I am apparently getting an incorrect ICC profile out of the coarse monitor profiling method in LPROF. Here is what I did: I set gamma and black point, then chose a generic Wide Gamut RGB primaries for the profile, and generated a profile based on these values. The file is correctly interpreted by Firefox 3.0, Geeqie and Gimp, which lets me look at my pictures with somewhat more realistic colors on my wide gamut monitor (HP LP2475).
However, Firefox 3.5 and later, which seem to use a new color management system, refuse the file as invalid. I posted a bug about the problem to Firefox, with the response that my color profile is incorrect as it can't represent a real monitor. Complete discussion of the details is beyond my knowledge:
Digging a little bit in the problem, I saw that LPROF doesn't actually fetch the primaries from EDID at all, as one enters them manually. I chose at that time Wide Gamut RGB, because that was the best result I could get from the proposed primaries values. I also did a try with the decoded EDID primaries information from my monitor, which worked, but gives me a profile with too saturated values. I will attach both profiles to the bug.
So, is LPROF generating a profile that can't possibly represent a monitor, as the Firefox color management guys explained to me? It seems strange that the conversion of the primaries values that my monitor gave me, which, at first glance, don't look obviously wrong, would generate values in the ICC profile that are as far off as described in the Firefox bug above, so might it be a bug in the conversion done by LPROF?
Or is it that Firefox acceptance criteria are too strict?