Re: C6X OSABI field

[Bernd Schmidt <bernds at codesourcery dot com>]

On 05/12/2011 10:41 PM, Joseph S. Myers wrote:
> On Thu, 12 May 2011, Bernd Schmidt wrote:
> 
>> ! tic6x-*-elf)		targ_emul=elf32_tic6x_elf_le
>> ! 			targ_extra_emuls="elf32_tic6x_elf_be elf32_tic6x_le elf32_tic6x_be"
>> ! 			targ_extra_libpath=elf32_tic6x_le
>> ! 			;;
>> ! tic6x-*-uclinux)	targ_emul=elf32_tic6x_linux_le
>> ! 			targ_extra_emuls="elf32_tic6x_linux_be elf32_tic6x_le elf32_tic6x_be"
>> ! 			targ_extra_libpath=elf32_tic6x_le
>>   			;;
> 
> I'm not sure why you're setting targ_extra_libpath, but I don't see how it 
> makes sense to have only one endianness in targ_extra_libpath in any case.  
> I think setting targ_extra_libpath=$targ_extra_emuls as various targets do 
> is safe in any case.

Yes, that seems to work.

>> + if { [istarget tic6x-*-elf] } {
>> +     set expected_osabi "Bare-metal C6000"
>> + } elseif { [istarget tic6x-*-linux] } {
> 
> That should be tic6x-*-uclinux.

Fixed.

>> +     catch "exec $READELF $READELFFLAGS --file-header $binary_file > readelf.out" got
> 
> Should be using remote_exec host (with appropriate adjustments for how 
> remote_exec arranges its return value) to work with remote host testing.

Ok. That's copied from ld-ifunc/ifunc.exp, so that probably needs fixing
too. It wasn't immediately obvious how to use remote_exec host; a few
versions that I thought should have worked didn't.

>> + if {! [check_osabi_tic6x tmpdir/libtest.so]} {
>> +     fail "C6X shared library OSABI, LE"
>> + }
> 
> I don't see any code to generate PASS lines here if the tests pass.

Fixed.


Bernd

Attachment: osabi-src2.diff
Description: Text document