This plugin uses auto-discovery to run checkers in most builds, via the CreateManifestTask. If I recall correctly, the original motivation for that decision was that the user might be relying on auto-discovery for other annotation processors, so if the user's build did not explicitly specify annotation processors, we couldn't do so either (because if we specified annotation processors directly, then auto-discovery would be disabled).
On JDK 23 with auto-discovery disabled, this plugin's job should be simplified: we can just modify the list of annotation processors directly, since an explicit empty list of processors will have the same semantics as the lack of any annotation processing options. However, we'll probably need to retain the current behavior of the plugin for the new -proc:full mode, which keeps the pre-JDK-23 behavior.
Annotation auto-discovery will be disabled by default starting in JDK 23: https://mail.openjdk.org/pipermail/jdk-dev/2024-May/009028.html
This plugin uses auto-discovery to run checkers in most builds, via the
CreateManifestTask. If I recall correctly, the original motivation for that decision was that the user might be relying on auto-discovery for other annotation processors, so if the user's build did not explicitly specify annotation processors, we couldn't do so either (because if we specified annotation processors directly, then auto-discovery would be disabled).On JDK 23 with auto-discovery disabled, this plugin's job should be simplified: we can just modify the list of annotation processors directly, since an explicit empty list of processors will have the same semantics as the lack of any annotation processing options. However, we'll probably need to retain the current behavior of the plugin for the new
-proc:fullmode, which keeps the pre-JDK-23 behavior.