1 * NORMAL_INSTALL / NORMAL_UNINSTALL -vs- recursive rules
3 * copyrights on m4 files, aclocal output
5 * if no AM_INIT_AUTOMAKE, then don't handle `missing' stuff.
8 * should not put texiname_TEXINFOS into distribution
9 should rename this macro anyway, to foo_texi_DEPENDENCIES
11 * should be able to specify path to texinfo.tex so that 'make dvi'
12 will work even when it isn't in this dir.
14 * *all* installed scripts should support --version, --help
16 * have aclocal diagnose unrecognized AM_ macros
18 For now I guess I'll just have automake give an error if it encounters
19 non-C source in a libtool library specification.
21 * must split $obj into two parts: one for libtool and one for
22 deansification. Otherwise .S files will be deansified!
24 * if program has the same name as a target, do something sensible:
25 - if the target is internal, rename it
26 - if the target is mandated (eg, "info"), tell the user
27 consider auto-modifying the program name to work around this
29 * should separate actual options from strictness levels
30 strictness should only cover requirements
31 You should be able to pick and choose options
33 should clean up texinfos.am; one rule is repeated 3 times, but
36 should always use perl -w
42 * put parser.h into distribution if "yacc -d" is used
44 * only remove libtool at top level?
46 * clean up source directory by moving stuff into subdirs
48 * consider adding pkglibexecdir, maybe others?
49 requests for pkg-dirs with version included
54 Avoid loops when installing; instead unroll them in automake
57 * completely handle multi-":" mode for AC_CONFIG_HEADER
58 * Scan multiple input files when Makefile is generated?
59 This would provide flexibility for large projects; subsumes
60 the "Makefile.tmpl" idea
62 [ can't do this. must explain why in manual.
63 basically, solving all the problems is too hard
64 like: how to remove redundancies between generated .in files
65 instead should implement `include' directive for Makefile.am ]
66 * for multi-":" mode and AC_OUTPUT, it might be good to pick the
67 first input file that has a corresponding .am file.
69 Some long-term projects:
70 * if $(FOO) is used somewhere, ensure FOO is defined, either by
71 user or by automake if possible
72 * Don't rearrange order of `include' lines relative to += assignments.
73 * Handle += assignments at all.
75 consider putting all check-* targets onto @check?
76 To support --help/--version checking?
78 take diff-n-query code from libit
81 Per> 1) Being able to build a set of non-source programs
82 Per> from source programs, without necessarily linking them together.
83 Per> I.e. one should be able to say something like:
84 Per> dummy_SOURCES=foo.c bar.c
85 Per> and automake should realize that it needs to build foo.o and bar.o.
86 Per> 2) Being intelligent about new kinds of suffixes.
88 Per> SUFFIXES = .class .java
89 Per> and a suffix rule of the form:
91 Per> then it should be able to realize it can build .class files from
92 Per> .java files, and thus be able to generate a list of
93 Per> .class files from a list of .java source files.
96 * allow a way to use "gzip --best"
97 [ Try GZIP=--best make dist ]
98 * don't assume GNU tar is "tar" (eg in distcheck)
100 !! Must fix require_file stuff. It is really gross, and I don't
101 understand it any more.
103 * error messages should print ``[info blah blah]'' command when a
104 certain part of the standards apply. saw idea in message from
105 Craig Burley. wouldn't it be really cool if compile-mode in Emacs
106 understood this convention, and you could click on such text to
107 go to the appropriate info page?
109 !! should write autoconf-style doc entries for each m4 macro
111 Jim's idea: should look for @setfilename and warn if filenames too long
114 ** many requests for a way to omit a file from the distribution.
115 Should be done like `!foo' or `~foo' in _SOURCES, etc.
116 Such files should be removed explicitly after the copy step!
117 Doing this requires rewriting macros before generating Makefile.in.
119 from joerg-martin schwarz:
120 -- If Makefile.am contains $(CC), $(COMPILE), $(YLWRAP), ....
121 in an explicitly written rule, you should emit the corresponding
122 Makefile variables automatically.
124 Configuring in the large:
125 * allow hierarchy of dirs to share one aclocal.m4
128 consider printing full file name of Makefile.am or configure.in when
129 giving error. This would help for very large trees with many
132 From the GNU Standards. These things could be checked, and probably
134 * Make sure that the directory into which the distribution unpacks (as
135 well as any subdirectories) are all world-writable (octal mode 777).
136 * Make sure that no file name in the distribution is more than 14
138 * Don't include any symbolic links in the distribution itself.
140 * Make sure that all the files in the distribution are world-readable.
141 ** also, check --help output and --version output. Idea from François
142 * standards no longer prohibit ANSI C. What does this imply
143 for the de-ansi-fication feature?
145 consider supporting "var+= stuff" syntax. rewrite to just var=... on
146 output. This is sometimes convenient when you want to write a
147 Makefile.am in more-or-less modular parts
149 should be able to determine what is built by looking at rules (and
150 configure.in). Then built man pages (eg) could automatically be
151 omitted from the distribution.
153 Idea from Joerg-Martin Schwarz: allow passing different -D flags to
154 different compiles. This can be done, but with the restriction that a
155 .c cannot appear in 2 different "objects" (programs/libraries)
156 compiled with different -D options (because -c and -o do not always
157 work together and parallel makes must work). This could be
158 implemented by noticing whenever a ".o" target with no rules is being
159 emitted, and adding the appropriate compilation rule as appropriate.
160 This should work with targets from Makefile.am as well as from .P
161 files, which means rewriting so that the Makefile.am contents aren't
162 copied into the output immediately.
163 [ this could be probably done more directly by examining the sources
164 as we scan Makefile.am ]
166 Henrik Frystyk Nielsen says:
167 Henrik> 4) Flags like --include-deps are lost when you make changes to
168 Henrik> Makefile.am files and automake is run automatically. It would
169 Henrik> be nice to keep these flags as I now have to redo everything
171 ... what about other options here too?
173 Think about: maybe "make check" should just bomb if error occurs?
174 Then user must use "make -k check". This is probably more natural.
176 Consider: "cvs" option adds some cvs-specific rules?
178 Right now, targets generated internally (eg "install") are not
179 overridable by user code. This should probably be possible, even
180 though it isn't very important. This could be done by generating all
181 internal rules via a function call instead of just appending to
183 [ this will be harder to implement when scanning a rule like all-recursive
186 * Should be a way to have "nobuild_PROGRAMS" which aren't even built,
187 but which could be by running the magic make command.
190 * Must rewrite am_install_var. Should break into multiple functions.
191 This will allow the callers to be a little smarter.
192 * Rewrite clean targets.
193 * Must rewrite error handling code. Right now it is a real mess
194 Should fix up require_file junk at the same time
196 djm wants ``LINKS'' variable; list of things to link together after
197 install. In BSD environment, use:
198 LINKS = from1 to1 from2 to2 ...
200 Need way to say there are no suffixes in a Makefile (Franc,ois'
201 "override" idea suffices here)
203 Check to make sure various scripts are executable (IE when looking for
206 Use recode in dist target when MAINT_CHARSET specified. Read caveats
207 in automake.in before doing this. Note the same problem used to apply
208 to the no-dependencies option; maybe it still should? Note also that
209 each Makefile.am must be rewritten at "make dist" time if
210 MAINT_CHARSET and DIST_CHARSET are not identical. NOTE: gettext must
211 arrange for all .po files not to be recoded. In the long term this
212 might be a problem (consider when some systems use Unicode but the
214 MAINT_CHARSET *must* be local to each Makefile.am, to enable
215 merged distributions.
216 DIST_CHARSET must be passed down to subdir makes during a "make dist"
218 Handle dist-zoo. Generally add more DOS support. Maybe run "doschk"
219 (why isn't this merged with "pathchk"?) when doing a dist. Do
220 whatever else François says here...
222 Add support for html via an option. Use texi2html. Use
223 "html_TEXINFOS", and htmldir = .../html. Include html files in
224 distribution. Also allow "html_DATA", for raw .html files.
225 [ when will texinfo directly support html? ]
227 uninstall and pkg-dirs should rm -rf the dir.
229 a potential bug: configure puts "blah.o" into LIBOBJS, thus implying
230 these files can't be de-ansified. Not a problem?
231 [ fix by using ansi2knr wrapper program ]
233 In general most .am files should be merged into automake. For
234 instance all the "clean" targets could be merged by keeping lists of
235 things to be removed. This would be a lot nicer looking. Note that
236 the install targets probably should not be merged; it is sometimes
237 useful to only install a small part.
240 * Order rules sensibly
241 * Ensure every line has a purpose. Omit unused stuff
242 * Eliminate extraneous rules when possible (eg 'install-am' stuff)
243 * Make sure vertical spacing is correct
244 Omit program transform vars from header if no program installed. This
245 is currently pretty hard to do. (But with beautification code it
246 would probably be easy)
249 * It would be nice to automatically support using bison's better features
250 to rename the output files. This requires autoconf support
251 * Consider supporting syntax from autoconf "derived:source", eg:
253 for yacc and lex source
254 * what if you use flex and the option to avoid -lfl?
257 Multi-language support:
258 * should have mapping of file extensions to languages
259 * should automatically handle the linking issue (special-case C++)
260 * must get compile rules for various languages; FORTRAN probably
261 most important unimplemented language
262 This should be integrated in some way with Per's idea.
263 Eg .f.o rules should be recognized & auto-handled in _SOURCES
264 That way any random language can be treated with C/C++ on a first-class
267 It might be cool to generate .texi dependencies by grepping for
268 @include. (If done, it should be done the same way C dependencies are
271 It would be good to check some parts of GNU standards. Already check
272 for install-sh and mkinstalldirs. What else is required to be in
273 package by GNU standards or by automake?
274 Some things for --strictness=gnits:
275 * "cd $(foo); something" is an error in a rule. Should be:
276 "cd $(foo) && something"
277 * Look for 'ln -s' and warn about using $(LN) and AC_PROG_LN_S
278 * Look for $(LN) and require AC_PROG_LN_S
280 Auto-distribute "ChangeLog.[0-9]+"? "ChangeLog.[a-z]+"?
282 Internationalize. [ gettext doesn't have the necessary machinery yet ]
283 am_error should use printf-style arguments (for eventual gettext scheme)
285 François says the ordering of files in a distribution should be as follows:
289 I agree, but I don't see how to implement this yet.
290 It might be easier if "derived files" is limited to those that
291 Automake itself knows about, eg output of yacc.
293 Check all source files to make sure that FSF address is up-to-date.
294 --gnits or --gnu only.
296 Merge each -vars.am file with corresponding ".am" file. Can do this
297 because of changes to &file_contents.
299 Should libexec programs have the name transform done on them?
301 Order the output rules sensibly, so FOO_SOURCES and FOO_OBJECTS are
302 together and rules are in the usual order.
304 Make the output minimal: only output definitions for variables that
308 David> To avoid comments like the one about subdirs getting buried in
309 David> the middle of a Makefile.in, how about pushing comments that
310 David> start with ### to the top of the Makefile.in (in order)? Sort
311 David> of like how Autoconf uses diversions to force initialization
312 David> code to the top of configure.
315 Karl> 2) Your Makefile variable names are generally uppercase, but GNU
316 Karl> generally uses lowercase. Not that it matters :-).
318 ================================================================
322 probably should put each group of m4 files into a subdir owned by the
323 containing application.
325 ================================================================
329 how to use the generated makefiles
332 - NORMAL_INSTALL junk
334 what goes in AC_CONFIG_AUX_DIR
338 SUBDIR entry must be direct subdir of this dir
342 multi-":" mode in AC_OUTPUT -- automake only looks at the first file
343 also a note on how a .am file is found in this case
345 rationale for avoiding
346 make CFLAGS="$CFLAGS" ...
349 a package that installs its own aclocal macros
351 write example of using automake with dejagnu
352 follow calc example in dejagnu docs
354 document which variables are actually scanned and which are not.
356 Document customary ordering of Makefile.am. From François.
358 Should include extended version of diagram from Autoconf (suggested by
361 Make a definition of the term "source"
363 document how to use Automake with CVS. Idea from Mark Galassi. Also
364 include Greg Woods' more sophisticated "cvs-dist" target.
366 document rebuilding configure. CONFIGURE_DEPENDENCIES
368 -- must document all variables that are supposed
369 to be public knowledge
371 automake must be run in each directory with a configure.in
372 This is insufficiently clear
374 must document the targets required for integration with
375 non-automake-using subdirs
377 use of (eg) EXTRA_PROGRAMS is not very clear right now
378 document EXTRA_foo_SOURCES
379 document why EXTRA_* vars must be statically knowable
381 document the "make SHELL='/bin/sh -x'" trick for debugging
383 section on relationship to GNU make
387 ================================================================
389 Things to do for autoconf:
391 * patch autoreconf to run automake and aclocal. I've done this but it is
392 not really available. It can't be made available until automake
393 is officially released
395 ================================================================
399 * Should support standalone library along with subdir library in same
400 Makefile.am. Maybe: turn off "standalone" mode if library's Makefile.am
401 is not only one specd? [ add an option for this ]
403 ================================================================
407 Would it be useful to integrate in some way with the Debian package
408 building utility? Must check. maybe it would be possible to deal
409 with all the different package utilities somehow. Lately I've been
410 hearing good things about the RedHat packaging utilities. Why are
411 there so many of these? Are they fun to write or something?
412 The RedHat package utility is called RPM; see
413 ftp://ftp.redhat.com/pub/code/rpm
414 It actually has problems, like no configure script and no documentation.
416 For Cygnus it would probably be good to be able to handle the native
417 package utility on each platform. There are probably 3 or 4 of these
418 (sysv, solaris?, aix?)
420 tcl/unix/Makefile.in has some code to generate a Solaris package.
422 Automake probably can't do all of this on its own. A new tool might
425 ================================================================
427 A tool to guess what the local Makefile.am should look like:
428 (see Gord's Maint program!)
430 * Probably integrate with autoscan
431 * Use various simple rules to determine what to do:
432 * get name of top directory, sans version info
433 * search for .c files with 'main' in them
434 * if in main.c, use directory name for program
435 * if in more than one, generate multiple programs
436 * if not found, generate a library named after directory
437 * order subdir searches correctly: lib first, src last
438 * assume 'testsuite' dir means we are using dejagnu
439 * maybe be smart about reading existing Makefile.am, so tool
440 can be run for incremental changes? You could imagine:
443 autoproject --incremental
445 ================================================================
447 Stuff NOT to do, and why:
449 consider auto-including any file that matches "*.in".
450 [ no: po/Makefile.in shouldn't be included ]
452 must look at mkid to see how it works (for subdir usage)
453 [ right now, it doesn't. i don't see a simple fix right now ]
455 if configure.in not found, move up a directory and try again? This
456 could eliminate a common source of problems.
457 [ this is just a bad idea ]
459 * scripts are installed in $exec_prefix/bin, not $prefix/bin
461 [ the consensus on Gnits is that this isn't required.
462 doubters can work around it anyway ]
464 * make the auto-dep code crash if GNU make not in use?
465 (doesn't it already?)
467 Looked at a program called 'ezmake', which seems to do something
468 similar. The only idea there that is possibly worth stealing is using
469 globs in definitions. Also has negations. Eg in a directory with
470 files a.c, b.c and c.c, the line:
471 foo_SOURCES = *.c ~c.c
472 would be equivalent to:
473 foo_SOURCES = a.c b.c
474 Is this worth implementing?
475 [ No... it is more reliable to spell everything out. ]
477 Scan source directories and warn about missing files, eg .c/.h files
478 that aren't mentioned?
479 [ distcheck makes this less useful ]