Re: Probable m4 quoting problem

[Ralf Wildenhues]

Hello David,

* David Bruce wrote on Tue, Jun 22, 2010 at 06:37:12AM CEST:
> PKG_CHECK_MODULES([SDL_MIXER],
>                   [SDL_mixer],
>                   [],
>                   [AC_CHECK_LIB([SDL_mixer],
>                               [Mix_OpenAudio],
>                                 [],
>                                 [AC_MSG_ERROR([SDL_mixer not found!
> http://www.libsdl.org/projects/SDL_mixer])])])
> 
> 
> AC_DEFINE([HAVE_LIBSDL_MIXER],[1],[Define to 1 if you have the
> `SDL_mixer` library])
> CFLAGS="$CFLAGS $SDL_MIXER_CFLAGS"
> LIBS="$LIBS $SDL_MIXER_LIBS"

This looks correctly quoted, and if I stick it between
  AC_INIT
  AC_PROG_CC

and AC_OUTPUT, there is no warning for me.

> Problem is, I get errors on autoreconf -i such as:
> 
> address@hidden:/usr/local/src/git/tuxmath/build$ autoreconf -i ..
> configure.ac:116: error: possibly undefined macro: AC_CHECK_LIB
>       If this token and others are legitimate, please use m4_pattern_allow.
>       See the Autoconf documentation.
> autoreconf: /usr/bin/autoconf failed with exit status: 1

Well, the quoting bug may be in pkg.m4 not in your code.  Mine came from
pkg-config 0.22-1 as packaged by Debian.  Which is yours, do you have a
pointer to the exact version?

> If I run autoreconf a second time, it succeeds, so perhaps this
> "error" is actually a warning, but nevertheless I want to get rid of
> it.

That's a caching bug in Autoconf, I think.

> Also, this approach generates somewhat misleading ./configure output,
> as it shows the unsuccessful pkg-config check, followed by the
> successful AC_CHECK_LIB.  I've tried adding explanatory "echo"
> statements but have encountered more quoting errors. Is AC_MSG_NOTICE
> better, and if so, how do I group multiple macros within the
> "action_if_not_found" of PKG_CHECK_MODULES.

I think you should show what you tried, preferably with a short but
complete configure.ac file.  Generally, you can avoid intertwining
multiple 'checking for ...' with their results by moving the second test
out of the IF-FAILS argument, setting a variable there (you can use a
cache variable for a better recheck experience) and invoking the second
test later conditionally only, wrapped in AS_IF.  But I'm not sure
whether this is your issue.

Hope that helps.

Cheers,
Ralf