Re: PATCH PING: Build machinery: Set poststage1_ldflags for all builds

[Alexandre Oliva <aoliva at redhat dot com>]

On Nov 29, 2010, Ian Lance Taylor <iant@google.com> wrote:

> - if test "$ENABLE_BUILD_WITH_CXX" = "yes" -a "$poststage1_libs" = ""; then
> + if test "$poststage1_libs" = ""; then
>    poststage1_ldflags="-static-libstdc++ -static-libgcc"
>   fi

In principle, I think the change fixes an inconsistency, so I don't
oppose it.

However, given the presented concern:

> the newly built and installed compiler will, in general, only run if
> the gcc libraries are on the dynamic linker search path

I get a feeling we might be better served by using libtool to link GCC
executables.  This would take care of finding libgcc and libstdc++ both
inside and outside the build tree, and would enable us to take advantage
of dynamic linking to save disk space and, more importantely, memory for
compiler users.  If we feel adventurous, even libbackend could be turned
into a dynamic library, avoiding the long link times I experience,
mostly out of copying libbackend, when linking in parallel all
front-ends with a bunch of concurrent (-O1/-O2/-O3 -m32/-m64) builds.

> People who build distros, or in general people who know that they will
> install the newly built libraries in a place where the dynamic linker
> will find them, may use the existing configure option
> --with-boot-ldflags to control this.

Hmm, I don't think it's that easy.  We want to use the stage1 libstdc++
while running the stage2 compiler to build both stage2 libstdc++ (stage1
in prev-) and stage3 (stage1 in stage1-) front-ends, and we want to use
the stage2 libstdc++ (in prev-) while running stage3 g++ to build stage3
libstdc++.  I don't see how --with-boot-flags can be set so that this
works, and so that this doesn't encode build-time library search paths
in binaries that are going to be installed.  We don't want installed
compilers to search these temporary directories for libstdc++.

Now, it's clearly far too late to make this sort of major change at this
point in the 4.6 cycle, and -static-* is such an easy work-around for
these complexities, that I'm inclined to accept it.

So, if you get no objections in the next 48 hours or so, please go ahead
with the patch.

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist      Red Hat Brazil Compiler Engineer