RFR: 8326661: sun/java2d/cmm/ColorConvertOp/ColConvTest.java assumes profiles were generated by LCMS [v2]

[Dmitry Markov]

On Fri, 8 Mar 2024 01:09:27 GMT, Sergey Bylokhov <serb at openjdk.org> wrote:

>> I agree, I reverted almost all changes except the ones in tolerance for non-LCMS profiles. That values need to be updated to make the test pass on JDKs (e.g. JDK 8u, etc) where non-LCMS profiles are used
>
> Could you please clarify what is the root cause of the problem. I assume you did not update the profiles itself, so what is the problem in the new lcms library?
> 
>>It used to be OK until recent LCMS update where the CMM started to keep original profile ID instead of writing ‘lcms’ to returned header. 
> 
> Do you mean that previously we always use lcms thresholds even for kcms related profiles?

As far as I know there is no any issues with the new LCMS library. There was a bug in LCMS: it used to set ‘lcms’ as profile ID to the header. As a result we used lcms threshold even for non-LCMS (kcms, etc.) profiles in the test.
In LCMS 2.16 the problem with profile ID was fixed, see https://github.com/openjdk/jdk/pull/17382/files#diff-738d14c3b278e6d7297dbc4943e90cef33fc04d61eba085218a1229c92ea9a33R941
And the tests started failing for non-LCMS profiles.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/18097#discussion_r1517499065