On Thu, 20 Nov 2025 11:05:59 GMT, Severin Gehwolf <[email protected]> wrote:

>> Please review this simple fix to `jlink`'s `--release-info` plugin to handle 
>> non-ASCII in vendor strings. The JDK build uses UTF-8 encoding for the 
>> produced `release` file that is being passed to `jlink` at build-time via 
>> the `--release-info` plugin. However, the plugin internally uses 
>> `java.util.Properties.load(InputStream)` API which assumes `ISO-8859-1` 
>> encoding of the input stream. The proposed fix is to use the 
>> `java.util.Prorperties.load(Reader)` API instead and pass it an 
>> `InputStreamReader` with `UTF-8` encoding.
>> 
>> Testing:
>> - [x] GHA
>> - [x] `test/jdk/tools/jlink` tests including the new reg-test which fails 
>> prior and passes after the fix.
>> 
>> Thoughts?
>
> Severin Gehwolf has updated the pull request incrementally with one 
> additional commit since the last revision:
> 
>   Specify that the input file shall be UTF-8

> Question is if it's worth supporting arbitrary _input_ encondings for 
> `--release-info <file>` usages. `<file>` is currently in UTF-8 in the JDK 
> build, so went with that expectation. But arguably it could be any other 
> encoding a user chooses. So to make it generic, it's conceivable to allow 
> `--release-info=<file>,<encoding>` or some such to allow non-UTF-8 as input 
> and still do the right thing (Note: output would still be UTF-8). Perhaps 
> it's not worth the trouble?

That seems unnecessary complexity, I think we should just keep it as UTF-8.

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

PR Comment: https://git.openjdk.org/jdk/pull/28399#issuecomment-3557683372

Reply via email to