APZ_ZOOM_ACTIVITY is wrongly incremented in about:telemetry
Categories
(Core :: Panning and Zooming, defect, P3)
Tracking
()
People
(Reporter: sbadau, Assigned: kats)
References
(Blocks 1 open bug)
Details
Attachments
(2 files)
Affected versions
Firefox Nightly 84.0a1
Affected platforms
Mac OS X 10.15
Windows 10
Preconditions
fx.webrender.all = true OR fx.webrender.all = false
apz.allow_zooming = true
apz.windows.use_direct_manipulation = true
Steps to reproduce
- Open Firefox with a new profile
- Navigate to any website of your choice and pinch-zoom
- Go to about:telemetry and look for APZ_ZOOM_ACTIVITY
Expected result
The value of APZ_ZOOM_ACTIVITY should be incremented with 1 each time pinch zooming is done on a new website.
Actual Result
The value of APZ_ZOOM_ACTIVITY is incremented usually with 2 (seen ever higher, like 5 or 10). APZ_ZOOM_ACTIVITY is incremented also when opening a new website, even without pinch-zooming (this behavior is encountered also for APZ_ZOOM_ACTIVITY_RDM, it's enough to open RDM mode for the value to be incremented)
Reproduced this issue also on Firefox 83 beta 6.
Severity-Suggestion
S2
|
Assignee | |
Comment 1•4 years ago
|
||
Given this probe is going to expire in 85 (bug 1672576) and it seems unreliable I'm inclined to just remove it. Doesn't seem like we've used this data for anything in particular.
|
||
Comment 2•4 years ago
|
||
@Martin: have you or are you plannin to use this probe?
|
||
Comment 3•4 years ago
|
||
I planned to use the probe to track the adoption of the feature. I would appreciate if we could fix the counting, if possible. Over counting is generally easier for me to deal with than under counting.
It would be important to me to set APZ_ZOOM_ACTIVITY and APZ_ZOOM_PINCHSOURCE to not expire. I'll comment on bug 1672576.
|
|
Assignee | |
Updated•4 years ago
|
|
|
Assignee | |
Comment 4•4 years ago
|
||
I added a print statement with the document URL to see what was happening here. Mostly the telemetry is correct, but on browser startup there's a bunch of spurious about:blank presshells that get loaded and unloaded, so there's ~5 telemetry entries that don't really capture user intent there. It seems hard to distinguish that from other loads of about:blank (e.g. if I go to about:blank by typing that into the URL bar manually) but I can try. There's also telemetry entries from about:newtab. Both about:blank and about:newtab can actually be zoomed so I think it makes sense to keep recording zoom activity on those URLs (as opposed to, e.g. stripping out presShells with those URLs or any about: URL).
|
|
Assignee | |
Comment 5•4 years ago
|
||
The spurious about:blank loads seem to be coming from the extension-type process, as opposed to regular about:blank which gets loaded in a privilegedabout-type process. So I guess I can record stuff from web and privilegedabout process and ignore all the rest.
|
|
Assignee | |
Updated•4 years ago
|
|
|
Assignee | |
Comment 6•4 years ago
|
||
There's also some about:blank page loads in the web process, but usually just one per process it seems like, so I think it's probably safe to ignore.
|
|
Assignee | |
Comment 7•4 years ago
|
||
|
|
Assignee | |
Comment 8•4 years ago
|
||
Depends on D96893
|
||
Updated•4 years ago
|
|
|
||
Updated•4 years ago
|
Pushed by kgupta@mozilla.staktrace.com: https://hg.mozilla.org/integration/autoland/rev/414f92a9b18d Fix typo in function name. r=nika https://hg.mozilla.org/integration/autoland/rev/cec7a6eee54f Further restrict where we record APZ_ZOOM_ACTIVITY. r=nika
|
||
Comment 10•4 years ago
|
||
| bugherder | ||
https://hg.mozilla.org/mozilla-central/rev/414f92a9b18d
https://hg.mozilla.org/mozilla-central/rev/cec7a6eee54f
|
||
Updated•4 years ago
|
|
||
Comment 11•4 years ago
|
||
The patch landed in nightly and beta is affected.
:kats, is this bug important enough to require an uplift?
If not please set status_beta to wontfix.
For more information, please visit auto_nag documentation.
|
|
Assignee | |
Comment 12•4 years ago
|
||
It's probably not worth uplifting. It's kind of nice that the patch landed right at the start of the 85 cycle, which means we know that all data from 85 is corrected and data from older versions is not. If we uplift then data from the first few days of 84 beta will not be comparable to later in 84 beta.
Description
•