trr_request_count_per_conn incorrectly reports http protocol version of 1.1
Categories
(Core :: Networking: DNS, defect, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox143 | --- | fixed |
People
(Reporter: acreskey, Assigned: kershaw)
References
(Blocks 2 open bugs)
Details
(Whiteboard: [necko-triaged][necko-priority-next])
Attachments
(1 file)
In this query we see a very high % of Fenix TRR connections being made over HTTP/1.1.
However, this conflicts with the reporting of other probes -- see https://bugzilla.mozilla.org/show_bug.cgi?id=1976996#c4
I believe there's a bug in how we mark trr_request_count_per_conn as "h1"
https://searchfox.org/mozilla-central/rev/ac81a39dfe0663eb40a34c7c36d89e34be29cb20/netwerk/protocol/http/nsHttpConnection.cpp#68-84
We test if (!mEverUsedSpdy) and I suspect we end up counting http/3 connections.
Reporter | ||
Updated•29 days ago
|
Assignee | ||
Comment 1•26 days ago
|
||
I believe there's a bug in how we mark trr_request_count_per_conn as "h1"
https://searchfox.org/mozilla-central/rev/ac81a39dfe0663eb40a34c7c36d89e34be29cb20/netwerk/protocol/http/nsHttpConnection.cpp#68-84
We test if (!mEverUsedSpdy) and I suspect we end up counting http/3 connections.
hmm... but nsHttpConnection
should only be used for TCP connections.
Maybe the reason is that we increase mHttp1xTransactionCount
without checking the status or those h1 connections were from some user specified trr server. Anyway, I think it'd be good to improve this metric. I'll prepare a patch.
Assignee | ||
Comment 2•26 days ago
|
||
Updated•26 days ago
|
Comment 4•24 days ago
|
||
bugherder |
Description
•