[FFmpeg-devel] [PATCH]Change check_exec_crash in configure

[Carl Eugen Hoyos]

M?ns Rullg?rd <mans <at> mansr.com> writes:

> >> > +__attribute__((noinline)) void foo() {
> >> 
> >> This is a gcc extension.  We should avoid using those in generic
> >> tests.
> >
> > Without attribute noinline, the test fails again for icc 11.0.
> 
> Then we need to find a better way to test it.

Unfortunately, I don't have other ideas (the original idea wasn't mine, either).
Have you?

> > So should I always force enable ebp_available for intel compilers?
> 
> Only if you can guarantee that it's always safe.  Can you?

Only for current source and current compilers.
Note that according to the comments, I believe passing our current test could
possibly be made necessary (it currently isn't, see icc) for using ebp, but not
safe.

Carl Eugen