519915 - Firefox does not honor the monitor profile (try Microsoft test suit)

Status: NEW

(Core :: Graphics: Color Management, defect)

Description

[polet82]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.1.3) Gecko/20090824 Firefox/3.5.3
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.1.3) Gecko/20090824 Firefox/3.5.3

Firefox has a problem with Microsoft CMS test image and profile (it is attached to the following page):
http://blogs.msdn.com/color_blog/archive/2006/09/29/Profile-utilization-test-image-and-profile.aspx

The bike on the test image is blue. It's embedded profile swaps B and G channels, and FF does show it green. Now, set the same profile as a monitor profile. It should flip B and G one more time and the bike have to be blue again. It does not work, the bike is still green. Same with any image with sRGB profile, this BGR monitor profile is not honored.

Reproducible: Always

Steps to Reproduce:
1. Download test suit from the link above.
2. Set BGR profile for the monitor.
3. Open the bike photo in both FF and Safari.
Actual Results:  
In Firefox bike is green. In Safari bike is blue (because of double conversion, not absence of CMS support).

Expected Results:  
In Firefox bike should be blue.

Same problem on Windows 7 and Mac OSX 10.5.8.
Safari and Photoshop do this double conversion properly and shows blue bike.

Comment 1

[Jeff Muizelaar [:jrmuizel]]

This happens because qcms detects the monitor profile as bogus and chooses not to use it. Chris, do you think our behavior should be different here?

Comment 2

[polet82]

Jeff, 
my point here is that qcms does apply the BGR profile ones (embedded in the picture) but not the second time (i.e. the monitor profile).
I'd expect it to either discard BGR profile both times, or apply it twice (in both cases the bike'd be blue). Not the half way (i.e. green).

Comment 3

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #2)
> Jeff, 
> my point here is that qcms does apply the BGR profile ones (embedded in the
> picture) but not the second time (i.e. the monitor profile).
> I'd expect it to either discard BGR profile both times, or apply it twice (in
> both cases the bike'd be blue). Not the half way (i.e. green).

The reason this happens is because we only check the "bogus"ness of the display profile because a bogus display profile will ruin all images, whereas a bogus image profile only affects the individual image.

Comment 4

[Chris Murphy]

(In reply to comment #1)
> This happens because qcms detects the monitor profile as bogus and chooses not
> to use it. Chris, do you think our behavior should be different here?

With the test profile set as a display profile on Mac OS 10.6.1, upon launch, Photoshop brings up a dialog, that says the profile appears to be defective, and asks the user what to do.

As an embedded profile it is honored, which again is the correct behavior.

The current Firefox behavior is the correct default behavior. I would entertain the idea of having some sort of about:config preferences for a color management test mode, which would be off by default, and then optionally disable the fail safes we've put into place. This could come in handy for testing purposes, and also conceivably someone might have special needs for strict spec adherence.