This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: RFC: COMDAT group names become anonymouse local symbols
- From: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- To: mark at codesourcery dot com
- Cc: iant at google dot com, binutils at sourceware dot org
- Date: Sun, 24 Oct 2010 03:02:28 +0200
- Subject: Re: RFC: COMDAT group names become anonymouse local symbols
> Date: Sat, 23 Oct 2010 11:05:49 -0700
> From: Mark Mitchell <mark@codesourcery.com>
> >> 2010-10-21 Mark Mitchell <mark@codesourcery.com>
> >>
> >> * config/obj-elf.c (elf_adjust_symtab): New. Move group section
> >> processing here from elf_frob_file. Ensure that group signature
> >> symbols have the name of the group.
> >> (elf_frob_file): Move group section processing to
> >> elf_adjust_symtab.
> >> * config/obj-elf.h (elf_adjust_symtab): Declare.
> >> (obj_adjust_symtab): Define.
> >> * config/tc-arm.c (arm_adjust_symtab): Call elf_adjust_symtab.
> Committed, thanks.
It looks like this patch caused my cris-elf autotester to
regress. For the ld tests, after your commit:
Running /tmp/hpautotest-binutils/bsrc/src/ld/testsuite/ld-cris/cris.exp ...
FAIL: ld-cris/def2
FAIL: ld-cris/undef2
FAIL: ld-cris/warn1
FAIL: ld-cris/warn3
>From the ld.log:
Executing on host: sh -c {/tmp/hpautotest-binutils/cris-axis-elf/ld/../gas/as-new --em=crisaout -o tmpdir/dump0.o /tmp/hpautotest-binutils/bsrc/src/ld/testsuite/ld-cris/start1.s 2>&1} /dev/null ld.tmp (timeout = 300)
sh: line 1: 6009 Segmentation fault /tmp/hpautotest-binutils/cris-axis-elf/ld/../gas/as-new --em=crisaout -o tmpdir/dump0.o /tmp/hpautotest-binutils/bsrc/src/ld/testsuite/ld-cris/start1.s 2>&1
Same for the other FAILs.
I guess your new code is missing an adjustment to
gas/config/obj-multi.[hc] to add obj_adjust_symtab there to
handle (not break for) multi-objformat targets: when obj-elf.c
is compiled as not *the* object-format but one of several, and
when another format is selected.
(I'm open to just deprecating and removing the whole
multi-format machinery too, and cris-aout too, as long as you
make the --em=... option an assert. Patches welcome; I'm sure
global maintainers agree. :)
brgds, H-P