Open Bug 1687584 Opened 5 years ago Updated 5 years ago

Uptick in A11Y_INSTANTIATED_FLAG during Firefox 80

Categories

(Core :: Disability Access APIs, defect)

Unspecified
Windows
defect

Tracking

()

People

(Reporter: mstange, Assigned: m_kato)

References

(Regression)

Details

(Keywords: regression)

The telemetry evolution graph for A11Y_INSTANTIATED_FLAG shows an uptick from ~11% to ~13% between the build IDs 20200715 and 20200717.
Graph: https://mzl.la/2M7cJTk

It's a small change, but the rest of the graph is extremely flat so it definitely stands out.
Having accessibility enabled causes a large performance overhead (and parent process freezes), so we should try hard to minimize the number of users that are affected.

Here's the pushlog range for those build IDs.

The only change that stands out to me in that list is https://hg.mozilla.org/mozilla-central/rev/8636b6b1f95f2f34548d725254dd0d7fe3516791 , which touched something related to touch screens.

Makoto, is it possible that your patch in bug 1648534 could have caused this?
It seems like it was supposed to just be a regression fix, but there was no corresponding decrease in the numbers when bug 1618759 landed.

Flags: needinfo?(m_kato)

In the meantime, there's work happening in bug 1687535 to stop the mere presence of a touch screen from instantiating accessibility. So if that ends up happening, the "extra accessibility population" from this regression may also be fixed.

Hmm, IME will access a11y module, so I guess that this data is incremented. Now, I am considering that I re-implement software keyboard code via bug 1670539. Although I have a prototype for it, it won't access IME directly...

Assignee: nobody → m_kato
Flags: needinfo?(m_kato)
Type: defect → task
Type: task → defect
Severity: -- → S3
Depends on: 1681800
Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.