This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [patch] fix MIPS gas 'lineno' testsuite entry.


Nick Clifton <nickc@redhat.com> writes:
> [ ... MIPS gas lineno test fix ... ]
> Approved and applied.

Thanks.


FYI, the MIPS gas testsuite currently shows the following unexpected
failures with the following targets.  I'm going to look into them as I
find time.

mips-elf:

* "MIPS ELF reloc"

  Dunno what this is.  Symptom is 0 padding at end of section that
  didn't used to show up.  I believe you've already noted this in mail
  to others.


mips-ecoff:

* "MIPS jal-empic"

  Looks to me like a real regression.

* "MIPS la-empic"

  Not a regression, but instead breakage caused by the (reasonable)
  addition of an entry in the opcodes table (the one which does "la"
  -> "addiu" when possible).  There's a patch to address this in
  (hoping for somebody to give it a look-over):

	http://sources.redhat.com/ml/binutils/2001-02/msg00432.html

* "MIPS lifloat-empic"

  Not had a chance to look at this one yet.


(those are the only targets that i've tried recently.)

In case you're wondering why I've been testing mips-ecoff, it's not
that i have a great love for it.  From what I can tell, it included
the most complete and tested support for -membedded-pic, which I do
care about.  I've been using the known-somewhat-goodness for ECOFF to
help debug problems with ELF.

Waiting on my assignment paperwork to be finished, I've got some
additional patches relating to other -membedded-pic tests, including
adding a new 'la-empic2' test which actually tests the HI16 parts of
relocs.  (They're wrong right now for ELF.  Also, there's a patch that
fixes them to work properly for ELF... 8-)


Finally, as noted previously (the thread starting at
http://sources.redhat.com/ml/binutils/2001-02/msg00130.html), my
previous change to the 'empic' test was mostly wrong (the comment
fixes were right!), and should be backed out.  (The tc-mips.c patch in
that message is bogus tho.)

The existing code is somewhat broken w.r.t. 'la' and -membedded-pic.
My plan is, when I get a chance, I'm going to:

(1) back out my changes to the empic test case,

(2) make the trivial fix to tc-mips.c which will cause it not mistakenly
match 16-bit immediates where they don't actually make sense
(i.e. nuke the bogus chunk of code at ca. line 7828 of current
tc-mipc.c).

(3) actually fix the 'la' (etc.) macro handling to do the right thing
in the generated code.  I've got patches to fix this "properly" but
they're kinda big -- too big to go in w/o proper assignment in my
opinion.

I'll post some more patches about this when I get the chance.



chris





Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]