Re: PATCH: Fix ia64 gcc warnings

["H. J. Lu" <hjl at lucon dot org>]

On Thu, Nov 03, 2005 at 02:36:16PM -0800, James E Wilson wrote:
> On Thu, 2005-11-03 at 12:23, H. J. Lu wrote:
> > Jim, did you use "--enable-targets=all" when you were configuring
> > binutils on ia64?
> 
> I've answered one of my questions.  The code that emits this warning is
> pretty stupid, and doesn't fully take the CFG and variable lifetime info
> into account.  We can do much better.  Meanwhile, whether you get a
> warning depends on what optimizations have been performed, and what the
> optimizations did.  So we are going to have to work around this until
> gcc is fixed.  Which your patch does.
> 
> Incidentally, I'm getting a different error for gcc-4.1.  It is
> complaining about the argument info not the local variable buffer.  I
> doubt that your patch fixes that.  I don't have a convenient gcc-4.0
> install tree to try to see what warning I get there.
> 
> So yes, we do need your patch, and probably a few more, until gcc is
> fixed.

Even with my patch, gcc 4.0 gave me:

/net/gnu-13/export/gnu/src/binutils-import/binutils/opcodes/m68k-dis.c:
In function 'print_insn_m68k':
/net/gnu-13/export/gnu/src/binutils-import/binutils/opcodes/m68k-dis.c:1348:
warning: argument 'info' might be clobbered by 'longjmp' or 'vfork'
make[2]: *** [m68k-dis.lo] Error 1

Gcc 3.2, 3.4 and 4.1 are OK. I guess it is a gcc 4.0 regression.


H.J.