1285765 - The elements that css filter: is applied are not visible when hardware acceleration is enabled

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect, P3)

Description

[eminozlem7777]

Attachment: bugzm.jpg

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:47.0) Gecko/20100101 Firefox/47.0
Build ID: 20160623154057

Steps to reproduce:

Updated Firefox and / or my graphics driver. 
I dont know which one came first, and which one the problem stemmed from but I think it was the Firefox update.


Actual results:

CSS filters stopped working.

I tried updating my graphics drivers and Firefox switched between several versions to no avail. 
Firefox v: 46-47
GPU: Radeon R9 380
gpu software: radeon settings 16.3 , 16.4


Expected results:

CSS filters should be working.

Comment 1

[eminozlem7777]

Forgot to add; i disabled all plugins, refreshed firefox even uninstall and reinstalled.

Comment 2

[Tooru Fujisawa [:arai]]

Can you provide the URL of the page?

Comment 3

[eminozlem7777]

(In reply to Tooru Fujisawa [:arai] from comment #2)
> Can you provide the URL of the page?

http://my.studiopress.com/themes/pretty-creative/#demo-full

or my own site eminozlem.com or any other site that uses filters.

Comment 4

[eminozlem7777]

Needless to say but forgot to mention; the problem goes away when hardware acceleration is disabled.

Comment 5

[Tooru Fujisawa [:arai]]

thanks :)

can you also provide more information about the expected result and actual result?
which part of the webpage do you mean and what's the difference between expected result and actual result?
(which filter do you apply to which element, and how is it displayed with and without the filter?)

also, is there any action should be done to reproduce the issue, like clicking or hovering some element?

Comment 6

[eminozlem7777]

Attachment: info

Comment 7

[eminozlem7777]

(In reply to Tooru Fujisawa [:arai] from comment #5)
> thanks :)
> 
> can you also provide more information about the expected result and actual
> result?
> which part of the webpage do you mean and what's the difference between
> expected result and actual result?
> (which filter do you apply to which element, and how is it displayed with
> and without the filter?)
> 
> also, is there any action should be done to reproduce the issue, like
> clicking or hovering some element?

The elements that css filter: is applied are not visible at all (as if you did visibility:hidden). I think it has something to do with a clash between ATI drivers and firefox, so I dont think you can reproduce it unless you have exact software as i do on this PC unfortunately. PS: this is a brand new PC i put together recently, so I am pretty sure it is not a hardware issue.

Comment 8

[Tooru Fujisawa [:arai]]

(In reply to eminozlem7777 from comment #7)
> The elements that css filter: is applied are not visible at all (as if you
> did visibility:hidden). I think it has something to do with a clash between
> ATI drivers and firefox, so I dont think you can reproduce it unless you
> have exact software as i do on this PC unfortunately. PS: this is a brand
> new PC i put together recently, so I am pretty sure it is not a hardware
> issue.

So, the top half of https://bug1285765.bmoattachments.org/attachment.cgi?id=8769447 is the actual result and
bottom half of the image is expected result, right?

Comment 9

[eminozlem7777]

(In reply to Tooru Fujisawa [:arai] from comment #8)
> (In reply to eminozlem7777 from comment #7)
> > The elements that css filter: is applied are not visible at all (as if you
> > did visibility:hidden). I think it has something to do with a clash between
> > ATI drivers and firefox, so I dont think you can reproduce it unless you
> > have exact software as i do on this PC unfortunately. PS: this is a brand
> > new PC i put together recently, so I am pretty sure it is not a hardware
> > issue.
> 
> So, the top half of
> https://bug1285765.bmoattachments.org/attachment.cgi?id=8769447 is the
> actual result and
> bottom half of the image is expected result, right?

right.

Comment 10

[Loic]

Attachment: Capture d’écran 2016-07-10 à 04.55.19.png

That looks fine on my machine with FF47.

Comment 11

[congxiong]

Hi guys, 

This is Carlos from AMD graphics driver team. I have checked this issue and it happens on all AMD R9 380, Fury and RX 480 ASICs. The root cause is that AMD starts reporting 16 bit precision shader instruction support caps (D3D11_DDI_SHADER_MIN_PRECISION_16_BIT) on above these ASICs; but still doesn't support 10 bit precision shader instructions (D3D11_DDI_SHADER_MIN_PRECISION_10_BIT) yet. However, Firefox browser doesn't check the shader precision support caps and uses 10 bit precision shader instructions directly, which causes the issue for our driver. 

The correct way is that Firefox should query shader precision caps (D3D11_1DDICAPS_SHADER_MIN_PRECISION_SUPPORT) and check if 10 bit precision is supported before using it (the same way as 16 bit precision support, no use it if not supported).

I also have a workaround for this issue from driver side but it needs a couple of weeks to go into the public driver.

BTW, the issue can be observed by visiting https://developer.mozilla.org/en-US/docs/Web/CSS/filter
Scroll down to the opacity filter; notice in the live example pictures, some of them are empty. You can turn off hardware acceleration for comparison to be aware of this issue.

Comment 12

[eminozlem7777]

Thank you for the info, I hope it gets fixed soon, its very annoying. I ve updated to the most recent -16.7.2- Radeon Settings and its still there

Comment 13

[congxiong]

Will let you know the published driver contains the fix. Unfortunately I think you need wait 3~4 weeks. All of the changes need go through lots of test stages including quality and performance.

Comment 14

[eminozlem7777]

I just updated to the newest 16.7.3 driver and to my disappointment, its still there. Isnt there a beta branch where this is fixed ? or cant I install an older version ? Thanks in advance

Comment 15

[congxiong]

I just checked our release plan: the driver with the fix will be available in 16.8.2 which will be published at end of August.

Comment 16

[congxiong]

Sorry for the late reply as company has changed the release plan so the driver isn't published as I mentioned. I confirm the fix should be in 16.9.2 driver.

http://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-Edition-16.9.2-Release-Notes.aspx

Comment 17

[Emma Humphries ☕️🎸🧞‍♀️✨ (she/they) [:emceeaich] (Pacific Time) use needinfo]

Bulk move of gfx-noted bugs without priority to P3 for tracking.

Comment 18

[Gregory Pappas [:gregp]]

Closing per comment 16