This is the mail archive of the
mailing list for the binutils project.
Re: Commit: Improve GAS testsuite results for various targets
- From: "Maciej W. Rozycki" <macro at imgtec dot com>
- To: Nick Clifton <nickc at redhat dot com>
- Cc: <binutils at sourceware dot org>
- Date: Wed, 22 Jun 2016 15:53:36 +0100
- Subject: Re: Commit: Improve GAS testsuite results for various targets
- Authentication-results: sourceware.org; auth=none
- References: <87porieapk dot fsf at redhat dot com>
> With a new branch coming up, I thought that it would be worthwhile
> spending a little time trying to improve the testsuite results for
> various targets. This first patch fixes various simple GAS testsuite
> failures, either by correcting the problem, or skipping them or
> marking them as expected failures if there is no reasonable
> expectation of them passing. Some failures do still remain, notably
> for MIPS based targets, and I hope to find time to look at these in
> the future.
What MIPS failures do you have specifically in mind? I know of just 3
that are arch-wide:
FAIL: objcopy -shared -z relro (tbss1)
FAIL: objcopy -shared -z relro (tbss2)
FAIL: objcopy -shared -z relro (tbss3)
which I have planned to look into, but kept pushing to a lower priority.
There is also:
.../ld/testsuite/ld-mips-elf/tls-multi-got-1-1.s: Assembler messages:
.../ld/testsuite/ld-mips-elf/tls-multi-got-1-1.s:39: Error: attempt to store non-zero value in section `.tbss'
ERROR: -EB -march=mips1 -32 -KPIC .../ld/testsuite/ld-mips-elf/tls-multi-got-1-1.s: assembly failed
UNRESOLVED: Shared library with multiple GOTs and TLS
which I think I'll just bisect to see what started it. I also need to
understand what the reason was to put initialised data in `.tbss' in this
test case -- maybe using `.space' with `tlsvar_ld' like with the other
symbols there will be equally good for coverage. Unfortunately the
circumstances around the addition of this test case look a bit obscure in
our repo history as the commit predates actual MIPS TLS support, so I'll
have to spend a bit to sort it out. Maybe it has never actually worked.
Then there a bunch of target specific issues. I think all the IRIX ones
can be ignored for the purpose of the impending release. Many of them are
a result of test framework shortcomings. I have an idea and a WIP change
to address some of them, selective testing will be more robust then.
Of the rest I've had a brief look at and I have some test case fixes to
address them, e.g. more STB_GNU_UNIQUE exclusions which I missed in my
previous clean up in this area. There are some generic bugs I've
encountered too while looking into these failures and and which I will
post fixes for once I'm done with the current outstanding MIPS work I
asked Tristan to wait for (I think I need to limit myself WRT bug
investigations as I tend to discover more bugs than I fix ;) ).