Expose the chromium waterfall in sheriff-o-matic.

Expose the chromium waterfall in sheriff-o-matic.

-Store every waterfall's failures in ct-failure-analyzer
-Pick the list of failures based off the tree and link to the
right flakiness dashboard groups based off the tree.

NOTRY=true
R=eseidel@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=179197

[ojan, 2014-07-25 05:25:24]

I think this is the last step in getting a minimal viable product for chromium
sheriff's to use in terms of the frontend. There's still a bit of work to do on
the backend.

I took a bit of a shortcut with the tree->master url mapping. I really should
grab that out of gatekeeper, but I wanted to get this up and that list doesn't
change very often, so it's fine to hard-code for now. It has a FIXME.

[eseidel, 2014-07-25 16:40:17]

lgtm

I don't feel super well versed in this code.  This seems fine.  But would be
nice to have someone who's hacked on GOM more approve this.

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/script...
File Tools/GardeningServer/scripts/ui.js (right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/script...
Tools/GardeningServer/scripts/ui.js:32: // FIXME: Replace the flakiness
dashboard's sense of groups with sheriff-o-matic's
IMO the flakiness dashboard should just operate in terms of masters.  Or we
should just get rid of masters entirely and have globally unique builder names.

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
File Tools/GardeningServer/ui/ct-embedded-flakiness-dashboard-tests.html
(right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
Tools/GardeningServer/ui/ct-embedded-flakiness-dashboard-tests.html:25: var tree
= "blink";
auto-sheriff happily handles failures which don't have a 'tree'.  Right now I
only have it polling masters in the gatekeeper.json file, but I could trivially
change it to poll all masters in teh known universe.  Are the set of masters
defined by 'blink' and 'chromium' trees the only ones SOM plans to support?

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
File Tools/GardeningServer/ui/ct-failure-analyzer.html (right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
Tools/GardeningServer/ui/ct-failure-analyzer.html:18: _trees: {
Isn't this just as simple as atob(xhr.responseText)?

[ojan, 2014-07-26 17:31:51]

Thx for the review. I forgot to mention that this patch depends on
https://codereview.chromium.org/413073004 and
https://codereview.chromium.org/416673003. Will land this once those land.

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/script...
File Tools/GardeningServer/scripts/ui.js (right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/script...
Tools/GardeningServer/scripts/ui.js:32: // FIXME: Replace the flakiness
dashboard's sense of groups with sheriff-o-matic's
On 2014/07/25 16:40:17, eseidel wrote:
> IMO the flakiness dashboard should just operate in terms of masters.  Or we
> should just get rid of masters entirely and have globally unique builder
names.

The flakiness dashboard at least needs the tree concept here so that we can show
you the right list of builders when we show you the embedded flakiness dashboard
in sheriff-o-matic, e.g. if Foo.Test fails on linux and windows, we want to show
you both builders, but those are on different masters.

Another way we could do that is that the flakiness dashboard could take a list
of masters and sheriff-o-matic would be responsible for passing it the list of
masters.

Globally unique builder names SGTM, but that requires changing a lot on the
buildbot side and enforcing it is difficult.

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
File Tools/GardeningServer/ui/ct-embedded-flakiness-dashboard-tests.html
(right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
Tools/GardeningServer/ui/ct-embedded-flakiness-dashboard-tests.html:25: var tree
= "blink";
On 2014/07/25 16:40:17, eseidel wrote:
> auto-sheriff happily handles failures which don't have a 'tree'.  Right now I
> only have it polling masters in the gatekeeper.json file, but I could
trivially
> change it to poll all masters in teh known universe.  Are the set of masters
> defined by 'blink' and 'chromium' trees the only ones SOM plans to support?

The other masters will be on some tree somewhere. So we'll just need the right
mapping. I've been punting on this problem since the short-term goal is just to
support these two trees. It's hard to design around supporting other things when
we don't know what we need to support.

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
File Tools/GardeningServer/ui/ct-failure-analyzer.html (right):

https://codereview.chromium.org/418253002/diff/1/Tools/GardeningServer/ui/ct-...
Tools/GardeningServer/ui/ct-failure-analyzer.html:18: _trees: {
On 2014/07/25 16:40:17, eseidel wrote:
> Isn't this just as simple as atob(xhr.responseText)?

Pretty much. It needs doing the network request, atob and then munging the
response into this format. It's not difficult. I just didn't feel like doing it
as part of this patch. I was also thinking it might make sense to skip pulling
this file again if we've already got the data since the data changes so
infrequently. After waffling on that for a few seconds I decided to just punt
for now and leave it as a FIXME.

[ojan, 2014-07-30 04:37:38]

Committed patchset #2 manually as r179197 (tree was closed).

[ojan, 2014-07-30 05:04:28]

Whoops. Accidentally committed https://codereview.chromium.org/413073004 with
this.