The following layout tests are failing on Win.
shinyak: If you can, in the future, add a link to where the tests started failing, that's helpful in getting these resolved.
These are forgotten rebaselines after: http://trac.webkit.org/changeset/131358.
Yes garden-o-matic refused to rebaseline these for me yesterday. I'll try again today.
OIC. This is because the win7 release bot is having trouble running. I wonder if webkit-patch should try to detect this case and fallback to the debug bot? Maybe that's crazy-talk.
(In reply to comment #3)
> OIC. This is because the win7 release bot is having trouble running. I wonder if webkit-patch should try to detect this case and fallback to the debug bot? Maybe that's crazy-talk.
Yeah, crazy talk :).
What would be _really_ useful is a way of getting results before landing a change.
We have the cbuild bots and ews but yet the only way to get platform specific results is to either manually run the tests and collect the results on all platforms or commit without results and update the expected results after the fact.
Not only would this make it easier to keep the tree green it would also greatly reduce the number of commits and thus improve cycle times for the cbuild. Not to mention allow one to see all test changes during the review process.
It would require a way to side-slip a change onto the bots though and a way to gather the results.
Getting results off the EWS bots is only blocked on having someone taking the time to do the legwork. Everyone agrees that would be better.
It would also require being able to request try jobs without uploading a patch (either to the ews or cbuild bots) and if we use the ews bots then we'll need to add one for each supported platform.
Not a trivial amount of work sadly.
(In reply to comment #7)
> It would also require being able to request try jobs without uploading a patch (either to the ews or cbuild bots) and if we use the ews bots then we'll need to add one for each supported platform.
I think just being able to 1) easily view the new results + diffs for each EWS bot and 2) choose some to get merged into your local patch, or the patch on bugzilla would be sufficient.
I don't think uploading a patch without a bug is a blocker here. If you can only do this for patches uploaded to bugzilla, that will cover 99% of patches. Getting EWS bots running tests for each platform will naturally happen once we have this. It's in each port's interest to setup such a bot to keep their trees green with minimal maintenance work.
Committed r131531: <http://trac.webkit.org/changeset/131531>
garden-o-matic finally let met rebaseline these. the CL only has 4 in it for some reason though it claims to have rebaselined 7.
There might be some further rebaselining to do still.