Re: [sr #108201] libtool problems with -export-symbols-regex on solaris with gcc-4.7.x

[Dennis Clarke]

> Follow-up Comment #20, sr #108201 (project libtool):
<snip>
> There was a (substantially long) time where active libtool maintainers 
> were no longer able to bootstrap the project from git.  Luckily that 
> issue was resolved so we are able to work with the master version
> again and usefully apply well-tested patches.

This feels like a release gate has closed and a new, perhaps significant,
release is on the way.  My interest in this is exclusively related to a ld
and as variant that may work well on a modern commercial UNIX such
as Solaris. For a bit of history there was a project within OpenSolaris
in which we ( OpenSolaris people ) actually showed an interest in GNU
as and ld and gcc being used to bootstrap the OS. There was, of course,
no corporate interest in such a thing and the project languished in dust
with little or no support from Sun engineers.  That is all history and we 
now have the illumos project which managed to get fairly full support
from a corporation and even garnered some Sun kernel engineers and 
other heavy hitting code gurus.  There yet may be life in the idea of a 
modern GNU linker, assember and compiler. 

To date I have been unable to get a reasonable result with GCC 4.7.2 on
Solaris 10 using ( or not ) GNU as and GNU ld. Strangely I can get really
excellent results on very very old obsolete Solaris 8. This is a mystery to
me. ( see http://gcc.gnu.org/gcc-4.7/buildstat.html )

So for quite some time now I have been watching the GNU binutils and the
GNU compiler collection project with interest and have been running test
builds in various ways on a slew of strange servers, from the obscure ( I have
a functional DEC Alpha and Freescale MPC5200B 32-bit PPC ) to latest Fujitsu
Sparc ( Niagara based units ).  Some on Debian, some on Solaris, some on 
other stuff like Red Hat Enterprise Linux 6.  All do what they do and are 
seen running, strolling or sitting powered off on a regular basis. It seems, to
me, that lately the GCC and binutils code simply won't bootstrap and pass
its own testsuite.  Most of the time.  Therefore it is my hope that this gate
that is swinging in the wind closes and the maintainers are able to make a
significant jump/leap/fall forwards with code that can bootstrap once again.

Sorry for the long message but I have been lurking a lot and waiting for a 
shoe to drop.

Dennis Clarke 
UNIX Systems programmer