This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR32843, testsuite bug wrt sign/zero-extension of return values
Richard Guenther writes:
> On Thu, 26 Jul 2007, Andrew Haley wrote:
>
> > Richard Guenther writes:
> > > On Thu, 26 Jul 2007, Paolo Bonzini wrote:
> > >
> > > >
> > > > > This patch was approved by Andreas Tobler in the PR audit trail and
> > > > > committed to mainline.
> > > >
> > > > Couldn't it be also a FFI bug? If the ABI is unclear, I would have expected
> > > > libffi to do the cast itself...
> > >
> > > Maybe. Still it shouldn't matter -- the only valid thing to do with
> > > a signed char is to use it as signed char. Everything else requires
> > > a proper conversion.
> >
> > Perhaps so, but even if this is not strictly speaking a libffi bug, it
> > is a libffi quality of implementation issue, and it does potentially
> > break code.
> >
> > The intention of the testcase that has been broken is precisely to
> > check that signed characters are sign extended. When this fails
> > libffi should be changed, not the testcase.
>
> If you are familiar with libffi can you try to fix this then and
> revert the testsuite change?
Sure.
Andrew.