Re: [PATCH] MIPS bfd: abi 64 ld -r

[Alexandre Oliva <aoliva at redhat dot com>]

On Sep  3, 2002, Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> wrote:

> I did test some slightly more complex testcase on IRIX, and I haven't
> seen such a case.

But did you actually look at the relocations generated by the IRIX
assembler for such a case, and verify that they had the addend in
place?  I know I've run into many problems lately, and simplified a
number of them to smaller testcases, but unfortunately they were in my
local build trees that didn't survive a reinstall of my desktop, in
which I reorganized all the local disk and thus erased all temporary
files that were not backed up.  I regret having forgotten I didn't
have other copies of those testcases, but they're now gone, so I can't
verify that this is one of the problems I've actually simplified or I
was just satisfied with the obvious fix.  But I know when I made the
change it was pretty obvious that we had a compatibility problem, in
which the assembler would put the addend of the relocation in place,
but the linker would just disregard it.  Both the IRIX native tools
and our tools were self-consistent, and giving the IRIX linker an
object file produced by our assembler worked just fine, but if my
recollection is correct, doing it the other way round didn't work,
precisely because our linker would ignore the in-place addends.

If this doesn't mean we should have partial_inplace set to true, I
don't know what does.

> I would like to see the exact case so I can try to reproduce it here.

I'll try to come up with something for you soonish.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer