 * should not put texiname_TEXINFOS into distribution
157 should rename this macro anyway, to foo_texi_DEPENDENCIES
159 * *all* installed scripts should support --version, --help
161 For now I guess I'll just have automake give an error if it encounters
162 non-C source in a libtool library specification.
164 * must split $obj into two parts: one for libtool and one for
165 deansification. Otherwise .S files will be deansified!
167 * ansi2knr must currently appear in a directory that has some source
169 * if program has the same name as a target, do something sensible:
170 - if the target is internal, rename it
171 - if the target is mandated (eg, "info"), tell the user
172 consider auto-modifying the program name to work around this
174 * should separate actual options from strictness levels
175 strictness should only cover requirements
176 You should be able to pick and choose options
178 should clean up texinfos.am; one rule is repeated 3 times, but
181 should always use perl -w
183 rewrite in guile (RMS request)
184 at the same time, consider adding a GUI
185 could use the same parsing code for the GUI and the standalone version
186 that means figuring out a better representation of internal state
187 [ that's easy -- anything is better than what we have now ]
189 having just one Makefile for a project would give a big speed increase
190 for a project with many directories, eg glibc. ideally (?) you'd
191 still be able to have a Makefile.am in each directory somehow; this
192 might make editing conceptually easier.
194 * finish up TAGS work
196 * put parser.h into distribution if "yacc -d" is used
198 * only remove libtool at top level?
200 * clean up source directory by moving stuff into subdirs
202 * consider adding pkglibexecdir, maybe others?
203 requests for pkg-dirs with version included
205 Avoid loops when installing; instead unroll them in automake
208 * completely handle multi-":" mode for AC_CONFIG_HEADER
209 * Scan multiple input files when Makefile is generated?
210 This would provide flexibility for large projects; subsumes
211 the "Makefile.tmpl" idea
213 [ can't do this. must explain why in manual.
214 basically, solving all the problems is too hard
215 like: how to remove redundancies between generated .in files
216 instead should implement `include' directive for Makefile.am ]
217 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
218 first input file that has a corresponding .am file.
220 Some long-term projects:
221 * if $(FOO) is used somewhere, ensure FOO is defined, either by
222 user or by automake if possible
224 [ include, += support ]
225 * even better would be allowing targets in different included
226 fragments to be merged. e.g., `install-local'.
228 consider putting all check-* targets onto @check?
229 To support --help/--version checking?
231 take diff-n-query code from libit
234 Per> 1) Being able to build a set of non-source programs
235 Per> from source programs, without necessarily linking them together.
236 Per> I.e. one should be able to say something like:
237 Per> dummy_SOURCES=foo.c bar.c
238 Per> and automake should realize that it needs to build foo.o and bar.o.
239 Per> 2) Being intelligent about new kinds of suffixes.
241 Per> SUFFIXES = .class .java
242 Per> and a suffix rule of the form:
244 Per> then it should be able to realize it can build .class files from
245 Per> .java files, and thus be able to generate a list of
246 Per> .class files from a list of .java source files.
248 !! Must fix require_file stuff. It is really gross, and I don't
249 understand it any more.
251 * error messages should print ``[info blah blah]'' command when a
252 certain part of the standards apply. saw idea in message from
253 Craig Burley. wouldn't it be really cool if compile-mode in Emacs
254 understood this convention, and you could click on such text to
255 go to the appropriate info page? [ I think you can ]
257 Jim's idea: should look for @setfilename and warn if filenames too long
260 ** many requests for a way to omit a file from the distribution.
261 Should be done like `!foo' or `~foo' in _SOURCES, etc.
262 Such files should be removed explicitly after the copy step!
263 Doing this requires rewriting macros before generating Makefile.in.
265 from joerg-martin schwarz:
266 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
267 in an explicitly written rule, you should emit the corresponding
268 Makefile variables automatically.
270 Configuring in the large:
271 * allow hierarchy of dirs to share one aclocal.m4
274 consider printing full file name of Makefile.am or configure.in when
275 giving error. This would help for very large trees with many
278 From the GNU Standards. These things could be checked, and probably
280 * Make sure that the directory into which the distribution unpacks (as
281 well as any subdirectories) are all world-writable (octal mode 777).
282 * Make sure that no file name in the distribution is more than 14
284 * Don't include any symbolic links in the distribution itself.
286 * Make sure that all the files in the distribution are world-readable.
287 ** also, check --help output and --version output. Idea from François
288 * standards no longer prohibit ANSI C. What does this imply
289 for the de-ansi-fication feature? [ must keep it -- some users rely on it ]
291 should be able to determine what is built by looking at rules (and
292 configure.in). Then built man pages (eg) could automatically be
293 omitted from the distribution.
295 Idea from Joerg-Martin Schwarz: allow passing different -D flags to
296 different compiles. This can be done, but with the restriction that a
297 .c cannot appear in 2 different "objects" (programs/libraries)
298 compiled with different -D options (because -c and -o do not always
299 work together and parallel makes must work). This could be
300 implemented by noticing whenever a ".o" target with no rules is being
301 emitted, and adding the appropriate compilation rule as appropriate.
302 This should work with targets from Makefile.am as well as from .P
303 files, which means rewriting so that the Makefile.am contents aren't
304 copied into the output immediately.
305 [ this could be probably done more directly by examining the sources
306 as we scan Makefile.am ]
308 Henrik Frystyk Nielsen says:
309 Henrik> 4) Flags like --include-deps are lost when you make changes to
310 Henrik> Makefile.am files and automake is run automatically. It would
311 Henrik> be nice to keep these flags as I now have to redo everything
313 ... what about other options here too?
315 Think about: maybe "make check" should just bomb if error occurs?
316 Then user must use "make -k check". This is probably more natural.
318 Consider: "cvs" option adds some cvs-specific rules?
320 Right now, targets generated internally (eg "install") are not
321 overridable by user code. This should probably be possible, even
322 though it isn't very important. This could be done by generating all
323 internal rules via a function call instead of just appending to
325 [ this will be harder to implement when scanning a rule like all-recursive
328 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
329 but which could be by running the magic make command.
332 * Must rewrite am_install_var. Should break into multiple functions.
333 This will allow the callers to be a little smarter.
334 * Rewrite clean targets.
335 * Must rewrite error handling code. Right now it is a real mess
336 Should fix up require_file junk at the same time
338 djm wants ``LINKS'' variable; list of things to link together after
339 install. In BSD environment, use:
340 LINKS = from1 to1 from2 to2 ...
342 Need way to say there are no suffixes in a Makefile (Franc,ois'
343 "override" idea suffices here)
345 Check to make sure various scripts are executable (IE when looking for
348 Use recode in dist target when MAINT_CHARSET specified. Read caveats
349 in automake.in before doing this. Note the same problem used to apply
350 to the no-dependencies option; maybe it still should? Note also that
351 each Makefile.am must be rewritten at "make dist" time if
352 MAINT_CHARSET and DIST_CHARSET are not identical. NOTE: gettext must
353 arrange for all .po files not to be recoded. In the long term this
354 might be a problem (consider when some systems use Unicode but the
356 MAINT_CHARSET *must* be local to each Makefile.am, to enable
357 merged distributions.
358 DIST_CHARSET must be passed down to subdir makes during a "make dist"
360 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
361 (why isn't this merged with "pathchk"?) when doing a dist. Do
362 whatever else François says here...
364 Add support for html via an option. Use texi2html. Use
365 "html_TEXINFOS", and htmldir = .../html. Include html files in
366 distribution. Also allow "html_DATA", for raw .html files.
367 [ when will texinfo directly support html? ]
369 uninstall and pkg-dirs should rm -rf the dir.
371 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
372 these files can't be de-ansified. Not a problem?
373 [ fix by using ansi2knr wrapper program ]
375 In general most .am files should be merged into automake. For
376 instance all the "clean" targets could be merged by keeping lists of
377 things to be removed. This would be a lot nicer looking. Note that
378 the install targets probably should not be merged; it is sometimes
379 useful to only install a small part.
382 * Order rules sensibly
383 * Ensure every line has a purpose. Omit unused stuff
384 * Eliminate extraneous rules when possible (eg 'install-am' stuff)
385 * Make sure vertical spacing is correct
386 Omit program transform vars from header if no program installed. This
387 is currently pretty hard to do. (But with beautification code it
388 would probably be easy)
391 * It would be nice to automatically support using bison's better features
392 to rename the output files. This requires autoconf support
393 * Consider supporting syntax from autoconf "derived:source", eg:
395 for yacc and lex source
396 * what if you use flex and the option to avoid -lfl?
399 Multi-language support:
400 * should have mapping of file extensions to languages
401 * should automatically handle the linking issue (special-case C++)
402 * must get compile rules for various languages; FORTRAN probably
403 most important unimplemented language
404 This should be integrated in some way with Per's idea.
405 Eg .f.o rules should be recognized & auto-handled in _SOURCES
406 That way any random language can be treated with C/C++ on a first-class
409 It might be cool to generate .texi dependencies by grepping for
410 @include. (If done, it should be done the same way C dependencies are
413 It would be good to check some parts of GNU standards. Already check
414 for install-sh and mkinstalldirs. What else is required to be in
415 package by GNU standards or by automake?
416 Some things for --strictness=gnits:
417 * "cd $(foo); something" is an error in a rule. Should be:
418 "cd $(foo) && something"
419 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
420 * Look for $(LN) and require AC_PROG_LN_S
422 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
424 Internationalize. [ gettext doesn't have the necessary machinery yet ]
425 am_error should use printf-style arguments (for eventual gettext scheme)
427 François says the ordering of files in a distribution should be as follows:
431 I agree, but I don't see how to implement this yet.
432 It might be easier if "derived files" is limited to those that
433 Automake itself knows about, eg output of yacc.
435 Check all source files to make sure that FSF address is up-to-date.
436 --gnits or --gnu only.
438 Merge each -vars.am file with corresponding ".am" file. Can do this
439 because of changes to &file_contents.
441 Should libexec programs have the name transform done on them?
443 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
444 together and rules are in the usual order.
446 Make the output minimal: only output definitions for variables that
450 David> To avoid comments like the one about subdirs getting buried in
451 David> the middle of a Makefile.in, how about pushing comments that
452 David> start with ### to the top of the Makefile.in (in order)? Sort
453 David> of like how Autoconf uses diversions to force initialization
454 David> code to the top of configure.
457 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
458 Karl> generally uses lowercase. Not that it matters :-).
460 ================================================================
464 probably should put each group of m4 files into a subdir owned by the
465 containing application.
467 ================================================================
473 how to use the generated makefiles
476 - NORMAL_INSTALL junk
478 what goes in AC_CONFIG_AUX_DIR
480 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
481 also a note on how a .am file is found in this case
483 rationale for avoiding
484 make CFLAGS="$CFLAGS" ...
487 a package that installs its own aclocal macros
489 write example of using automake with dejagnu
490 follow calc example in dejagnu docs
492 document which variables are actually scanned and which are not.
494 Document customary ordering of Makefile.am. From François.
496 Should include extended version of diagram from Autoconf (suggested by
499 Make a definition of the term "source"
501 document how to use Automake with CVS. Idea from Mark Galassi. Also
502 include Greg Woods' more sophisticated "cvs-dist" target.
504 document rebuilding configure. CONFIGURE_DEPENDENCIES
505 CONFIG_STATUS_DEPENDENCIES
507 -- must document all variables that are supposed
508 to be public knowledge
510 must document the targets required for integration with
511 non-automake-using subdirs
513 document the "make SHELL='/bin/sh -x'" trick for debugging
515 section on relationship to GNU make. include notes on parallel makes
519 move discussion of cygwin32, etags, mkid under other gnu tools
523 ================================================================
525 Things to do for gcc:
527 Regularize dependency generation. Add new flags:
529 -MH Generate a dummy dependency for each header file mentioned.
533 Set name of output file
535 Then automake can use -MD -MH -MT 'foo.o foo.lo' -MF .deps/...
537 ================================================================
539 Things to do for autoconf:
541 * patch autoreconf to run automake and aclocal. I've done this but it is
542 not really available. It can't be made available until automake
543 is officially released
545 ================================================================
549 * Should support standalone library along with subdir library in same
550 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
551 is not only one specd? [ add an option for this ]
553 ================================================================
557 Would it be useful to integrate in some way with the Debian package
558 building utility? Must check. maybe it would be possible to deal
559 with all the different package utilities somehow. Lately I've been
560 hearing good things about the RedHat packaging utilities. Why are
561 there so many of these? Are they fun to write or something?
562 The RedHat package utility is called RPM; see
563 ftp://ftp.redhat.com/pub/code/rpm
564 It actually has problems, like no configure script and no documentation.
566 For Cygnus it would probably be good to be able to handle the native
567 package utility on each platform. There are probably 3 or 4 of these
568 (sysv, solaris?, aix?)
570 tcl/unix/Makefile.in has some code to generate a Solaris package.
572 Automake probably can't do all of this on its own. A new tool might
575 I have some notes from a Debian developer on how the integration
578 ================================================================
580 A tool to guess what the local Makefile.am should look like:
581 (see Gord's Maint program!)
583 * Probably integrate with autoscan
584 * Use various simple rules to determine what to do:
585 * get name of top directory, sans version info
586 * search for .c files with 'main' in them
587 * if in main.c, use directory name for program
588 * if in more than one, generate multiple programs
589 * if not found, generate a library named after directory
590 * order subdir searches correctly: lib first, src last
591 * assume 'testsuite' dir means we are using dejagnu
592 * maybe be smart about reading existing Makefile.am, so tool
593 can be run for incremental changes? You could imagine:
596 autoproject --incremental
598 ================================================================
600 Stuff NOT to do, and why:
602 consider auto-including any file that matches "*.in".
603 [ no: po/Makefile.in shouldn't be included ]
605 must look at mkid to see how it works (for subdir usage)
606 [ right now, it doesn't. i don't see a simple fix right now ]
608 if configure.in not found, move up a directory and try again? This
609 could eliminate a common source of problems.
610 [ this is just a bad idea ]
612 * scripts are installed in $exec_prefix/bin, not $prefix/bin
614 [ the consensus on Gnits is that this isn't required.
615 doubters can work around it anyway ]
617 * make the auto-dep code crash if GNU make not in use?
618 (doesn't it already?)
620 Looked at a program called 'ezmake', which seems to do something
621 similar. The only idea there that is possibly worth stealing is using
622 globs in definitions. Also has negations. Eg in a directory with
623 files a.c, b.c and c.c, the line:
624 foo_SOURCES = *.c ~c.c
625 would be equivalent to:
626 foo_SOURCES = a.c b.c
627 Is this worth implementing?
628 [ No... it is more reliable to spell everything out. ]
630 Scan source directories and warn about missing files, eg .c/.h files
631 that aren't mentioned?
632 [ distcheck makes this less useful ]
635 - how to install file with a space in its name?
636 [ don't bother with this -- make is just too losing ]