Open Bug 1478217 Opened 2 years ago Updated 2 years ago

Bisection stuck in infinite loop


(Testing :: mozregression, defect)

Not set


(Not tracked)


(Reporter: gingerbread_man, Unassigned)




(1 file)

I was testing for bug 1478185. Bisection options were the defaults; I don't recall what starting range I picked. The problem occurs around this spot:

1. Bisecting on mozilla-central [9302fd8c - 3edc9c3a] 70f901964f9725e6dfd09750f48996e9d6670492 rated bad
2. Bisecting on autoland [22caf5b0 - 12f49235] e7872b6bba62397f3849e924869711d3ba868578 rated good
3. 9cbe048f7395f806579c6e878a4166672ecd1639 rated good
4. 3f8d4ac2adae867b0b6bcc98059fb9c035c07017 rated good
5. b2cd0e01b6556147c4b0c51960f668ec71580d95 rated good
6. b4cd5a640564cee2dda926724bf2faeb8934d702 rated good
7. Bisecting on mozilla-central [9302fd8c - 3edc9c3a] ...

Actual results:
Builds that were already tested are tested again.

Expected results:
For the bisection to end at some point.
Flags: needinfo?(wlachance)
This looks like a sheriff mistake, I'm seeing both inbound- and autoland-landing bugs in that central push[0].  I seem to remember something recently...
Ah, indeed:

>Jul 10 23:25:16 <RyanVM>	CosminS|sheriffduty: it looks like there was more in that merge than you meant there to be?
>Jul 10 23:25:29 <RyanVM>	i.e. you pushed inbound tip to m-c
>Jul 10 23:25:37 <CosminS|sheriffduty>	yeah, that's right
>Jul 10 23:25:40 <Aryx>	RyanVM: yes, got pinged about that
>Jul 10 23:25:41 <RyanVM>	and called it an autoland merge
>Jul 10 23:25:49 <RyanVM>	mozregression is going to love that
>Jul 10 23:26:07 <CosminS|sheriffduty>	I'm monitoring the latest pushes from inbound closely
>Jul 10 23:26:45 <CosminS|sheriffduty>	did not update correctly to central and was on the tip of inbound
>Jul 10 23:27:25 <RyanVM>	eww, so that push has inbound *and* autoland commits in it?
>Jul 10 23:27:39 <Aryx>	yes
>Jul 10 23:27:49 <CosminS|sheriffduty>	yes, the tip from inound and the merge candidate from autoland
>Jul 10 23:28:04 <RyanVM>	yeah, mozregression is going to love that :)

Not sure what mozregression could do in this situation.  Maybe try and remember that it's already visited a range on a tree and not repeat it, but that seems like unjustified effort for a very rare situation.

I suppose another possibility is manually blacklisting some m-c pushes, and bailing out explaining the problem.

Flags: needinfo?(wlachance)
Duplicate of this bug: 1483903
You need to log in before you can comment on or make changes to this bug.