Add a --disable-fission option to mach xpcshell-test
Categories
(Testing :: XPCShell Harness, task, P3)
Tracking
(firefox102 fixed)
Tracking | Status | |
---|---|---|
firefox102 | --- | fixed |
People
(Reporter: mccr8, Assigned: jmaher)
References
Details
Attachments
(2 files)
It would be nice if there was a way to disable Fission when running XPCShell tests via mach.
Right now, you have to do it with --setpref fission.autostart=false
which is a little awkward.
Assignee | ||
Comment 1•3 years ago
|
||
well right now we don't run with fission enabled, but overall I think we should add this in. We will only be running xpcshell tests without fission on linux- would you prefer that we make this available on all platforms?
Reporter | ||
Comment 2•3 years ago
|
||
(In reply to Joel Maher ( :jmaher ) (UTC -0800) from comment #1)
well right now we don't run with fission enabled, but overall I think we should add this in. We will only be running xpcshell tests without fission on linux- would you prefer that we make this available on all platforms?
mach xpcshell-test
runs with Fission enabled, as far as I can tell. Which is a little confusing given that we run without Fission in automation. I don't know how useful this will really be, but yeah, it should be on all platforms. Really, it is just a bit of convenience and consistency.
Assignee | ||
Comment 3•3 years ago
|
||
when I get to cleaning up fission by default, then making --no-fission or --disable-fission will be the mode everywhere; I will keep this on my list for when I land the xpcshell w/fission changes (bug 1751492)
Assignee | ||
Comment 4•3 years ago
|
||
Updated•3 years ago
|
Comment 7•3 years ago
|
||
Backed out for causing multiple failures (exit status 2)
- Backout link
- Push with failures
- Failure Log
- Failure line: [taskcluster:error] exit status 2
Assignee | ||
Comment 8•3 years ago
|
||
thanks, I overlooked android hardware tests, fixed the patch and it is up for review.
Comment 10•3 years ago
|
||
Comment 11•3 years ago
|
||
Backed out for causing jit failures.
Backout link: https://hg.mozilla.org/integration/autoland/rev/292df8ed886d5b62860f0d88054c5cbb7e71aea1
Failure log: https://treeherder.mozilla.org/logviewer?job_id=377512045&repo=autoland&lineNumber=1526
The xpcshell failures jumped from tier 2 to tier 1
Assignee | ||
Comment 12•3 years ago
|
||
thanks; I see I overlooked an edge case on desktop- a couple extra try pushes and my patch looks good for another try
Comment 13•3 years ago
|
||
Comment 14•3 years ago
|
||
bugherder |
Comment 15•3 years ago
|
||
Is it possible this broke some of the benchmarks on AWFY? For example this hasn't been updating the past days: https://arewefastyet.com/win10/benchmarks/raptor-desktop-matrix-react-bench?numDays=60
Assignee | ||
Comment 16•3 years ago
|
||
this is a good question, I see tests still running on mozilla-central every commit:
https://treeherder.mozilla.org/jobs?repo=mozilla-central&tier=1%2C2%2C3&searchStr=matrix-react%2Cwin&fromchange=70a9f214d8d3a5e97b928d2a650a823ba368dea2
the task name hasn't changed at all, but the metadata for the replicates lost the fission
tag. I will look into this, it does seem like a regression from this patch.
Comment 17•3 years ago
|
||
How can we run fission geckoview-junit
tests after this change? Using --setpref fission.autostart=true
doesn't work and the test defaults to non-fission mode.
Assignee | ||
Comment 18•3 years ago
|
||
:agi, thanks for catching this. I assume you mean via ./mach
. in CI we run geckoview-junit-fis successfully. I think the area of code to fix is:
https://hg.mozilla.org/mozilla-central/rev/dfcdb4347811#l16.2
I suspect there is a simple fix for it. On my short list to look at today.
Comment 19•3 years ago
|
||
Yeah I meant locally, I verified in CI that we do the right thing and enable fission for fis
jobs. Thanks.
Assignee | ||
Comment 20•3 years ago
|
||
Comment 21•3 years ago
|
||
Comment 22•3 years ago
|
||
bugherder |
Assignee | ||
Comment 23•3 years ago
|
||
:jandem, benchmark data is now back- sorry for the ~2 day interruption.
:agi, your issue with junit is still high on my list.
Comment 24•3 years ago
|
||
(In reply to Joel Maher ( :jmaher ) (UTC -0800) from comment #23)
:jandem, benchmark data is now back- sorry for the ~2 day interruption.
Thank you! \o/
Description
•