Re: PATCH: Fix PR other/37463 (All Solaris/x86 eh tests fail)

[Rainer Orth <ro at techfak dot uni-bielefeld dot de>]

Alexandre Oliva writes:

> On Oct 10, 2008, Rainer Orth <ro@techfak.uni-bielefeld.de> wrote:
> 
> > Indeed: results for all three combinations tested (gas 2.15/sun ld, gas
> > 2.15/gld 2.17.50 and gas 2.19.50/sun ld) are now practically identical.
> 
> > Any takers for the patch at hand
> 
> > 	http://gcc.gnu.org/ml/gcc-patches/2008-10/msg00249.html
> 
> > which fixes EH tests on Solaris/x86?
> 
> Patch looks good, thanks.  I'd have quoted "x$gcc_cv_objdump" in the

Thanks.

> test for != x, but that's just me ;-)

I've mimiced the other similar constructs in configure.ac here, and none of
them were quoted.  But I think you're right and will fix this.

> I wonder if it would make sense to take the more conservative path in
> case we don't have a working objdump to test, or perhaps try to
> exercise the bug in this case.  Thoughts?

The conservative path should already be taken, since without objdump
present gcc_cv_as_cfi_directive will not be set at all, which later breaks

AC_DEFINE_UNQUOTED(HAVE_GAS_CFI_DIRECTIVE,
  [`if test $gcc_cv_as_cfi_directive = yes; then echo 1; else echo 0; fi`],

since that assumes that the variable is set somehow.  So I'll make sure it
will be set on all paths.

>From problems like those, it seems configure.ac could considerably benefit
from a lookover by a shell/portability expert :-)

I'll revise the patch accordingly and resubmit once testing completes.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University