This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: Result of failures for first test release for 2.11
- To: philb at gnu dot org
- Subject: Re: Result of failures for first test release for 2.11
- From: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- Date: Sun, 18 Mar 2001 21:39:34 +0100
- CC: binutils at sources dot redhat dot com
> Date: Sun, 18 Mar 2001 10:54:40 +0000
> From: Philip Blundell <philb@gnu.org>
> >check-cris-elf.log:FAIL: finding out ELF size with readelf -h
> >check-cris-elf.log:FAIL: readelf -h
>
> Most, or maybe all,
All. (I do run the testsuite now and then. ;-)
> of these CRIS failures are caused by an oddity in the way
> configuration works for this target. Configuring for "cris-elf" actually gets
> you an assembler that defaults to producing a.out. This seems strange, but I
> guess it was done that way for a reason.
Hysterical raisins. People invoking the assembler by itself
(not through gcc, and with no specified object-format) were
historically expecting a.out output. Actually, cris-elf isn't
used as much as just cris, which defaults to cris-axis-none,
(which might be better expressed as cris-axis-multi) which used
to be an a.out-only target.
I think I'll change that to default to ELF, at least for the
cris-axis-elf configuration(!), as it helps people doing bulk
testing of binutils, for one, and being more sane for another...
FWIW, with current sources configured to cris-elf, to get the
expected results, use:
env AS_EMULATION=criself LDEMULATION=criself make check
People configuring for --target=cris or perhaps
--target=cris-aout will see no FAIL:s. The testsuites assume
config triples ending in "elf" are fine to run ELF-specific
tests on.
Sorry for the confusion.
brgds, H-P