On Thu, 6 Feb 2025 10:30:39 GMT, Sergey Bylokhov <s...@openjdk.org> wrote:
>>> > Non-ICC intents are not something we'd support. I see absolutely no >>> > problem with that. >>> > If a profile has an intent that is unknown to JDK APIs, it doesn't matter >>> > what LCMS might do with it >>> > if you were programming directly to LCMS as a CMM. >>> >>> non-cc intents in lcms are just examples of profiles in the wild that don't >>> violate the icc spec, currently such profiles can be loaded and used for >>> conversion, this patch will break that. Why we should apply this limit and >>> do that now? >>> >>> Note that it is possible to use custom intents as well, and we will break >>> it: >>> >>> > Little CMS plug-in architecture allows to implement user-defined intents >>> >>> > Who knows how many things we'd have to do in order to support 'custom' >>> > ICC intents. >>> >>> Then let's at least not break it intentionally. Then test and fix if it >>> does not work. >> >> I am 100% for breaking it. We should never have allowed it. > >> I am 100% for breaking it. We should never have allowed it. > > But why? It is allowed by the icc spec, such profiles do not break any rules. @mrserb Thank you for taking time to review this PR. Your suggestions to move up the validation check logic in `.getInstance()` and to check deserialization path helped to make the fix better. At this point, going with what is explicitly stated in ICC Spec seems to be a reasonable fix and the current fix will be integrated. ------------- PR Comment: https://git.openjdk.org/jdk/pull/23044#issuecomment-2640645806
