[Mesa-dev] [PATCH] st/dri: Remove useless flush front.

Jose Fonseca jfonseca at vmware.com
Wed Jan 18 00:26:29 PST 2012



----- Original Message -----
> On Tue, Jan 17, 2012 at 15:47, Stéphane Marchesin
> <stephane.marchesin at gmail.com> wrote:
> > On Wed, Jan 11, 2012 at 10:13, Jose Fonseca <jfonseca at vmware.com>
> > wrote:
> >> ----- Original Message -----
> >>> Where else would the flush go?
> >>
> >> Maybe one of the callers already invoked glFlush, or something
> >> like that.
> >>
> >>> I'll look into fixing it differently
> >>> but if the flush is anywhere in the makecurrent path, the same
> >>> issue
> >>> will happen. Again, none of the classic dri drivers do it.
> >>
> >> If you wanna a short term fix, it might be OK with just commenting
> >> out the flush call, but please leave a note there saying that a
> >> flush is needed for strict conformance.
> >>
> >
> > Ok I'll push this for now as it fixes a bunch of Chrome bugs. I
> > guess
> > I'm confused why no other drivers do it, but maybe they're just all
> > non-conformant....
> >
> 
> Ah I understand how things work a little better now. I see two cases
> where the removal of this flush could be an issue:
> - The unbind-after-destroy case.  In that case we're good since the
> flush has happened in dri_destroy_context already.
> - Makecurrent path. Again in that case dri_make_current flushes the
> old context.
> 
> Therefore I think the flush in dri_unbind_context is not needed.
> Please let me know if you see a flaw in this analysis.

It looks you got all covered. And it would explain why other drivers don't need it. So your patch looks good.

Before commiting, please put a comment describing that flush already happens elsewhere, so that this issue doesn't pop up again. Thanks

Jose


More information about the mesa-dev mailing list