2 - how to install file with a space in its name?
4 * "make diff" capability
5 look at gcc's Makefile.in to see what to do
6 or look at maint program
8 * Karl wants to be able to set LIBS and LDFLAGS at build time, like CFLAGS
9 maybe we need something more general?
11 * add $(srcdir)/ before some dependencies?
13 * mention PACKAGE/VERSIOn -vs- autoheader in manual
15 * define LINK if a program is mentioned, even if no C sources appear
17 * BUILT_SOURCES should not be distributed, even when they appear in
18 another _SOURCES line. [? or maybe just leave this up to the
19 to-be-defined generic distribution method ]
20 must completely revisit the entire BUILT_SOURCES idea
21 probably should generate dependencies as side effect of compile
22 (but that will mess up guile distributions... sigh)
24 * in --cygnus, clean-info not generated at top level
26 * what if an element of a scanned variable looks like
28 or some other arbitrary thing?
29 right now we try to cope, but not very well
31 * if `interlock' exists, that should be an error (?)
32 should also warn about using new ylwrap and not old one
33 only do this when looking for ylwrap
35 ** make sure every variable that is used is also defined
37 * make sure `missing' defines are generated
38 * if no AM_INIT_AUTOMAKE, then don't handle `missing' stuff.
40 * missing should handle install -d and rmdir -p (for uninstall)
42 * allow 'cygnus' in AUTOMAKE_OPTIONS
44 * a couple ways to be smarter:
45 - notice when a .c file is a target somewhere, and auto-add it to
47 - notice a target of the form `.x.y:' and assume it is a suffix rule
50 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
51 [ requires changes to the standard ]
53 * cross-compilation support:
54 programs built and used by the build process need to be
55 built for CC_FOR_BUILD
56 introduce a new variable for this
58 * if foo.y is a source, foo.h isn't auto-distributed?
60 * dependency tracking doesn't work well when a file is removed
61 the new code to track header dependencies exacerbates this
63 it would probably be better to use "gcc -MD" and move the .d
64 file into the .deps directory. That is, create the dependencies
65 as a side effect of compilation
66 This still won't solve the file-deletion problem
67 [ also: jim makes distributions by checking out, configuring,
68 and running "make dist". This scheme would cause that to fail ]
70 * copyrights on m4 files, aclocal output
72 * is there a way to add a directory and then have "make" do all the
75 * put standards.texi into distribution
78 * should not put texiname_TEXINFOS into distribution
79 should rename this macro anyway, to foo_texi_DEPENDENCIES
81 * *all* installed scripts should support --version, --help
83 For now I guess I'll just have automake give an error if it encounters
84 non-C source in a libtool library specification.
86 * must split $obj into two parts: one for libtool and one for
87 deansification. Otherwise .S files will be deansified!
89 * ansi2knr must currently appear in a directory that has some source
91 * if program has the same name as a target, do something sensible:
92 - if the target is internal, rename it
93 - if the target is mandated (eg, "info"), tell the user
94 consider auto-modifying the program name to work around this
96 * should separate actual options from strictness levels
97 strictness should only cover requirements
98 You should be able to pick and choose options
100 should clean up texinfos.am; one rule is repeated 3 times, but
103 should always use perl -w
105 rewrite in guile (RMS request)
106 at the same time, consider adding a GUI
107 could use the same parsing code for the GUI and the standalone version
108 that means figuring out a better representation of internal state
109 [ that's easy -- anything is better than what we have now ]
111 having just one Makefile for a project would give a big speed increase
112 for a project with many directories, eg glibc. ideally (?) you'd
113 still be able to have a Makefile.am in each directory somehow; this
114 might make editing conceptually easier.
116 * finish up TAGS work
118 * put parser.h into distribution if "yacc -d" is used
120 * only remove libtool at top level?
122 * clean up source directory by moving stuff into subdirs
124 * consider adding pkglibexecdir, maybe others?
125 requests for pkg-dirs with version included
130 Avoid loops when installing; instead unroll them in automake
133 * completely handle multi-":" mode for AC_CONFIG_HEADER
134 * Scan multiple input files when Makefile is generated?
135 This would provide flexibility for large projects; subsumes
136 the "Makefile.tmpl" idea
138 [ can't do this. must explain why in manual.
139 basically, solving all the problems is too hard
140 like: how to remove redundancies between generated .in files
141 instead should implement `include' directive for Makefile.am ]
142 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
143 first input file that has a corresponding .am file.
145 Some long-term projects:
146 * if $(FOO) is used somewhere, ensure FOO is defined, either by
147 user or by automake if possible
148 * Don't rearrange order of `include' lines relative to += assignments.
149 * Handle += assignments at all.
150 * Handle `include' lines by scanning other files, and adding
151 to Makefile.in dependency
153 consider putting all check-* targets onto @check?
154 To support --help/--version checking?
156 take diff-n-query code from libit
159 Per> 1) Being able to build a set of non-source programs
160 Per> from source programs, without necessarily linking them together.
161 Per> I.e. one should be able to say something like:
162 Per> dummy_SOURCES=foo.c bar.c
163 Per> and automake should realize that it needs to build foo.o and bar.o.
164 Per> 2) Being intelligent about new kinds of suffixes.
166 Per> SUFFIXES = .class .java
167 Per> and a suffix rule of the form:
169 Per> then it should be able to realize it can build .class files from
170 Per> .java files, and thus be able to generate a list of
171 Per> .class files from a list of .java source files.
173 !! Must fix require_file stuff. It is really gross, and I don't
174 understand it any more.
176 * error messages should print ``[info blah blah]'' command when a
177 certain part of the standards apply. saw idea in message from
178 Craig Burley. wouldn't it be really cool if compile-mode in Emacs
179 understood this convention, and you could click on such text to
180 go to the appropriate info page?
182 Jim's idea: should look for @setfilename and warn if filenames too long
185 ** many requests for a way to omit a file from the distribution.
186 Should be done like `!foo' or `~foo' in _SOURCES, etc.
187 Such files should be removed explicitly after the copy step!
188 Doing this requires rewriting macros before generating Makefile.in.
190 from joerg-martin schwarz:
191 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
192 in an explicitly written rule, you should emit the corresponding
193 Makefile variables automatically.
195 Configuring in the large:
196 * allow hierarchy of dirs to share one aclocal.m4
199 consider printing full file name of Makefile.am or configure.in when
200 giving error. This would help for very large trees with many
203 From the GNU Standards. These things could be checked, and probably
205 * Make sure that the directory into which the distribution unpacks (as
206 well as any subdirectories) are all world-writable (octal mode 777).
207 * Make sure that no file name in the distribution is more than 14
209 * Don't include any symbolic links in the distribution itself.
211 * Make sure that all the files in the distribution are world-readable.
212 ** also, check --help output and --version output. Idea from François
213 * standards no longer prohibit ANSI C. What does this imply
214 for the de-ansi-fication feature?
216 consider supporting "var+= stuff" syntax. rewrite to just var=... on
217 output. This is sometimes convenient when you want to write a
218 Makefile.am in more-or-less modular parts
220 should be able to determine what is built by looking at rules (and
221 configure.in). Then built man pages (eg) could automatically be
222 omitted from the distribution.
224 Idea from Joerg-Martin Schwarz: allow passing different -D flags to
225 different compiles. This can be done, but with the restriction that a
226 .c cannot appear in 2 different "objects" (programs/libraries)
227 compiled with different -D options (because -c and -o do not always
228 work together and parallel makes must work). This could be
229 implemented by noticing whenever a ".o" target with no rules is being
230 emitted, and adding the appropriate compilation rule as appropriate.
231 This should work with targets from Makefile.am as well as from .P
232 files, which means rewriting so that the Makefile.am contents aren't
233 copied into the output immediately.
234 [ this could be probably done more directly by examining the sources
235 as we scan Makefile.am ]
237 Henrik Frystyk Nielsen says:
238 Henrik> 4) Flags like --include-deps are lost when you make changes to
239 Henrik> Makefile.am files and automake is run automatically. It would
240 Henrik> be nice to keep these flags as I now have to redo everything
242 ... what about other options here too?
244 Think about: maybe "make check" should just bomb if error occurs?
245 Then user must use "make -k check". This is probably more natural.
247 Consider: "cvs" option adds some cvs-specific rules?
249 Right now, targets generated internally (eg "install") are not
250 overridable by user code. This should probably be possible, even
251 though it isn't very important. This could be done by generating all
252 internal rules via a function call instead of just appending to
254 [ this will be harder to implement when scanning a rule like all-recursive
257 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
258 but which could be by running the magic make command.
261 * Must rewrite am_install_var. Should break into multiple functions.
262 This will allow the callers to be a little smarter.
263 * Rewrite clean targets.
264 * Must rewrite error handling code. Right now it is a real mess
265 Should fix up require_file junk at the same time
267 djm wants ``LINKS'' variable; list of things to link together after
268 install. In BSD environment, use:
269 LINKS = from1 to1 from2 to2 ...
271 Need way to say there are no suffixes in a Makefile (Franc,ois'
272 "override" idea suffices here)
274 Check to make sure various scripts are executable (IE when looking for
277 Use recode in dist target when MAINT_CHARSET specified. Read caveats
278 in automake.in before doing this. Note the same problem used to apply
279 to the no-dependencies option; maybe it still should? Note also that
280 each Makefile.am must be rewritten at "make dist" time if
281 MAINT_CHARSET and DIST_CHARSET are not identical. NOTE: gettext must
282 arrange for all .po files not to be recoded. In the long term this
283 might be a problem (consider when some systems use Unicode but the
285 MAINT_CHARSET *must* be local to each Makefile.am, to enable
286 merged distributions.
287 DIST_CHARSET must be passed down to subdir makes during a "make dist"
289 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
290 (why isn't this merged with "pathchk"?) when doing a dist. Do
291 whatever else François says here...
293 Add support for html via an option. Use texi2html. Use
294 "html_TEXINFOS", and htmldir = .../html. Include html files in
295 distribution. Also allow "html_DATA", for raw .html files.
296 [ when will texinfo directly support html? ]
298 uninstall and pkg-dirs should rm -rf the dir.
300 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
301 these files can't be de-ansified. Not a problem?
302 [ fix by using ansi2knr wrapper program ]
304 In general most .am files should be merged into automake. For
305 instance all the "clean" targets could be merged by keeping lists of
306 things to be removed. This would be a lot nicer looking. Note that
307 the install targets probably should not be merged; it is sometimes
308 useful to only install a small part.
311 * Order rules sensibly
312 * Ensure every line has a purpose. Omit unused stuff
313 * Eliminate extraneous rules when possible (eg 'install-am' stuff)
314 * Make sure vertical spacing is correct
315 Omit program transform vars from header if no program installed. This
316 is currently pretty hard to do. (But with beautification code it
317 would probably be easy)
320 * It would be nice to automatically support using bison's better features
321 to rename the output files. This requires autoconf support
322 * Consider supporting syntax from autoconf "derived:source", eg:
324 for yacc and lex source
325 * what if you use flex and the option to avoid -lfl?
328 Multi-language support:
329 * should have mapping of file extensions to languages
330 * should automatically handle the linking issue (special-case C++)
331 * must get compile rules for various languages; FORTRAN probably
332 most important unimplemented language
333 This should be integrated in some way with Per's idea.
334 Eg .f.o rules should be recognized & auto-handled in _SOURCES
335 That way any random language can be treated with C/C++ on a first-class
338 It might be cool to generate .texi dependencies by grepping for
339 @include. (If done, it should be done the same way C dependencies are
342 It would be good to check some parts of GNU standards. Already check
343 for install-sh and mkinstalldirs. What else is required to be in
344 package by GNU standards or by automake?
345 Some things for --strictness=gnits:
346 * "cd $(foo); something" is an error in a rule. Should be:
347 "cd $(foo) && something"
348 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
349 * Look for $(LN) and require AC_PROG_LN_S
351 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
353 Internationalize. [ gettext doesn't have the necessary machinery yet ]
354 am_error should use printf-style arguments (for eventual gettext scheme)
356 François says the ordering of files in a distribution should be as follows:
360 I agree, but I don't see how to implement this yet.
361 It might be easier if "derived files" is limited to those that
362 Automake itself knows about, eg output of yacc.
364 Check all source files to make sure that FSF address is up-to-date.
365 --gnits or --gnu only.
367 Merge each -vars.am file with corresponding ".am" file. Can do this
368 because of changes to &file_contents.
370 Should libexec programs have the name transform done on them?
372 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
373 together and rules are in the usual order.
375 Make the output minimal: only output definitions for variables that
379 David> To avoid comments like the one about subdirs getting buried in
380 David> the middle of a Makefile.in, how about pushing comments that
381 David> start with ### to the top of the Makefile.in (in order)? Sort
382 David> of like how Autoconf uses diversions to force initialization
383 David> code to the top of configure.
386 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
387 Karl> generally uses lowercase. Not that it matters :-).
389 ================================================================
393 probably should put each group of m4 files into a subdir owned by the
394 containing application.
396 ================================================================
402 how to use the generated makefiles
405 - NORMAL_INSTALL junk
407 what goes in AC_CONFIG_AUX_DIR
409 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
410 also a note on how a .am file is found in this case
412 rationale for avoiding
413 make CFLAGS="$CFLAGS" ...
416 a package that installs its own aclocal macros
418 write example of using automake with dejagnu
419 follow calc example in dejagnu docs
421 document which variables are actually scanned and which are not.
423 Document customary ordering of Makefile.am. From François.
425 Should include extended version of diagram from Autoconf (suggested by
428 Make a definition of the term "source"
430 document how to use Automake with CVS. Idea from Mark Galassi. Also
431 include Greg Woods' more sophisticated "cvs-dist" target.
433 document rebuilding configure. CONFIGURE_DEPENDENCIES
435 -- must document all variables that are supposed
436 to be public knowledge
438 must document the targets required for integration with
439 non-automake-using subdirs
441 document the "make SHELL='/bin/sh -x'" trick for debugging
443 section on relationship to GNU make. include notes on parallel makes
447 move discussion of cygwin32, etags, mkid under other gnu tools
449 ================================================================
451 Things to do for autoconf:
453 * patch autoreconf to run automake and aclocal. I've done this but it is
454 not really available. It can't be made available until automake
455 is officially released
457 ================================================================
461 * Should support standalone library along with subdir library in same
462 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
463 is not only one specd? [ add an option for this ]
465 ================================================================
469 Would it be useful to integrate in some way with the Debian package
470 building utility? Must check. maybe it would be possible to deal
471 with all the different package utilities somehow. Lately I've been
472 hearing good things about the RedHat packaging utilities. Why are
473 there so many of these? Are they fun to write or something?
474 The RedHat package utility is called RPM; see
475 ftp://ftp.redhat.com/pub/code/rpm
476 It actually has problems, like no configure script and no documentation.
478 For Cygnus it would probably be good to be able to handle the native
479 package utility on each platform. There are probably 3 or 4 of these
480 (sysv, solaris?, aix?)
482 tcl/unix/Makefile.in has some code to generate a Solaris package.
484 Automake probably can't do all of this on its own. A new tool might
487 I have some notes from a Debian developer on how the integration
490 ================================================================
492 A tool to guess what the local Makefile.am should look like:
493 (see Gord's Maint program!)
495 * Probably integrate with autoscan
496 * Use various simple rules to determine what to do:
497 * get name of top directory, sans version info
498 * search for .c files with 'main' in them
499 * if in main.c, use directory name for program
500 * if in more than one, generate multiple programs
501 * if not found, generate a library named after directory
502 * order subdir searches correctly: lib first, src last
503 * assume 'testsuite' dir means we are using dejagnu
504 * maybe be smart about reading existing Makefile.am, so tool
505 can be run for incremental changes? You could imagine:
508 autoproject --incremental
510 ================================================================
512 Stuff NOT to do, and why:
514 consider auto-including any file that matches "*.in".
515 [ no: po/Makefile.in shouldn't be included ]
517 must look at mkid to see how it works (for subdir usage)
518 [ right now, it doesn't. i don't see a simple fix right now ]
520 if configure.in not found, move up a directory and try again? This
521 could eliminate a common source of problems.
522 [ this is just a bad idea ]
524 * scripts are installed in $exec_prefix/bin, not $prefix/bin
526 [ the consensus on Gnits is that this isn't required.
527 doubters can work around it anyway ]
529 * make the auto-dep code crash if GNU make not in use?
530 (doesn't it already?)
532 Looked at a program called 'ezmake', which seems to do something
533 similar. The only idea there that is possibly worth stealing is using
534 globs in definitions. Also has negations. Eg in a directory with
535 files a.c, b.c and c.c, the line:
536 foo_SOURCES = *.c ~c.c
537 would be equivalent to:
538 foo_SOURCES = a.c b.c
539 Is this worth implementing?
540 [ No... it is more reliable to spell everything out. ]
542 Scan source directories and warn about missing files, eg .c/.h files
543 that aren't mentioned?
544 [ distcheck makes this less useful ]