1204946 - Classifying failures can be midaired by someone else, returns an HTTP 500

Status: RESOLVED DUPLICATE of bug 1221936

(Tree Management :: Treeherder, defect, P2)

Description

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

Reproducing this takes a bit of good timing.
If two people have a Treeherder tab open to the same tree, it's possible for person A to classify a failure and then person B tries to classify that same failure (before the UI updates to reflect A's new classification), causing Treeherder to throw an error when B tries to submit their classification.

The end result is that A's classification is posted successfully to the bug, but B's classification isn't.

If B looks at their network monitor while submitting their classification, they'd see that the POST to https://treeherder.mozilla.org/api/project/[tree]/bug-job-map/ is returning a 500 status.


I'm not sure exactly what could be done here. Maybe attempt to fetch the job's updated classification prior to submitting it to see if it was already classified by someone else? Catch that 500 ISE internally and throw a more useful error?

Comment 1

[Ed Morley [:emorley]]

We should definitely do something about the HTTP 500 - it means the API is not handling this properly (a more sensible response code would be say 409 - http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.4.10 ).

Ideally the UI would also handle this by updating to show the mid-air too.

Comment 2

[Phil Ringnalda (:philor)]

The race possibilities include even more fun possibilities than that: some recent day, you and I raced so closely that we both commented on the bug, and treeherder showed both of us as classifying it intermittent but didn't show either of us as mentioning the bug.

Comment 3

[Ed Morley [:emorley]]

The 500 was fixed in bug 1221936, however there's likely still UI work to do that makes the mid-air easier to follow.