Leakage of CDP connections
Categories
(Remote Protocol :: Agent, defect, P3)
Tracking
(Not tracked)
People
(Reporter: whimboo, Unassigned)
References
Details
(Keywords: memory-leak)
The following log shows that we leak CDP connections:
https://treeherder.mozilla.org/logviewer?job_id=352634137&repo=autoland&lineNumber=48403-48467
The problem is that the Target class holds the connections, but only the destructor actually closes them all.
If a client closes the connection itself, the reference to the connection will still remain. As such we should inform consumers of the class by emitting a closed
event that the given connection is no longer existent.
The same should actually also happen with the WebDriver BiDi connections given that they have been designed identically.
The actual memory leak shouldn't be that dramatic, but it would be good to get fixed anyway. Also it would improve the investigation of test failures, which would have less cluttered log files.
Reporter | ||
Updated•3 years ago
|
Comment 1•3 years ago
|
||
The severity field is not set for this bug.
:whimboo, could you have a look please?
For more information, please visit auto_nag documentation.
Reporter | ||
Updated•3 years ago
|
Updated•2 years ago
|
Reporter | ||
Comment 2•2 years ago
|
||
I think that the WebDriver BiDi specific issue has been fixed by bug 1742359. As such this should only apply to CDP now.
Currently CDPConnection.onclosed
calls session.destructor()
for each registered session. But Session.destructor
which owns the target, doesn't call the Target.destructor
.
That said it's not that important and should not block any WebDriver BiDi milestone.
Reporter | ||
Comment 3•2 years ago
|
||
I quickly tested that but it breaks quite a lot of tests. So it might be more complicated to do that correctly.
Description
•