Starting with Firefox 37 there is more of a delay with speech when typing on the Web with typing echo enabled

[nvaccessAuto]

Reported by dgoldfield on 2015-04-13 08:27
When using previous versions of Firefox, typing text on the Web, such as using this particular form to file this ticket, produced instant feedback when typing mode was enabled. The response was nearly instantaneous, as it is with standard word processing software. Starting with Firefox 37.0.1 I am noticing more of a delay between a keypress and when speech reads the key which was typed.

[nvaccessAuto]

Comment 1 by jteh on 2015-04-13 23:11
Marco, have you heard any reports of major performance regressions with respect to accessibility with Firefox 37? Either way, do you know what might be causing this? I've been running nightlies as my primary browser for years now and I never saw this, and I'm pretty sure I did some testing with Firefox 37 when it was in beta a while ago. Regardless, this (along with #5032) is the second report we've had of this. The first report was via email and is even more concerning. I've included it below:

I've been a Firefox and NVDA user for just over 5 years, and have found it to be the most stable combination to use. Several days ago, however, I decided to upgrade to the recently-released version of Firefox (37.0.1), which I can unfortunately say without a doubt has been the worst version of the browser that I've ever used. The main issue is down to lag; as I'm sure you know, navigation is generally quick but this update appears to have changed that, resulting in delays of up to 3 or more seconds when multiple tabs are opened.

Thoughts appreciated.

[nvaccessAuto]

Comment 3 by MarcoZehe on 2015-04-14 11:26
Hi jamie, this is the first I've heard of it, and I cannot reproduce with Firefox 37 release or the nightly build that is on level 40. I in fact have several tabs open as I write this, and there is no delay whatsoever. I am also using a braille display. We definitely need more info on the system specifics etc.

[nvaccessAuto]

Comment 4 by elliott94 (in reply to comment 3) on 2015-04-17 15:47
Replying to MarcoZehe:

We definitely need more info on the system specifics etc.

Hello, thank you for taking the time to look into this.

Upon carrying out several more tests, my initial thought was that this particular issue was related to hardware; version 37.0.1 runs much better, for example, on one of my test machines with a Core 2 Duo T9300 (2.5GHz) than it does on a machine with a Pentium E5400 (2.7GHz). However, on both machines performance is made 100 times worse when upgrading to 37.0.1 from 36.0.4, so much so that on certain websites the tab becomes so unresponsive that it must be closed to continue normal browsing. Unfortunately as of yet I have no other ideas of use; both machines that I tested on have significant amounts of memory (2/3 GB respectively), and are both running Windows 7 Home Premium X64. In addition, I don't think that NVDA is to blame at all here; I tried several older releases (including 2015.1), all of which produced the same result. All do, however, work without issue if downgrading to version 36.0.4 of Firefox, but this obviously can't be recommended as a longterm solution due to security fixes that are routinely included in updates.

If nothing else, I hope that helps clarify that the problem isn't related to low-specification hardware; @dgoldfield, if it's not too much trouble would it be possible to get an overview of your setup also? I should finally add that I'm not using a braille display and that the issue isn't just limited to typing in text fields, but also when navigating in Browse Mode. I'd be extremely grateful if anybody else could shed some light on this - the latest iteration is literally unusable in its current state.

[nvaccessAuto]

Comment 5 by dgoldfield on 2015-04-17 18:19
I will provide you with system specs generated from the System Information app. This is the system where I am noticing a slight delay in typing but it is not awful or problematic, just a bit slower than FF 36 and prior. Interestingly, a similarly configured machine at work seems to have no performance issues. I will provide specs for that machine once I'm able to do so. For now, here are the specs for the machine where I am noticing the slight delay while typing, especially when I type quickly. If you need more info, please let me know.

OS Name Microsoft Windows 7 Professional
Version 6.1.7601 Service Pack 1 Build 7601

System Manufacturer Dell Inc.
System Model OptiPlex 740 Enhanced
System Type X86-based PC
Processor AMD Athlon(tm) 64 X2 Dual Core Processor 5600+, 2900 Mhz, 2 Core(s), 2 Logical Processor(s)
BIOS Version/Date Dell Inc. 2.2.7, 8/3/2011
SMBIOS Version 2.4
Windows Directory C:\Windows
System Directory C:\Windows\system32
Boot Device \Device\HarddiskVolume1
Locale United States
Hardware Abstraction Layer Version = "6.1.7601.17514"

Time Zone Eastern Daylight Time
Installed Physical Memory (RAM) 3.00 GB
Total Physical Memory 2.94 GB
Available Physical Memory 1.81 GB
Total Virtual Memory 5.87 GB
Available Virtual Memory 4.66 GB
Page File Space 2.94 GB
Page File C:\pagefile.sys

[nvaccessAuto]

Comment 6 by dgoldfield on 2015-04-17 18:41
I have two additional questions.

1. Since FF 37.0.1, virtual cursor navigation is noticeably slow while playing a Youtube video using the HTML5 version of Youtube. Is this likely relevant to this ticket or should another ticket be created for this issue?
2. In light of a previous comment from @elliott94 is it worth removing FF 37.0.1 along with all profiles and performing a fresh install?

[nvaccessAuto]

Comment 7 by elliott94 (in reply to comment 6) on 2015-04-17 19:24
Hello.

Thank you for taking the time to post your system information - I'm unfortunately not a developer myself, but hopefully it'll help in some way. As I speculated in comment:4, I believe that the issue is not caused by lack of memory. See below for more comments.

Replying to dgoldfield:

I have two additional questions.

1. Since FF 37.0.1, virtual cursor navigation is noticeably slow while playing a Youtube video using the HTML5 version of Youtube. Is this likely relevant to this ticket or should another ticket be created for this issue?

As mentioned above I'm in no way a developer, but I think that these issues are likely caused by the upgrade as a whole rather than individual points that need to be addressed. I could be totally wrong on this, however.

1. In light of a previous comment from @elliott94 is it worth removing FF 37.0.1 along with all profiles and performing a fresh install?

Instead of removing all profiles and starting off with a fresh copy of the browser, I have a suggestion that you could try beforehand which would support the theory that these issues are caused by the 37.0.1 update; running a portable copy of the browser should give you a good idea (see below for links). I should point out that I'm in no way affiliated with the PortableApps project, but I've used their re-packaged software many times in the past without issue and they're without a doubt a trusted source.

If this is something that you want to try, you can download the portable version of 36.0.4 from the following link (externally hosted at SourceForge):

http://sourceforge.net/projects/portableapps/files/Mozilla%20Firefox%2C%20Portable%20Ed./Mozilla%20Firefox%2C%20Portable%20Edition%2036.0.4/FirefoxPortable_36.0.4_English.paf.exe/download

Simply choose to either open or save the above file, and when prompted by the resulting installer choose a location in which you want to extract the files. Obviously since this is a portable app and no information is written to the registry or anything similar you can safely remove the entire directory when finished (although make sure during installation that you choose a location that is not a part of your existing installed copy). Finally, launch the FirefoxPortable exicutable file in the root of the extracted directory and observe whether there is less lag than with 37.0.1.

If you do choose to test this, would you be able to update the ticket with your findings?

Thanks.

[nvaccessAuto]

Comment 8 by elliott94 on 2015-04-21 20:20
I was really hoping that the recently-released 37.0.2 update of the browser would fix the issues described above, but after several minutes of testing this unfortunately doesn't appear to be the case. Has nobody else experienced anything similar?

[nvaccessAuto]

Comment 9 by dgoldfield on 2015-04-25 06:35
I have discovered an interesting solution to this issue which seems to have improved the overall responsiveness of Firefox. In the options dialog in the Advanced category in the "general" tab I unchecked "use hardware acceleration" and things seem to be back to normal. First, keyboard echo while typing text seems to be back to normal. Arrow navigation during a Youtube video also seems to be back to normal. I am going to do further testing to see if this solves the issues with bug 2039. I am not certain why this needs to be done as this setting was always enabled using FF 36 and earlier. However, at least it's a workaround which has improved responsiveness and so far has had no negative side effects.

[nvaccessAuto]

Comment 10 by elliott94 (in reply to comment 9) on 2015-04-25 08:11
Replying to dgoldfield:

I have discovered an interesting solution to this issue which seems to have improved the overall responsiveness of Firefox. In the options dialog in the Advanced category in the "general" tab I unchecked "use hardware acceleration" and things seem to be back to normal. First, keyboard echo while typing text seems to be back to normal. Arrow navigation during a Youtube video also seems to be back to normal.

Nice find! I can confirm also that disabling the same setting appears to have restored responsiveness to how it was before the 37 update; thank you for taking the time to test. Did you by any chance try the portable version of 36.0.4 on your system to see if it was definitely the 37.0.1 update that caused the issue?

@marco, do you have any ideas as to if any changes were made to the hardware acceleration code between 36.0.4 and 37? Obviously whilst this solution appears to restore the browser to a useable state, my worry is that new users who download/upgrade will also experience similar issues. My guess, however, is that sluggishness would occur whether or not the user had any assistive software active and that it's a browser issue. It'd be interesting to know why this actually happens.

[nvaccessAuto]

Comment 11 by jteh on 2015-04-25 09:46
I was starting to have a hunch this might be something to do with hardware acceleration. We've certainly seen regressions like this before. I also seem to recall I had to tweak advanced settings related to hardware acceleration to stop Thunderbird from being ridiculously sluggish on my system, but this was quite a while ago and I can't remember specifics.

Interestingly, I have hardware acceleration enabled in Firefox on my system and am not seeing this, so I suspect it's some weird and wonderful combination of things that triggers this.

[nvaccessAuto]

Comment 12 by elliott94 (in reply to comment 11) on 2015-04-25 10:44
Replying to jteh:

Interestingly, I have hardware acceleration enabled in Firefox on my system and am not seeing this, so I suspect it's some weird and wonderful combination of things that triggers this.

Interesting. I've got the same sluggishness on two machines that I've been able to test with, both now running 37.0.2 of Firefox. Each machine has drastically different hardware configurations, although from comment:5 it would appear that this occurs both on AMD and Intel CPUs, so I guess the next suspect could be down to the graphics hardware in use. I've experienced this both on a system utilizing both built-in and NVIDIA setups, so am unfortunately out of ideas for the time being.

[nvaccessAuto]

Comment 13 by MarcoZehe (in reply to comment 10) on 2015-04-27 06:27
Replying to elliott94:

@marco, do you have any ideas as to if any changes were made to the hardware acceleration code between 36.0.4 and 37?

I, myself, am not aware of any changes, and since I've never seen the bug and never heard of it until this ticket was filed, there were no flags raised. Furthermore, Hardware Graphics is a different team. I've therefore asked them in this Mozilla bug what could be happening.

[nvaccessAuto]

Comment 14 by elliott94 (in reply to comment 13) on 2015-04-27 14:00
Replying to MarcoZehe:

I, myself, am not aware of any changes, and since I've never seen the bug and never heard of it until this ticket was filed, there were no flags raised. Furthermore, Hardware Graphics is a different team. I've therefore asked them in this Mozilla bug what could be happening.

Hello.

Thank you for this - I'll definitely keep an eye on the Bugzilla entry to see if anybody has any suggestions. For now, at least, applying the fix mentioned in comment:9 on both of my test machines seems to have done the trick, but as previously mentioned in comment:10 my main concern is for users who are either new to the browser and arn't expecting slow performance, or users who upgrade and wonder why the browser suddenly appears to be a lot slower than it was before.

[nvaccessAuto]

Comment 15 by MarcoZehe on 2015-04-27 14:44
OK, a colleague came back with a first try at finding what's wrong. Can I ask those of you who experience the problem to do the following:

1. Re-enable the Hardware Acceleration checkbox in Tools/Options/Advanced/General Tab.
2. Confirm that you once again see the keyboard echo lag.
3. In the address bar, type about:config.
4. Okay the warning with "I'll be careful, I promise!".
5. Find the preference layers.d3d11.disable-warp. You should be able to type in warp into the search field, tab forward once and find it in the list.
6. Press the Context Menu key and choose the Toggle option.
7. Restart Firefox.
8. See if you still see a keyboard echo lag now.
9. Report back here please.

[nvaccessAuto]

Comment 16 by dgoldfield on 2015-04-28 09:22
I can confirm that this suggestion made a difference. The keyboard lag while typing had returned when I enabled hardware acceleration, although the lag I am noticing is not as significant as what others seem to experience. I then toggled the setting you mentioned in your previous comment, restarted Firefox and typing responsiveness was back to normal. Also, navigation during HTML5 Youtube videos seems to have returned back to normal as well.

[nvaccessAuto]

Comment 17 by elliott94 on 2015-04-28 09:48
I can confirm the exact same results as comment:16; after re-enabling Hardware Acceleration in Options/Advanced and disabling the preference in about:config, performance returns back to normal.

[nvaccessAuto]

Comment 18 by MarcoZehe on 2015-05-25 14:05
A request to all of you who are having this problem:

1. Open a new tab, and in the address bar type about:support
2. Locate the Graphics section of that new page and copy and paste all information contained in there, either into this ticket, or directly into the Mozilla bug quoted above if you have a Bugzilla account. Our engineers are asking for this information to nail down what is causing the problems with WARP.

Thanks for your help!

[nvaccessAuto]

Comment 19 by dgoldfield on 2015-05-25 15:11
Here is the information contained in the graphics section in the about:support page. I am currently not having this problem if I either disable the warp settings in about:config mentioned earlier or if I turn hardware acceleration off.

Graphics
Adapter Description
NVIDIA GeForce 6150 LE
Adapter Drivers
nvd3dum
Adapter RAM
64
Device ID
0x0241
DirectWrite Enabled
false (6.2.9200.17292)
Driver Date
1-31-2013
Driver Version
9.18.13.783
GPU #2 Active
false
GPU Accelerated Windows
0/1 Basic (OMTC) Blocked for your graphics card because of unresolved driver issues.
Subsys ID
01ec1028
Vendor ID
0x10de
WebGL Renderer
Google Inc. -- ANGLE (NVIDIA GeForce 6150 LE Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote
true
AzureCanvasBackend
skia
AzureContentBackend
cairo
AzureFallbackCanvasBackend
cairo
AzureSkiaAccelerated
0

[nvaccessAuto]

Comment 20 by bhavyashah on 2015-05-25 17:18
Hi,
Just to add to the existence of this bug, an Italian NVDA user named Mallard also confirmed this same behaviour and also confirmed that toggling hardware acceleration resolved these issues.
Thanks.

[nvaccessAuto]

Comment 21 by MarcoZehe (in reply to comment 19) on 2015-05-28 17:43
Hi David,

Replying to dgoldfield:

Here is the information contained in the graphics section in the about:support page. I am currently not having this problem if I either disable the warp settings in about:config mentioned earlier or if I turn hardware acceleration off.

I have a big favour to ask, passed on from our engineer currently looking into this problem: He requests that Warp or Hardware Accelleration be enabled again, and the graphics info from about:support collected with this enabled. Apparently the info mirrors what's actually there, so the important info he needs unfortunately wasn't there in the info you provided.

So if you could temporarily enable WARP or hardware accelleration, depending on which you disabled, confirm the problem reappeared, and then paste the graphics info about:support shows, that would be absolutely awesome!

[nvaccessAuto]

Comment 22 by MarcoZehe on 2015-07-14 07:01
David, did you have any chance to try what I asked in comment #21 above? E. G. get the support information with hardware acceeration enabled and the WARP thing also turned back on to its default?

Another question to all of you seeing this: Has anybody seen this on Windows versions other than 7? For example Windows 8.1? In all above comments, it only mentions Windows 7, so am wondering. And colleagues are asking that in the related Mozilla bug.

Thanks!

[jcsteh]

Mozilla believes this to be fixed, so closing. Please comment if the problem persists.