check mips abi x linker emulation compatibility

[Alexandre Oliva]

On Mar 28, 2003, Alexandre Oliva <aoliva@redhat.com> wrote:

> On Mar 23, 2003, Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> wrote:
>> IMHO it's better to change the GAS default to n32 for mips64-unknown-linux
>> and mips-sgi-irix6, because that is/will be the most heavily used ABI
>> variant on NewABI capable systems.

> Ok, here's a patch that fixes a typo in my earlier bfd patch, changes
> the assembler default ABI as you suggest, adjusting the assembler and
> the linker testsuite.  The assembler testsuite is currently very
> broken, because of a patch that Eric installed the other day, but I
> reverted his patch in my tree and, after the patch, got failures only
> in the empic tests, that have been broken for a while.  The linker
> testsuite was a bit trickier to fix: a number of tests seem to
> actually depend on o32 behavior, and a few of the flag-compatibility
> tests go further and actually rely on the no-abi default of the
> assembler, which doesn't hold for mips64-linux-gnu.  I'm thinking of
> simply disabling the 5 failing tests that depended on this behavior
> (perhaps only for mips64-linux-gnu?).  They are:

> FAIL: MIPS incompatible objects:  "-mips32" "-mips64"
> FAIL: MIPS incompatible objects:  "-mips32r2" "-mips64"
> FAIL: MIPS compatible objects:  "-mips4 -mgp32" "-mips2"
> FAIL: MIPS compatible objects:  "-mips2" "-mips4 -mabi=eabi -mgp32"
> FAIL: MIPS compatible objects:  "-march=sb1 -mgp32" "-march=4000 -mgp32"

It turned out I had a goof in my linker default.exp patch that caused
the failures in `incompatible' objects.  As for the compatible
failures, the problem was wrong expectations.  The -mips2 case
in the second `compatible' test would not have any ABI, but without a
NO_ABI as default, it would not be compatible with -mabi=eabi.  The
other two failed because 32bitmode would no longer be set, given that
it is only set if mips_abi == NO_ABI in the assembler.

Here's a revised patch, that passes all tests on mips64-linux, and
doesn't introduce any regression on mips-elf.  Ok?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: mips64-linux-bfd-check-abi-compat.patch
Type: text/x-patch
Size: 1388 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20030507/8e2103cf/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mips64-linux-gas-default-abi.patch
Type: text/x-patch
Size: 1973 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20030507/8e2103cf/attachment-0001.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mips64-linux-gas-default-abi-tests.patch
Type: text/x-patch
Size: 62973 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20030507/8e2103cf/attachment-0002.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mips64-linux-ld-test-abi.patch
Type: text/x-patch
Size: 7581 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20030507/8e2103cf/attachment-0003.bin>
-------------- next part --------------

-- 
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