--- Begin Message ---
Subject: |
AC_CHECK_HEADERS after GUILE_FLAGS loses $CPP |
Date: |
Thu, 17 Mar 2016 03:47:04 +0200 |
I have a reduced test case where adding GUILE_FLAGS in shell dead code
before a call to AC_CHECK_HEADERS causes configure to lose $CPP, which
results in warnings like this one:
checking stdarg.h usability... yes
checking stdarg.h presence... no
configure: WARNING: stdarg.h: accepted by the compiler, rejected by
the preprocessor!
configure: WARNING: stdarg.h: proceeding with the compiler's result
checking for stdarg.h... yes
Versions:
/usr/share/aclocal/guile.m4: guile-2.0-dev from Debian jessie (2.0.11+1-9)
autoconf: 2.69
Attached:
configure.ac
guile.m4
To reproduce:
Find a config.rpath somewhere (perhaps steal it from gettext).
autoreconf -fi
./configure
I noticed that a random recent git HEAD of autoconf avoids this
problem; it turns out that in commit
11f520c61d8b21f1522968d6e6afb899070f0a6f autoconf transitions to
compilation-only header checks.
configure.ac
Description: Binary data
guile.m4
Description: application/m4
--- End Message ---
--- Begin Message ---
Subject: |
Re: bug#23034: AC_CHECK_HEADERS after GUILE_FLAGS loses $CPP |
Date: |
Tue, 28 Feb 2017 16:03:28 +0100 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/25.1 (gnu/linux) |
On Thu 17 Mar 2016 02:47, Bernd Jendrissek <address@hidden> writes:
> I have a reduced test case where adding GUILE_FLAGS in shell dead code
> before a call to AC_CHECK_HEADERS causes configure to lose $CPP, which
> results in warnings like this one:
>
> checking stdarg.h usability... yes
> checking stdarg.h presence... no
> configure: WARNING: stdarg.h: accepted by the compiler, rejected by
> the preprocessor!
> configure: WARNING: stdarg.h: proceeding with the compiler's result
> checking for stdarg.h... yes
>
> Versions:
>
> /usr/share/aclocal/guile.m4: guile-2.0-dev from Debian jessie (2.0.11+1-9)
> autoconf: 2.69
>
> Attached:
>
> configure.ac
> guile.m4
>
> To reproduce:
>
> Find a config.rpath somewhere (perhaps steal it from gettext).
> autoreconf -fi
> ./configure
>
> I noticed that a random recent git HEAD of autoconf avoids this
> problem; it turns out that in commit
> 11f520c61d8b21f1522968d6e6afb899070f0a6f autoconf transitions to
> compilation-only header checks.
AFAIU then this is an autoconf bug? Weird. Weird also that autoconf is
just not releasing these days! I guess we close this from the Guile
point of view though, no?
Regarding autoconf releases:
http://lists.gnu.org/archive/html/autoconf/2016-12/msg00005.html
So, I will optimistically close this and this problem will go away once
autoconf 2.70 is out.
Andy
--- End Message ---