779066 - Increases in idle service test failures in June/July

Status: NEW

(Core :: Widget, defect)

Description

[:Gijs (he/him)]

(Splitting out finding a solution from a bug with 10+ comment spams a day)

It's rather hard to get a time range for this, but it seems the idle service unit tests are failing much more frequently than they used to. As best I can tell the first increase in occurrence happened on June 7 (since then it's failed at least once every day). It got worse on June 12, still worse from about halfway through July. Of course, this being random orange, I'm not sure if those dates are even vaguely correct, have anything to do with code changes or just with increase in the number of VMs/checkins. I do know that by now I would be surprised if there was no real problem with the code, so I'm filing an actual bug.

From briefly looking at hg logs and the associated bugs, it seems a fuzz factor was recently added. Could this have something to do with it? CC-ing some folks who've been working on this more recently than me (I mean, I might have checked in the original test, but that was more than 5 years ago...).

Oh, and setting platform to XP, as that seems to be where the grand majority of the failures are (which might provide another clue as to what's up?).

Comment 1

[bsurender]

The fuzz factor is disabled for the Idle API tests. I will need to look into it a little further.