Unable to run test-verify on try
Categories
(Firefox Build System :: Task Configuration, defect, P2)
Tracking
(firefox76 fixed)
Tracking | Status | |
---|---|---|
firefox76 | --- | fixed |
People
(Reporter: gbrown, Assigned: gbrown)
References
(Blocks 1 open bug)
Details
(Whiteboard: dev-prod-2020)
Attachments
(1 file)
'mach try fuzzy' and 'mach try fuzzy --full' do not present choices for 'test-verify'.
'mach try -b d -p linux64 -u test-verify' fails to run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=bb5d8daeaa5152631bc55636211249773b5c0559
Seems like a recent regression, but I'm not sure how recent...
Assignee | ||
Comment 1•4 years ago
|
||
test-verify seems to be running okay on mozilla-central and autoland.
test-verify is a tier 2 task, but I can select other tier 2 tasks.
The task definition explicitly includes 'try' for run-on-projects:
Assignee | ||
Comment 2•4 years ago
|
||
'mach taskgraph full -p project=try' displays expected test-verify tasks.
'mach taskgraph optimized -p project=try' does not.
Assignee | ||
Comment 3•4 years ago
|
||
Oh, but the list of tasks from 'mach taskgraph optimized -p project=try' is actually very short...reflecting what would run without any requests/parameters?
Assignee | ||
Updated•4 years ago
|
Assignee | ||
Comment 4•4 years ago
|
||
I suspect there is something wrong with the run-on-projects part of the task configuration; if I replace it with something simpler, I can push to try.
Assignee | ||
Updated•4 years ago
|
Comment 5•4 years ago
|
||
Note that -p project=try
grabs the parameters.yml
from the most recent push to try, which isn't all that meaningful, so will be random, depending on the most recent push to try.
I suspect that mach taskgraph target
(or ... target-graph
) is more intresting that mach taskgraph full
(if using a reasonable paramters file). The former will show the tasks requested to run (-graph
will include dependencies). Whereas full
will show all tasks that are defined.
It looks like what is going on in your pushes is that you have no local changes, so https://searchfox.org/mozilla-central/rev/5a10be606f2d76ef22f1f44565749490de991d35/taskcluster/taskgraph/util/perfile.py#58 returns nothing, and thus there are no chunks of test-verify created. Thus mach try fuzzy
does list them.
Assignee | ||
Comment 6•4 years ago
|
||
(In reply to Tom Prince [:tomprince] from comment #5)
It looks like what is going on in your pushes is that you have no local changes, so https://searchfox.org/mozilla-central/rev/5a10be606f2d76ef22f1f44565749490de991d35/taskcluster/taskgraph/util/perfile.py#58 returns nothing, and thus there are no chunks of test-verify created. Thus
mach try fuzzy
does list them.
It was possible in the (recent?) past to select test-verify with fuzzy or syntax without any modified files.
It was also possible to run test-verify with a specified path: mach try fuzzy <path/to/test> (bug 1400895); that doesn't work now -- test-verify is not included in the fuzzy list of tasks.
Now even if I modify a test and push to try with 'mach try fuzzy', test-verify is not included in the fuzzy list. Maybe we should debug the perfile.py code for that case, to start.
Assignee | ||
Comment 7•4 years ago
|
||
I find that 2 conditions are necessary to run test-verify (TV, TVg, TVwpt) on try:
- push must contain a test modification; AND
- --full must be included on the mach try command line.
Condition 1 is not required if I hack perfile_number_of_chunks().
Assignee | ||
Comment 8•4 years ago
|
||
Ensure a minimum of 1 chunk for "per-file" tests like test-verify when run on try, even when
no local file changes are detected. --full is still required.
Pushed by gbrown@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/03c92cf1a731 Make it easier to run test-verify on try; r=bc
Comment 10•4 years ago
|
||
bugherder |
Description
•