source-test-puppeteer-puppeteer-with-bidi logs too large to be parse by Treeherder
Categories
(Remote Protocol :: WebDriver BiDi, defect, P2)
Tracking
(firefox-esr115 unaffected, firefox126 unaffected, firefox127 unaffected, firefox128 fixed)
Tracking | Status | |
---|---|---|
firefox-esr115 | --- | unaffected |
firefox126 | --- | unaffected |
firefox127 | --- | unaffected |
firefox128 | --- | fixed |
People
(Reporter: aryx, Assigned: whimboo)
References
(Regression)
Details
(Keywords: regression, Whiteboard: [webdriver:m11])
Attachments
(1 file)
The source-test-puppeteer-puppeteer-with-bidi task logs are too large to be parsed by Treeherder (threshold: 5 MB gzipped). Failure lines are not shown, hence the log size should be reduced. Task example
Assignee | ||
Comment 1•6 months ago
|
||
This is a regression from the last Puppeteer sync as done on bug 1897183.
Comment 2•6 months ago
|
||
Set release status flags based on info from the regressing bug 1897183
:Sasha, since you are the author of the regressor, bug 1897183, could you take a look? Also, could you set the severity field?
For more information, please visit BugBot documentation.
Assignee | ||
Comment 3•6 months ago
•
|
||
This happens most likely because of the test AriaQueryHandler queryAllArray $$eval should handle many elements (ariaqueryhandler.spec.js)
.
Here 10.000 elements of type button
are getting created within a single evaluate call, then retrieved via browsingContext.locateNodes
as sharedId
, evaluated individually to get a handle, evaluated once to call reduce
, and finally the handles are disowned individually.
0:05.23 pid:22179 1716817055602 RemoteAgent DEBUG WebDriverBiDiConnection 7afd76bf-0fcd-4335-a43e-55b729ebe1de -> {"id":156,"method":"script.callFunction","params":{"functionDeclaration":"node => {\n return node;\n }\n//# sourceURL=pptr:internal\n","awaitPromise":true,"target":{"context":"0a3ff222-04e6-43df-aec1-8476cd80f724"},"arguments":[{"type":"node","sharedId":"ea2b5108-5e80-4095-a950-bdf65e7def14","value":{"nodeType":1,"localName":"button","namespaceURI":"http://www.w3.org/1999/xhtml","childNodeCount":1,"attributes":{},"shadowRoot":null}}],"resultOwnership":"root","userActivation":true,"serializationOptions":{"maxObjectDepth":0,"maxDomDepth":0}}}
0:06.41 pid:22179 1716817056164 RemoteAgent DEBUG WebDriverBiDiConnection 7afd76bf-0fcd-4335-a43e-55b729ebe1de <- {"type":"success","id":156,"result":{"realm":"c1657cae-c1a6-4ad8-a140-bb45bace99c3","type":"success","result":{"type":"node","handle":"ced784db-2d5f-429c-a087-7e048ce332f5","sharedId":"ea2b5108-5e80-4095-a950-bdf65e7def14","value":{"nodeType":1,"localName":"button","namespaceURI":"http://www.w3.org/1999/xhtml","childNodeCount":1,"attributes":{},"shadowRoot":null}}}}
[..]
0:06.79 pid:22179 1716817056972 RemoteAgent DEBUG WebDriverBiDiConnection 7afd76bf-0fcd-4335-a43e-55b729ebe1de -> {"id":10158,"method":"script.disown","params":{"target":{"context":"0a3ff222-04e6-43df-aec1-8476cd80f724"},"handles":["ced784db-2d5f-429c-a087-7e048ce332f5"]}}
Given that we have trace logs enabled this is causing quite a lot of output. So I've two questions:
- Are that many elements really needed?
- Evaluating nodes in Puppeteer always requires a handle?
- Why cannot all buttons dismissed at once by calling
script.disown
a single time?
If we cannot reduce the output we will have to most likely reduce the log output from trace
to debug
or even info
, or starting to chunk the tests.
Alex, mind to give some feedback? Thanks.
Assignee | ||
Comment 4•6 months ago
|
||
As discussed with Alex on Matrix yesterday the test will stay as is. That means we have to find another solution for us, which probably brings us down to these two:
- Add chunking to the pup (wd) jobs . It may work as long as this test isn't part of the job that has some other test with quite a lot of logging.
- Don't run with trace logs by default - even debug might be too much (we should check). Here we might not have the chance to easily investigate failures
Assignee | ||
Updated•6 months ago
|
Assignee | ||
Comment 5•6 months ago
|
||
As discussed lets reduce the number of elements that test is using so that we don't miss coverage. I'll file a follow-up bug so that we can work on a better fix for logging and at the same time get the log summaries for failure classification working again.
Assignee | ||
Comment 6•6 months ago
|
||
Updated•6 months ago
|
Updated•6 months ago
|
Comment 8•6 months ago
|
||
bugherder |
Description
•