https://sourceware.org/bugzilla/show_bug.cgi?id=33162
--- Comment #12 from Rainer Orth <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #11 from Nick Alcock <nick.alcock at oracle dot com> ---
> OK, so it's kind of acceptable to have required configurations for a GCC used
> for binutils testing, then. I wasn't sure. Time to build a new GCC locally on
> that cfarm box :)
I guess so. You may not run into (so many) issues on systems that
natively use GNU as and ld as /usr/bin/as, /usr/bin/ld.
One more point, btw.: be sure to use
configure 'CC=gcc -m32' 'CXX=g++ -m32'
when building 32-bit-default gcc's: while it should be possible to build
a 64-bit gcc binary that is 32-bit-default, there have been problems
trying that, so best avoid the possibility.
> (it *is* probably valuable to have a non-GNU linker compiler in the test
> matrix, if just because there are configury paths specific to it. I think I've
> figured out how to force the testsuite off in that case, too -- do a CTF link
> and see whether it dedups properly. If it just concatenated, it's not the
> right
> linker. I'll see if I can hack that up...)
Right, but I guess that's a completely different can of worms,
especially since there are tons of gld (and gas)-specific options used
in the testsuite that make Solaris ld choke.
As for checking for non-GNU ld, I believe there are checks if gcc -B.../
will use a freshly built ld. Maybe they just need extending...
--
You are receiving this mail because:
You are on the CC list for the bug.
