Bad performance degradation with NVDA under Windows XP in FF 32

RESOLVED FIXED in Firefox 33

Status

()

RESOLVED FIXED
5 years ago
4 years ago

People

(Reporter: MarcoZ, Unassigned)

Tracking

({perf, regression})

32 Branch
mozilla33
x86_64
Windows XP
perf, regression
Points:
---

Firefox Tracking Flags

(firefox32+ wontfix, firefox33+ fixed, firefox34+ fixed, relnote-firefox 32+)

Details

(Reporter)

Description

5 years ago
I cannot confirm this bug myself, and neither can Jamie. This problem seems to be related to Windows XP specifically. The related NVDA ticket is here:
http://community.nvda-project.org/ticket/4337

Related e-mail thread from the NVDA dev mailinglist:
http://sourceforge.net/p/nvda/lists/message/32651524/

The obvious candidate, unnoticed fall-out from bug 1014673, has been ruled out, as can be seen at the end of the thread. In fact, turning OMTC off seems to make things even worse for the user.

No idea what to suggest or how to proceed, I no longer have XP at my disposal, and it has been EOL'd in April anyway.
(Reporter)

Comment 1

5 years ago
OK, after some more discussion on the e-mail list, it turns out that we set the OMTC to false on XP by default. Turns out that setting it to TRUE improved things for the user drastically on at least two of his machines.

Why was the decision made to set this to FALSE for XP users? And can we still make a flip here in 32 so screen reader users on XP don't run into this issue when they update?
(Reporter)

Comment 2

5 years ago
Further investigation of the user shows that setting OMTC to enabled on his XP machines improved the situation on all of them.
(Reporter)

Comment 3

5 years ago
[Tracking Requested - why for this release]:

[Tracking Requested - why for this release]:

[Tracking Requested - why for this release]:

Confirming bug, since it is clear that we have a problem with OMTC set to False on XP.
Status: UNCONFIRMED → NEW
status-firefox32: --- → affected
status-firefox33: --- → affected
status-firefox34: --- → affected
tracking-firefox32: --- → ?
tracking-firefox33: --- → ?
tracking-firefox34: --- → ?
Ever confirmed: true
Can we get somebody cc'ed who knows impl of Windows part? That can be somehow related with messaging, right?
Hi Benoit, I forget (or never knew) why is OMTC not enabled on XP by default?
Flags: needinfo?(bgirard)
OMTC should be on by default on central, aurora for all windows configurations.
Flags: needinfo?(bgirard)
(Reporter)

Comment 7

5 years ago
And what about Beta? This is in 32, the first version that has OMTC, and the user consistently found that it was *off* by default for him.
Flags: needinfo?(bgirard)
(In reply to Benoit Girard (:BenWa) from comment #6)
> OMTC should be on by default on central, aurora for all windows
> configurations.

Hmmm... do we maybe turn off based on gfx driver detection or something?
OMTC -should- be off for beta. It was on briefly on 32 but was turned off when 32 became aurora. The user seeing it off on 32 beta is expected. Perhaps I'm misunderstanding the question?

Can we get some performance profile and see where the time is being spent?
Flags: needinfo?(bgirard)
Performance regression, tracking.
tracking-firefox32: ? → +
tracking-firefox33: ? → +
tracking-firefox34: ? → +
Marco - We need an option other than enabling OMTC if we are going to see a fix in 32. Can you drive this one?
Flags: needinfo?(marco.zehe)
(Reporter)

Comment 12

5 years ago
There is no real other solution. After talking to Jamie from the NVDA team: I'd suggest to relnote it and tell users to turn OMTC on if they run into this problem on XP. It's only for this release anyway, so other solutions would cause a much higher drain on resources than is warranted.
Flags: needinfo?(marco.zehe)
Release Note Request (optional, but appreciated)
[Why is this notable]:
[Suggested wording]:
[Links (documentation, blog post, etc)]:

(In reply to Marco Zehe (:MarcoZ) from comment #12)
> There is no real other solution. After talking to Jamie from the NVDA team:
> I'd suggest to relnote it and tell users to turn OMTC on if they run into
> this problem on XP. It's only for this release anyway, so other solutions
> would cause a much higher drain on resources than is warranted.

AFAIK, we should not suggest that anyone enable OMTC on 32 as it causes known stability problems (top crashes), which are likely worse than bad performance. I agree that we should consider adding a relnote for this issue.

Marco - Can you please help me write the relnote? Specifically, I'm unclear how the user will notice the performance degradation. Given that this is in the disability access API component, is the issue limited to one or more ally tools?

I'm going to flag this as won't fix in 32 and fixed in 33+ - but this assumes that OMTC sticks on 33.
Status: NEW → RESOLVED
Last Resolved: 5 years ago
status-firefox32: affected → wontfix
status-firefox33: affected → fixed
status-firefox34: affected → fixed
relnote-firefox: --- → ?
Flags: needinfo?(marco.zehe)
Resolution: --- → FIXED
(Reporter)

Comment 14

5 years ago
This will be noticed by users using assistive technologies on Windows XP, such as screen readers for the blind. The results will be slow page loads, slowed down reactions to key presses and focus changes, and over-all sluggishness of the browser.
Flags: needinfo?(marco.zehe)
Target Milestone: --- → mozilla33
I have added the following release note for Firefox 32 "Assistive technologies may cause performance issues on Windows XP"
relnote-firefox: ? → 32+
You need to log in before you can comment on or make changes to this bug.