Re: [patch added to 3.12-stable] compiler/gcc4+: Remove inaccurate comment about 'asm goto' miscompiles

[Ingo Molnar <mingo@kernel.org>]

* Jakub Jelinek <jakub@redhat.com> wrote:

> On Thu, Jan 28, 2016 at 11:56:53AM +0100, Jiri Slaby wrote:
> > From: Steven Noonan <steven@uplinklabs.net>
> > 
> > This patch has been added to the 3.12 stable tree. If you have any
> > objections, please let us know.
> > 
> > ===============
> > 
> > commit 5631b8fba640a4ab2f8a954f63a603fa34eda96b upstream.
> > 
> > The bug referenced by the comment in this commit was not
> > completely fixed in GCC 4.8.2, as I mentioned in a thread back
> > in February:
> > 
> >    https://lkml.org/lkml/2014/2/12/797
> 
> If you suspect a GCC bug, better try to narrow it down to a reasonably sized
> testcase, preprocess it and file in GCC bugzilla.  Otherwise it will never
> be fixed.

There's always the hope that someone hits an easier to analyze testcase! ;-)

OTOH the Linux kernel might be one of only a handful of large projects making 
heavy use of 'asm goto', right?

Btw., I think Linus complained at least about the debuggability of this:

  https://lkml.org/lkml/2014/2/12/805

  " [...] But I'm sure Jakub&co would love to have a test-case.

  Sadly, gcc has that really annoying habit of making small changes create *huge* 
  changes in label numbers etc, and that's definitely the case with the extra 
  empty asm - it's basically impossible to compare the generated asm with and 
  without the workaround, because all the label numbers change.

  I have no idea how gcc people debug things like this, when the output is so 
  unstable.

  Jakub, any suggestions to how Steven might be able to pinpoint where the code 
  generation problem lies?

                Linus"

Thanks,

	Ingo