This is the mail archive of the binutils@sourceware.org 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]
Other format: [Raw text]

Re: [PATCH] MIPS/gas: Redefined label alignment assertion failure fix


Hi Richard,

> >  The fix below removes the assertion failure making the "is already 
> > defined" test case behave as expected.
> >
> > 2009-09-12  Maciej W. Rozycki  <macro@linux-mips.org>
> >
> > 	* config/tc-mips.c (mips_align): Don't align the label if from 
> > 	another segment; remove the assertion failure for same.
> >
> >  Regression tested for the mipsel-linux and mips64-linux targets.  OK to 
> > apply to the trunk?  I suggest to propagate it to 2.20; we shouldn't be 
> > nourishing invalid assertion failures.
> 
> I suppose we first of all need to decide whether the target-independent
> code is doing the right thing by adding the faulty label to the
> segment's list in the first place, even though the label doesn't
> belong to that segment.  I've no opinion either way, just thought
> it needed to be asked.
> 
> Assuming the target-independent code is OK, then I agree your change is
> correct.  I'm just worried that, in order to see it's correct, you first
> need to go looking through the callers to see how they provide "label".
> I think it'd be cleaner to move the code inside mips_align itself, so
> that the logic is more obvious.  I realise that, with the branch in mind,
> you were probably wanting to be as non-invasive as possible, but what do
> you think of the following completely-untested alternative?

 Thanks for your notes -- unfortunately I had no time to look into it 
again and I agree it's a good idea to ckeck the target-independent code 
(though if the problem indeed lies there, then it's interesting why it 
should only hurt on MIPS -- there's nothing particularly specific to the 
MIPS architecture as far as label alignment is concerned).  I'm off for a 
holiday for some two weeks now, and while I may be able to have a look at 
this problem on my laptop some rainy evening, I expect not to have 
Internet connectivity for most of the time, so I may not be able to share 
my conclusions if any.  Therefore feel free to investigate it further, or 
otherwise if it can wait a fortnight, then I'll look into it when I am 
back.

  Maciej


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