[m-rev.] for pre-commit review: workaround Mantis bug #453

[Peter Wang]

On Sun, 19 Jan 2020 20:22:20 +1100 (AEDT), Julien Fischer <jfischer at opturion.com> wrote:
> 
> On Sun, 19 Jan 2020, Peter Wang wrote:
> 
> > On Sun, 19 Jan 2020 18:12:02 +1100 (AEDT), Julien Fischer <jfischer at opturion.com> wrote:
> >>
> >> For pre-commit review by Peter.
> >>
> >> In particular:
> >>
> >> - should we instead handle this by forcing the use of gcc -O0 in asm_fast
> >> grades on affected systes?
> >
> > Only in PIC mode.
> 
> So, force gcc -O0 in PIC mode only as opposed to this diff?

Yes.

> > We might want to disable the combination of gcc + asm_fast + x86 + PIC,
> 
> If you mean not install the PIC version of the libraries in asm_fast
> only, that's going to be complicated to arrange. I think just not
> allowing asm_fast in that case (as in the diff) is preferable.

Ok. I think forcing gcc -O0 in PIC mode is better.

> > or silently make asm_fast work like reg in that case.
> 
> I'm not sure what you mean by the latter.

I was idly wondering if mgnuc (etc.) could decline to pass
-DMR_USE_GCC_NONLOCAL_GOTOS -DMR_USE_ASM_LABELS if it detects the
combination of gcc + x86 + PIC; something like that.

Peter