Re: [PATCH] pop previous section in alternative.c

[Andy Whitcroft <apw@shadowen.org>]

On Thu, Apr 10, 2008 at 10:47:18AM +0200, Andi Kleen wrote:
> "H. Peter Anvin" <hpa@zytor.com> writes:
> >>>
> >> Oh, and this would not be complete without giving Andrew Pinski
> >> complete
> >> credit for telling me it wasn't a gcc bug but a bug in the toplevel asm
> >> code in the kernel. ;-)
> 
> We've actually had such bugs before, it isn't the first time.
> 
> >
> > In many ways it's kind of silly for this even to be in assembly, since
> > all it is is a sequence of comma-separated byte values; I guess it was
> > the easiest way to deal with it given the ".byte" prefix, but still...
> 
> The nops are primarily used in inline assembler statements (in alternative)
> and only once in this table. Not using this would have meant to write
> them all twice which would have been nasty. There is also no sane 
> way to get standard arrays into inline assembler as instructions.
> 
> BTW it looks like the problem was added with 121d7bf5a246d282ba91234d03a4edf9ccc9c940,
> signed off by me, sorry for not catching it in review.
> 
> Perhaps that is something that would make sense adding to checkpatch.pl?
> Complain for .section in inline assembler without .previous or popsection
> (cc Andy). I think such a check would make sense. 

Do you have an example of such a bad thing?  My only concern is this
sounds like a check which could potentially need to see more lines than are
available in the patch context and so the test might be rather unreliable.

-apw