1 * `DEFS += foo' won't work.
2 That's because DEFS is defined in header-vars.am, which is read
3 after the user's Makefile.am.
4 This will be a problem for any macro defined internally
6 * document user namespace for macro/target names
7 adopt some conventions and use uniformly
9 * make distcheck uses directories like `=build'.
10 Some (very rare) POSIX systems don't support `=' in filenames.
11 If this ever becomes a problem, fix it
13 a* distclean must remove config.status
14 can't this cause problems for maintainer-clean?
15 shouldn't maintainer-clean print the message before running
16 any part of the make? (just to slow things down long enough
17 for the user to stop it)
18 (maybe doesn't matter since people who even know about
19 maintainer-clean already have a clue)
21 * There are probably more bugs in variable_conditions_sub along
22 the lines of the one that caused cond4.test to fail.
24 * give user more control over -I flags
25 in particular document a way to override the defaults
27 * reintroduce AM_FUNC_FNMATCH which sets LIBOBJS
28 Then have automake know about fnmatch.h.
30 * Add a conditional for dependency tracking
31 (what to name it is the biggest problem here)
32 (because we want it to flag dist/no-dist -- not just deps)
34 * Allow per-object cflags:
37 * At the same time, allow sources in subdirs:
39 This requires `mkdir x' at build time
40 [ both of these require per-file rules, and not pattern rules ]
41 [ use user-written suffix rules to generate the per-file rules in
42 an automatic way -- this would be mucho cool ]
44 * Allow for multiple translations of a texinfo file:
45 LL_info_TEXINFOS = ...
46 will put info files for language LL into $(infodir)/LL.
48 * If you suppress an internal variable by specifying a variable
49 in a Makefile.am, but the variable is conditional, then automake
50 should generate the internal variable conditionally.
51 You have to scan the map of all conditions and fill in the holes here
53 * Every program foo has FOOFLAGS right now.
54 It should also have AM_FOOFLAGS, which can be set in Makefile.am.
55 DONE: but needs to be documented
57 * Should be able to update files that would be installed with -a
60 * "make diff" capability
61 look at gcc's Makefile.in to see what to do
62 or look at maint program
64 * Karl wants to be able to set LIBS and LDFLAGS at build time, like CFLAGS
65 maybe we need something more general?
67 * add $(srcdir)/ before some dependencies?
69 * define LINK if a program is mentioned, even if no C sources appear
71 * BUILT_SOURCES should not be distributed, even when they appear in
72 another _SOURCES line. [? or maybe just leave this up to the
73 to-be-defined generic distribution method ]
74 must completely revisit the entire BUILT_SOURCES idea
76 * in --cygnus, clean-info not generated at top level
78 * what if an element of a scanned variable looks like
80 or some other arbitrary thing?
81 right now we try to cope, but not very well
83 * if `interlock' exists, that should be an error (?)
84 should also warn about using new ylwrap and not old one
85 only do this when looking for ylwrap
87 ** make sure every variable that is used is also defined
89 * make sure `missing' defines are generated
90 * if no AM_INIT_AUTOMAKE, then don't handle `missing' stuff.
92 * missing should handle install -d and rmdir -p (for uninstall)
94 * a couple ways to be smarter:
95 - notice when a .c file is a target somewhere, and auto-add it to
97 - notice a target of the form `.x.y:' and assume it is a suffix rule
100 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
101 [ requires changes to the standard ]
103 * cross-compilation support:
104 programs built and used by the build process need to be
105 built for CC_FOR_BUILD
106 introduce a new variable for this
108 * if foo.y is a source, foo.h isn't auto-distributed?
110 * copyrights on m4 files, aclocal output
112 * is there a way to add a directory and then have "make" do all the
115 * put standards.texi into distribution
118 * should not put texiname_TEXINFOS into distribution
119 should rename this macro anyway, to foo_texi_DEPENDENCIES
121 * *all* installed scripts should support --version, --help
123 For now I guess I'll just have automake give an error if it encounters
124 non-C source in a libtool library specification.
126 * must split $obj into two parts: one for libtool and one for
127 deansification. Otherwise .S files will be deansified!
129 * ansi2knr must currently appear in a directory that has some source
131 * if program has the same name as a target, do something sensible:
132 - if the target is internal, rename it
133 - if the target is mandated (eg, "info"), tell the user
134 consider auto-modifying the program name to work around this
136 * should separate actual options from strictness levels
137 strictness should only cover requirements
138 You should be able to pick and choose options
140 should clean up texinfos.am; one rule is repeated 3 times, but
143 should always use perl -w
145 rewrite in guile (RMS request)
146 at the same time, consider adding a GUI
147 could use the same parsing code for the GUI and the standalone version
148 that means figuring out a better representation of internal state
149 [ that's easy -- anything is better than what we have now ]
151 having just one Makefile for a project would give a big speed increase
152 for a project with many directories, eg glibc. ideally (?) you'd
153 still be able to have a Makefile.am in each directory somehow; this
154 might make editing conceptually easier.
156 * finish up TAGS work
158 * put parser.h into distribution if "yacc -d" is used
160 * only remove libtool at top level?
162 * clean up source directory by moving stuff into subdirs
164 * consider adding pkglibexecdir, maybe others?
165 requests for pkg-dirs with version included
167 Avoid loops when installing; instead unroll them in automake
170 * completely handle multi-":" mode for AC_CONFIG_HEADER
171 * Scan multiple input files when Makefile is generated?
172 This would provide flexibility for large projects; subsumes
173 the "Makefile.tmpl" idea
175 [ can't do this. must explain why in manual.
176 basically, solving all the problems is too hard
177 like: how to remove redundancies between generated .in files
178 instead should implement `include' directive for Makefile.am ]
179 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
180 first input file that has a corresponding .am file.
182 Some long-term projects:
183 * if $(FOO) is used somewhere, ensure FOO is defined, either by
184 user or by automake if possible
186 [ include, += support ]
187 * even better would be allowing targets in different included
188 fragments to be merged. e.g., `install-local'.
190 consider putting all check-* targets onto @check?
191 To support --help/--version checking?
193 take diff-n-query code from libit
196 Per> 1) Being able to build a set of non-source programs
197 Per> from source programs, without necessarily linking them together.
198 Per> I.e. one should be able to say something like:
199 Per> dummy_SOURCES=foo.c bar.c
200 Per> and automake should realize that it needs to build foo.o and bar.o.
201 Per> 2) Being intelligent about new kinds of suffixes.
203 Per> SUFFIXES = .class .java
204 Per> and a suffix rule of the form:
206 Per> then it should be able to realize it can build .class files from
207 Per> .java files, and thus be able to generate a list of
208 Per> .class files from a list of .java source files.
210 !! Must fix require_file stuff. It is really gross, and I don't
211 understand it any more.
213 * error messages should print ``[info blah blah]'' command when a
214 certain part of the standards apply. saw idea in message from
215 Craig Burley. wouldn't it be really cool if compile-mode in Emacs
216 understood this convention, and you could click on such text to
217 go to the appropriate info page? [ I think you can ]
219 Jim's idea: should look for @setfilename and warn if filenames too long
222 ** many requests for a way to omit a file from the distribution.
223 Should be done like `!foo' or `~foo' in _SOURCES, etc.
224 Such files should be removed explicitly after the copy step!
225 Doing this requires rewriting macros before generating Makefile.in.
227 from joerg-martin schwarz:
228 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
229 in an explicitly written rule, you should emit the corresponding
230 Makefile variables automatically.
232 Configuring in the large:
233 * allow hierarchy of dirs to share one aclocal.m4
236 consider printing full file name of Makefile.am or configure.in when
237 giving error. This would help for very large trees with many
240 From the GNU Standards. These things could be checked, and probably
242 * Make sure that the directory into which the distribution unpacks (as
243 well as any subdirectories) are all world-writable (octal mode 777).
244 * Make sure that no file name in the distribution is more than 14
246 * Don't include any symbolic links in the distribution itself.
248 * Make sure that all the files in the distribution are world-readable.
249 ** also, check --help output and --version output. Idea from François
250 * standards no longer prohibit ANSI C. What does this imply
251 for the de-ansi-fication feature? [ must keep it -- some users rely on it ]
253 should be able to determine what is built by looking at rules (and
254 configure.in). Then built man pages (eg) could automatically be
255 omitted from the distribution.
257 Idea from Joerg-Martin Schwarz: allow passing different -D flags to
258 different compiles. This can be done, but with the restriction that a
259 .c cannot appear in 2 different "objects" (programs/libraries)
260 compiled with different -D options (because -c and -o do not always
261 work together and parallel makes must work). This could be
262 implemented by noticing whenever a ".o" target with no rules is being
263 emitted, and adding the appropriate compilation rule as appropriate.
264 This should work with targets from Makefile.am as well as from .P
265 files, which means rewriting so that the Makefile.am contents aren't
266 copied into the output immediately.
267 [ this could be probably done more directly by examining the sources
268 as we scan Makefile.am ]
270 Henrik Frystyk Nielsen says:
271 Henrik> 4) Flags like --include-deps are lost when you make changes to
272 Henrik> Makefile.am files and automake is run automatically. It would
273 Henrik> be nice to keep these flags as I now have to redo everything
275 ... what about other options here too?
277 Think about: maybe "make check" should just bomb if error occurs?
278 Then user must use "make -k check". This is probably more natural.
280 Consider: "cvs" option adds some cvs-specific rules?
282 Right now, targets generated internally (eg "install") are not
283 overridable by user code. This should probably be possible, even
284 though it isn't very important. This could be done by generating all
285 internal rules via a function call instead of just appending to
287 [ this will be harder to implement when scanning a rule like all-recursive
290 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
291 but which could be by running the magic make command.
294 * Must rewrite am_install_var. Should break into multiple functions.
295 This will allow the callers to be a little smarter.
296 * Rewrite clean targets.
297 * Must rewrite error handling code. Right now it is a real mess
298 Should fix up require_file junk at the same time
300 djm wants ``LINKS'' variable; list of things to link together after
301 install. In BSD environment, use:
302 LINKS = from1 to1 from2 to2 ...
304 Need way to say there are no suffixes in a Makefile (Franc,ois'
305 "override" idea suffices here)
307 Check to make sure various scripts are executable (IE when looking for
310 Use recode in dist target when MAINT_CHARSET specified. Read caveats
311 in automake.in before doing this. Note the same problem used to apply
312 to the no-dependencies option; maybe it still should? Note also that
313 each Makefile.am must be rewritten at "make dist" time if
314 MAINT_CHARSET and DIST_CHARSET are not identical. NOTE: gettext must
315 arrange for all .po files not to be recoded. In the long term this
316 might be a problem (consider when some systems use Unicode but the
318 MAINT_CHARSET *must* be local to each Makefile.am, to enable
319 merged distributions.
320 DIST_CHARSET must be passed down to subdir makes during a "make dist"
322 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
323 (why isn't this merged with "pathchk"?) when doing a dist. Do
324 whatever else François says here...
326 Add support for html via an option. Use texi2html. Use
327 "html_TEXINFOS", and htmldir = .../html. Include html files in
328 distribution. Also allow "html_DATA", for raw .html files.
329 [ when will texinfo directly support html? ]
331 uninstall and pkg-dirs should rm -rf the dir.
333 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
334 these files can't be de-ansified. Not a problem?
335 [ fix by using ansi2knr wrapper program ]
337 In general most .am files should be merged into automake. For
338 instance all the "clean" targets could be merged by keeping lists of
339 things to be removed. This would be a lot nicer looking. Note that
340 the install targets probably should not be merged; it is sometimes
341 useful to only install a small part.
344 * Order rules sensibly
345 * Ensure every line has a purpose. Omit unused stuff
346 * Eliminate extraneous rules when possible (eg 'install-am' stuff)
347 * Make sure vertical spacing is correct
348 Omit program transform vars from header if no program installed. This
349 is currently pretty hard to do. (But with beautification code it
350 would probably be easy)
353 * It would be nice to automatically support using bison's better features
354 to rename the output files. This requires autoconf support
355 * Consider supporting syntax from autoconf "derived:source", eg:
357 for yacc and lex source
358 * what if you use flex and the option to avoid -lfl?
361 Multi-language support:
362 * should have mapping of file extensions to languages
363 * should automatically handle the linking issue (special-case C++)
364 * must get compile rules for various languages; FORTRAN probably
365 most important unimplemented language
366 This should be integrated in some way with Per's idea.
367 Eg .f.o rules should be recognized & auto-handled in _SOURCES
368 That way any random language can be treated with C/C++ on a first-class
371 It might be cool to generate .texi dependencies by grepping for
372 @include. (If done, it should be done the same way C dependencies are
375 It would be good to check some parts of GNU standards. Already check
376 for install-sh and mkinstalldirs. What else is required to be in
377 package by GNU standards or by automake?
378 Some things for --strictness=gnits:
379 * "cd $(foo); something" is an error in a rule. Should be:
380 "cd $(foo) && something"
381 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
382 * Look for $(LN) and require AC_PROG_LN_S
384 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
386 Internationalize. [ gettext doesn't have the necessary machinery yet ]
387 am_error should use printf-style arguments (for eventual gettext scheme)
389 François says the ordering of files in a distribution should be as follows:
393 I agree, but I don't see how to implement this yet.
394 It might be easier if "derived files" is limited to those that
395 Automake itself knows about, eg output of yacc.
397 Check all source files to make sure that FSF address is up-to-date.
398 --gnits or --gnu only.
400 Merge each -vars.am file with corresponding ".am" file. Can do this
401 because of changes to &file_contents.
403 Should libexec programs have the name transform done on them?
405 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
406 together and rules are in the usual order.
408 Make the output minimal: only output definitions for variables that
412 David> To avoid comments like the one about subdirs getting buried in
413 David> the middle of a Makefile.in, how about pushing comments that
414 David> start with ### to the top of the Makefile.in (in order)? Sort
415 David> of like how Autoconf uses diversions to force initialization
416 David> code to the top of configure.
419 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
420 Karl> generally uses lowercase. Not that it matters :-).
422 ================================================================
426 probably should put each group of m4 files into a subdir owned by the
427 containing application.
429 ================================================================
435 how to use the generated makefiles
438 - NORMAL_INSTALL junk
440 what goes in AC_CONFIG_AUX_DIR
442 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
443 also a note on how a .am file is found in this case
445 rationale for avoiding
446 make CFLAGS="$CFLAGS" ...
449 a package that installs its own aclocal macros
451 write example of using automake with dejagnu
452 follow calc example in dejagnu docs
454 document which variables are actually scanned and which are not.
456 Document customary ordering of Makefile.am. From François.
458 Should include extended version of diagram from Autoconf (suggested by
461 Make a definition of the term "source"
463 document how to use Automake with CVS. Idea from Mark Galassi. Also
464 include Greg Woods' more sophisticated "cvs-dist" target.
466 document rebuilding configure. CONFIGURE_DEPENDENCIES
467 CONFIG_STATUS_DEPENDENCIES
469 -- must document all variables that are supposed
470 to be public knowledge
472 must document the targets required for integration with
473 non-automake-using subdirs
475 document the "make SHELL='/bin/sh -x'" trick for debugging
477 section on relationship to GNU make. include notes on parallel makes
481 move discussion of cygwin32, etags, mkid under other gnu tools
485 ================================================================
487 Things to do for gcc:
489 Regularize dependency generation. Add new flags:
491 -MH Generate a dummy dependency for each header file mentioned.
495 Set name of output file
497 Then automake can use -MD -MH -MT 'foo.o foo.lo' -MF .deps/...
499 ================================================================
501 Things to do for autoconf:
503 * patch autoreconf to run automake and aclocal. I've done this but it is
504 not really available. It can't be made available until automake
505 is officially released
507 ================================================================
511 * Should support standalone library along with subdir library in same
512 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
513 is not only one specd? [ add an option for this ]
515 ================================================================
519 Would it be useful to integrate in some way with the Debian package
520 building utility? Must check. maybe it would be possible to deal
521 with all the different package utilities somehow. Lately I've been
522 hearing good things about the RedHat packaging utilities. Why are
523 there so many of these? Are they fun to write or something?
524 The RedHat package utility is called RPM; see
525 ftp://ftp.redhat.com/pub/code/rpm
526 It actually has problems, like no configure script and no documentation.
528 For Cygnus it would probably be good to be able to handle the native
529 package utility on each platform. There are probably 3 or 4 of these
530 (sysv, solaris?, aix?)
532 tcl/unix/Makefile.in has some code to generate a Solaris package.
534 Automake probably can't do all of this on its own. A new tool might
537 I have some notes from a Debian developer on how the integration
540 ================================================================
542 A tool to guess what the local Makefile.am should look like:
543 (see Gord's Maint program!)
545 * Probably integrate with autoscan
546 * Use various simple rules to determine what to do:
547 * get name of top directory, sans version info
548 * search for .c files with 'main' in them
549 * if in main.c, use directory name for program
550 * if in more than one, generate multiple programs
551 * if not found, generate a library named after directory
552 * order subdir searches correctly: lib first, src last
553 * assume 'testsuite' dir means we are using dejagnu
554 * maybe be smart about reading existing Makefile.am, so tool
555 can be run for incremental changes? You could imagine:
558 autoproject --incremental
560 ================================================================
562 Stuff NOT to do, and why:
564 consider auto-including any file that matches "*.in".
565 [ no: po/Makefile.in shouldn't be included ]
567 must look at mkid to see how it works (for subdir usage)
568 [ right now, it doesn't. i don't see a simple fix right now ]
570 if configure.in not found, move up a directory and try again? This
571 could eliminate a common source of problems.
572 [ this is just a bad idea ]
574 * scripts are installed in $exec_prefix/bin, not $prefix/bin
576 [ the consensus on Gnits is that this isn't required.
577 doubters can work around it anyway ]
579 * make the auto-dep code crash if GNU make not in use?
580 (doesn't it already?)
582 Looked at a program called 'ezmake', which seems to do something
583 similar. The only idea there that is possibly worth stealing is using
584 globs in definitions. Also has negations. Eg in a directory with
585 files a.c, b.c and c.c, the line:
586 foo_SOURCES = *.c ~c.c
587 would be equivalent to:
588 foo_SOURCES = a.c b.c
589 Is this worth implementing?
590 [ No... it is more reliable to spell everything out. ]
592 Scan source directories and warn about missing files, eg .c/.h files
593 that aren't mentioned?
594 [ distcheck makes this less useful ]
597 - how to install file with a space in its name?
598 [ don't bother with this -- make is just too losing ]