This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH]: Set little endian flag when linking SH arch tests for little endian multilibs
Andrew Stubbs wrote:
For reasons I do not understand I found that if I included the above
part of the patch to arch.exp I ended up with *more* SH linker
testsuite failures (for little endian multilibs) than without it.
This was another thing that I was going to investigate when I had more
time. If you can show that it does reduce the number of testsuite
failures however then I would be happy to accept it.
Hmmm, did you use both parts of the patch? You say that I have had time
to test this, but actually the reason I had not submitted the patch was
that I had not yet finished testing it. I suspect that a few more bits
are needed. I also have a few more adjustments to the arch.exp script to
submit. Also, this has got put on the back burner for the last few weeks
so it isn't fresh in my mind any more.
I'll look into it. Not sure when though.
Well, I did the tests with and without my two-part patch.
Without:
=== ld Summary for sh-sim ===
# of expected passes 853
# of unexpected failures 15
# of expected failures 1
# of unresolved testcases 1
# of untested testcases 21
=== ld Summary for sh-sim/-ml ===
# of expected passes 133
# of unexpected failures 357
# of expected failures 1
# of unresolved testcases 14
# of untested testcases 378
=== ld Summary ===
# of expected passes 986
# of unexpected failures 372
# of expected failures 2
# of unresolved testcases 15
# of untested testcases 399
With:
=== ld Summary for sh-sim ===
# of expected passes 853
# of unexpected failures 15
# of expected failures 1
# of unresolved testcases 1
# of untested testcases 21
=== ld Summary for sh-sim/-ml ===
# of expected passes 797
# of unexpected failures 80
# of expected failures 1
# of untested testcases 21
=== ld Summary ===
# of expected passes 1650
# of unexpected failures 95
# of expected failures 2
# of unresolved testcases 1
# of untested testcases 42
So that's quite considerably better. I have not looked at what causes
the 65 additional failures in little endian, and I do not know when I
will get chance to. I do know that the assembler tests also fail some
little endian due to a lack of support for it in certain target
variants. At least I now know why I never got around to submitting the
patch.
The other testsuite patches I alluded to are, it turns out, related to
changes in behaviour caused by another patch I have not yet submitted,
so you don't need to worry about that just yet.
Andrew Stubbs