Re: Bugs which require external patches to reproduce

[Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>]

Zdenek Dvorak wrote:
> Hello,
>
>
> > > I just filed PR16282 which is a miscompilation of the tramp3d-v3 POOMA 
> > > testcase on ia64 with -fno-exeptions.  Unfortunately this failure 
> > > requires to have the leafify patch applied which seems to cause extra 
> > > stress on the inliner and loop optimizers.  Should I file such problems 
> > > with bugzilla?  I would probably be able to reproduce it without the 
> > > patch by imposing artificially high limits to the inliner but that would 
> > > cause compilation times for the testcase to grow significantly.
> >
> > I think, that, never-the-less, that is what you should do.  I don't 
> > think we should have PRs in our database that require the application of 
> > patches.
>
>
> in this particular case, could not just someone review his patch?  I see
> he has pinged it the second time some two weaks ago, without really ever
> receiving any comment about it, which seems quite annoying.

It may be because only global maintainers are able to approve the patch 
(I know of Jan who thought the cgraph parts are ok and I know of Andrew 
Pinski who thinks I should wait again for related new features to appear 
in mainline (the cfg inliner)).  Having the patch not in gcc is annoying 
for me only, because it complicates bug-reporting, building custom gcc's 
is not a problem for me.

But I was at a local CFD conference this week and got positive feedback 
on the general idea and it was suggested to try performance impact on 
libraries like MTL.

Thanks anyway,
Richard.