On Tue, 4 Mar 2025 13:05:30 GMT, Alexey Ivanov <aiva...@openjdk.org> wrote:

>>> This builtin flag check is exclusively for JDK built-in profiles created 
>>> within the private interface `BuiltInProfile` and not applicable if an 
>>> application loads the profile by serialization.
>> 
>> The built-in profiles can be serialized and deserialized, 
>> [see](https://github.com/openjdk/jdk/blob/c4b516dfe7c5a5fddd4d9c97a21f5f36bf845646/src/java.desktop/share/classes/java/awt/color/ICC_Profile.java#L1573).
>>  This means the application may get a reference to the profile not only via 
>> the getInstance(id) method. Therefore, the new spec is not strictly accurate 
>> regarding ICC_Profile.getInstance(int colorSpaceID), but I am not sure that 
>> we should mention serialisation here.
>> 
>> **Note:** 
>> The new Javadoc for the new private built-in flag is "public" via the 
>> serialized form of the class. It probably should be marked as transient, so 
>> it will only be set to true when the built-in profile is loaded via 
>> serialization. In all other cases, it will be false.
>
> This is an interesting observation.
> 
> Are the unserialized built-in profiles mapped to their corresponding 
> singleton instances? I presume it's the case. Then, the `builtIn` flag will 
> still be set to `true` in the returned object.
> 
> **Should the serialized form be preserved?**
> 
> If the `builtIn` field is `transient`, built-in profiles written by JDK 25 
> (mainline) could still be read in previous versions and vice versa. This will 
> improve the backward compatibility.

Good point. I had not considered the effect of serialization on builtIn flag, 
let me check and get back.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/23606#discussion_r1982109310

Reply via email to