On Tue, May 19, 2015 at 12:06:54PM +0200, Daniel Vetter wrote: > On Mon, May 18, 2015 at 05:12:17PM -0700, Dylan Baker wrote: > > I'm not sure this is a good idea, > > > > Ken Graunke and I talked this to death when I did the refactor that > > split the run and resume code into separate paths, and we agreed that > > changing the tests during a resume would make it impossible to reproduce > > a run, since the regex you pass might exclude tests that have already > > completed. > > > > That said, I'm opened to being convinced it is a good idea. > > With your patches to mark a test as incomplete, should we instead just > have an option on resume to mark all incomplete tests as failed and not > try to restart them? Or is that already what happens (tbh I haven't > tried). > -Daniel
Currently they're always retried. It seems reasonable to me to either change the behavior or add a switch to change the behavior to either mark them as failed or just leave them as incomplete. [snip]
signature.asc
Description: Digital signature
_______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
