(sorry for flooding) >From autoconf.info:
<< Avoid running test programs if possible, because this prevents people from configuring your package for cross-compiling. >> This means dropping the AC_TRY_RUN as a whole is a good idea after all... What we want to know is just the printf format string to use for size_t variables. IIRC, the gnu libc provides a %z modifier for exactly this, for any other libc implementation we need to know both size and signedness of it. Also, we want to know this for the *target* host, not for the build host. All in all, the simpler solution seems to be always casting the size_t variable into a known type and forget about SIZET_FMT as a whole. Comments ? --strk; On Wed, Jun 06, 2007 at 09:22:34AM +0200, strk wrote: > On Wed, Jun 06, 2007 at 09:08:34AM +0200, strk wrote: > > > Alright, found the problem. > > -g -O2 will NOT be pushed if CXXFLAGS is overridden by user (theoretically). > > Our setting CXXFLAGS=$save_CXXFLAGS seems to be taken as such an override. > > Even if $save_CXXFLAGS is empty !! > > > > To confirm just add the following in configure.ac: > > > > save_CXXFLAGS="$CXXFLAGS" > > CXXFLAGS="$save_CXXFLAGS" > > > > And the -g -O2 go away ... > > Now, I have a patch, that checks for "original" CXXFLAGS being empty or not. > If empty, it will be unset after the tests. > > This works fine, but prevents the user from setting CXXFLAGS to the empty > string > in his own WILL. > > Example: > > $ CXXFLAGS="" ./configure # will result in using default CXXFLAGS > > Same as: > > $ ./configure > > This is unless anyone knows a tip to check if a variable is set or not using > a bourne shell. > (ie: empty string counting as being set) > > --strk; > > > > > > > _______________________________________________ > Gnash-dev mailing list > [email protected] > http://lists.gnu.org/mailman/listinfo/gnash-dev -- () ASCII Ribbon Campaign /\ Keep it simple! _______________________________________________ Gnash-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash-dev
