Re: an autoconf expert challenge

[Jim Meyering]

Bruno Haible <address@hidden> wrote:
> Ralf Wildenhues wrote:
>> # M4 code to put before or right after AC_INIT:
>> m4_ifdef([_m4_require_call],
>> [m4_rename([_m4_require_call], [_m4_require_call_orig])dnl
>> m4_define([_m4_require_call],
>> [m4_pushdef([AC_LIBOBJ],
>>   [m4_warning([AC_LIBOBJ called inside a require chain])dnl
>>    m4_expansion_stack_dump()])dnl
>> _m4_require_call_orig($@)[]dnl
>> m4_popdef([AC_LIBOBJ])])dnl
>> ])

Very nice.
Thank you, Ralf!

>> configure.ac:921: warning: AC_LIBOBJ called inside a require chain
>> glm4/getopt.m4:11: gl_GETOPT_SUBSTITUTE is expanded from...
>> glm4/argp.m4:7: gl_ARGP is expanded from...
>> configure.ac:47: gl_INIT is expanded from...
>> configure.ac:921: the top level
>> ...
>
> Here is a proposed patch that should fix all 11 problems. The dependencies
> to gl_GETADDRINFO and gl_SAVE_CWD can be removed, since the module 
> dependencies
> already ensure the macros get called. For the other 9 cases, invoking
> instead of requiring the macro is the solution.
>
> Jim, is this OK to commit (mostly your territory)?
>
> 2007-01-23  Bruno Haible  <address@hidden>
>
>       Don't AC_REQUIRE autoconf macros that invoke AC_LIBOBJ. See
>       <http://lists.gnu.org/archive/html/bug-gnulib/2006-10/msg00279.html>.
>       * m4/argp.m4 (gl_ARGP): Invoke, don't require, gl_GETOPT_SUBSTITUTE.
>       * m4/canon-host.m4 (gl_PREREQ_CANON_HOST): Don't require
>       gl_GETADDRINFO.
>       * m4/fts.m4 (gl_FUNC_FTS, gl_FUNC_FTS_LGPL): Invoke, don't require,
>       gl_FUNC_FTS_CORE.
>       (gl_FUNC_FTS_CORE): Invoke, don't require, gl_FUNC_OPENAT.
>       * m4/lstat.m4 (gl_FUNC_LSTAT): Invoke, don't require,
>       AC_FUNC_LSTAT_FOLLOWS_SLASHED_SYMLINK.
>       * m4/memcmp.m4 (gl_FUNC_MEMCMP): Invoke, don't require, AC_FUNC_MEMCMP.
>       * m4/mktime.m4 (gl_FUNC_MKTIME): Invoke, don't require, AC_FUNC_MKTIME.
>       * m4/openat.m4 (gl_FUNC_OPENAT): Invoke, don't require,
>       gl_FUNC_FCHOWNAT.
>       (gl_PREREQ_OPENAT): Don't require gl_SAVE_CWD.
>       * m4/strftime.m4 (gl_FUNC_GNU_STRFTIME): Invoke, don't require,
>       gl_FUNC_STRFTIME.
>       * m4/strtod.m4 (gl_FUNC_STRTOD): Invoke, don't require, AC_FUNC_STRTOD.
>       Reported by Ralf Wildenhues.

Yes, please do.
However, for a full solution, we need a mechanism
to keep such problems from being reintroduced.
Ralf, can you suggest something that would be a little more durable?
At worst, we can just use that guaranteed-to-change-someday code as-is,
but only for testing.