This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: 3.0 PATCH: Fix mips-tfile to avoid Tru64 UNIX V5.1 linker warning
Richard Henderson writes:
> Speaking of mips-tfile, what test case caused us to use -oldas?
This one:
http://groups.google.com/groups?q=orth+sullivan&hl=en&lr=&group=comp.unix.tru64&safe=off&rnum=1&ic=1
I've later learned from a Compaq engineer that this is due to an
optimization in adu which causes it not to emit `empty' sections. I've
received a test version (3.04.33) which fixes this, but right now has at
least three other problems:
* TARGET_AS_CAN_SUBTRACT_LABELS is back (I'll add a autoconf test for this
instead of hardcoding in osf5.h, unless I can convince Compaq to again
remove that limitation)
* object files it creates from empty assembler sources crash mips-tfile (I
have a fix for that, but still need to make sure it's really correct and
mips-tfile in error)
* the debugging info created by mips-tfile from object files assembled with
the new adu are hosed (probably mips-tfile is in error again)
I've postponed further investigating those until after the GCC 3.0 release,
since it seemed unlikely that this version of adu would become publicly
available before the release and I may not redistribute the test version,
so I needed to work around the problems with what is currently available.
> Turns out that ldgp problem is an assembler bug that is fixed
> in the new assembler.
Seems like we need another autoconf test for this one ;-)
Rainer
-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University
Email: ro@TechFak.Uni-Bielefeld.DE