Open Bug 1714451 Opened 4 years ago Updated 4 years ago

Firefox 89 Eye Strain

Categories

(Core :: Graphics, defect)

Product:
Core
Component:
Graphics
Version:
Firefox 89
Type:
defect

Tracking

()

Status:
UNCONFIRMED

People

(Reporter: j.pospis, Unassigned, NeedInfo)

Details

Attachments

(1 file)

FF89-issue-support-info.txt
22.69 KB, text/plain
Details

User Agent: Mozilla/5.0 (Windows NT 6.2; Win64; x64; rv:88.0) Gecko/20100101 Firefox/88.0

Steps to reproduce:

Since Firefox 89 I am experiencing eye strain with Firefox similar to what Chrome 69 and above started to cause. Firefox was a browser very light on eyes and now it is similar to Chrome. It is probably the new color palette, not sure if it is using some form of temporal dithering? Is there a way to switch back to the old palette to test? I am falling back to FF ESR for now, which I can confirm is not causing the eye strain. I personally believe this issue may impact a lot of people (millions+) who will never report this as they will expect their eye strain issues are coming from their monitors, lighting, fatigue etc.

Actual results:

Eyes strain, picture looks "unstable" like using temporal dithering

Expected results:

No eye strain like with FF 88.

I am not sure whether the classification of "theme" is the correct one. This is not about how FF overall looks, it is about how its engine displays colors overall and especially on web pages.

Can you attach the contents of about:support?
I'll move to the Graphics component per comment #1.

Component: Theme → Graphics
Flags: needinfo?(j.pospis)
Product: Firefox → Core
Flags: needinfo?(j.pospis)

about:support info added from one of my laptops, however not sure to what extent it is relevant, this issue is very general with FF89, I see it on all PCs/laptops and OSes I have including Windows 7, Windows 10 and Linux.

Given what reply I received on the Firefox support channel, obviously I am not alone who sees this as a problem and people are reinstalling FF 88.0.1 to solve it.

To add to this: some may not call it a "bug" per se. However it is very limiting Firefox's usability for sensitive people - if you encounter eye strain shortly after using Firefox, the browser is very hard to use indeed. As I state earlier I believe this issue could be impacting a lot of people who only have symptoms after extended periods of time while working with FF. Chrome has a similar issue, this is why I've been using Firefox almost exclusively in the past years.
My best bet is that FF 89 was made to use a wider color range (yes, it is obvious that colors are more vibrant now) however it comes at a cost: new colors cannot be created out of nothing and it is using some kind of temporal dithering or a similar technology. This severely impacts people whose eyesight is sensitive to motion - yes, I am even getting motion sickness out of the new FF if I use it for more than half an hour. There are moments when I can even "sense" the "instability" of FF 89's picture.
I'd like to kindly ask the developers to provide an option within Firefox to revert back to the original color palette and the way colors are displayed. At the same time I'd like to kindly ask if someone can comment on the nature of the changes that were made with FF 89 as other software was impacted by a similar effect after updates in the past and I'd like to finally understand if it indeed is temporal dithering or something else.
It is again important to emphasize this is NOT about the new Proton look of Firefox - it is about how colors are displayed both in the browser UI and especially on web pages themselves.

Jeff, do you know if any recent colour management changes could have caused this?

jpos, thank you for the bug report. How easily are you able to detect if a version of firefox is affected by this? We have a tool called mozregression that can be used to test a series of versions of firefox. For each one you can answer "good" or "bad", and eventually it will determine what change caused the issue. If it's quick to determine whether Firefox is affected by the issue, perhaps you could try to use that?

  • You can install it from the website I linked there, and then open the application, click the "scissor" icon to "run a new bisection".
  • Click "next" for the first couple of screens ("Basic configuration" and "Profile selection")
  • On the build selection screen change the "date" dropdown menus to "release".
  • Select 88 for "Last known good build", and 89 for "First known bad build"
  • Then click "Finish" and it should start downloading versions of firefox. Select "good" or "bad" for each one.

However, if it takes you half an hour and causes lots of discomfort to determine whether a version is good or bad then please don't put yourself through that!

Flags: needinfo?(jmuizelaar)
Flags: needinfo?(j.pospis)

I can't think of any color related changes that would've caused this. WebRender has been enabled for a while on this hardware so it's unlikely to be that either. A regression window from mozregression would be really helpful in determining what change caused this.

Flags: needinfo?(jmuizelaar)

Jamie, I will certainly try to do that. Per what I observe now when I compare the 78 ESR/88.0.1 and 89 I should be able to say rather easily which version is impacted by this and therefore where it all started, but given my busy schedule now it may take several days before I get to it.

Jeff, the official release highlights of 89 state:
Cohesive, calmer visuals: Lighter iconography, a refined color palette, and more consistent styling throughout

When I noticed something was wrong I was searching for the info on what had been changed and my impression was it could be hidden under this (a refined color palette). But perhaps Mozilla is talking about the UI colors here?

Yes, the "refined color palette" is referring to the colours throughout the new UI theme.

But that's not to say something didn't accidentally change in the way we render web pages that is causing this issue, so when you get a chance to run mozregression that will be really helpful!

Severity: -- → S3

I've tried the first run of mozregression, 88 -> 89, I've narrowed it down to an event between April 5 and April 7, but I must say it's harder than I thought as after staring at the screen version after version my eyes get tired easily and I get slight nausea. This issue has some form of a "memory effect" on my eyesight as looking at a "bad" version for some time makes me feel like even the 88.0.1 is suddenly bad too and it takes a while to "reset" my eyesight and realize it is OK. In any case I am quite confident I am not imagining things and this issue is indeed real.

If I understand it correctly for the next iterations I can choose dates (instead of releases), good April 5, bad April 8 (to be on the safe side). Frankly I did complete the first run till the end but I am not very confident I found the right commit, I want to be sure before I let you guys look at it further.

I looked through some of the logs and nothing jumped out at me. Is it possible that you can see the effect with a screenshot or does the content need to be in motion?

Certainly must be in motion. It is hard to describe it, Jeff, it's very subtle. If you remember the CRT days the feeling is similar to using a lower Hz monitor compared to a 100 Hz one. Back in the day I was able to detect up to 85 Hz, that's probably why I am more sensitive to this. FF 88 is like looking at a 100 Hz monitor and FF 89 is something like 75-85 Hz... it's OK at first, the picture is the same... but there is something different and something that does not feel right and causes eye strain and motion sickness.

As I write above, Chrome has exactly the same issue, although even FF 89 is NOT that bad as the current Chrome is. I found out about this by the end of last year when I first used Chrome after several years. Later I discovered that it had all started with Chrome 69 (and I am not alone who has the same feeling) where with Chrome 69 they were boasting that they now had better colors.

Are we sure that the latest mozregression (4.0.17) for Windows is clean? Firefox itself is warning me that it contains a virus inside during the download! Although my antivirus seems to be OK with it then... Just weird that Mozilla's own product is having trouble with it :)

Previously I used the linux version but wanted to give the Windows platform a try for the next run...

Thanks!

Yeah, it should be. It's just not signed so that might be what's causing the problem.

My sincere apologies for not updating the bug report for a long time. Several last attempts to figure out where it started exactly caused me severe eye strain and then I was very busy with my work and could not find the right time to give it another try. In the meantime I tried the latest 90.x FF and I am very positive there is something wrong there. Returning back to ESR fixes the eye issues for me. I will report back as soon as I get the right version with which it was introduced.

Additionally - the mozregression runs I tried more than a month ago pointed to something between April 5 and 7. If there were no changes to the renderer during that time, are you sure that something external did not change? I mean - if there is some external software you use inside FF and you suddenly could have changed the version of it, like Skia or something similar...

You should be able to restart a mozregression directly from the range April 5 to April 7. That should cut down the number of revision that you need to try. Any narrowing down of the range will be very helpful. Mozregression prints a link to the current range as it narrows down. So you could always post those as you produce them.

You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: