On Fri, 11 Oct 2024 07:55:44 GMT, Adam Sotona <asot...@openjdk.org> wrote:
>> Please review this patch that adds a test by @nizarbenalla to perform null >> checks across the ClassFile API. This is an updated version of #20556 that >> minimizes impact on our implementation code. >> >> Notes: >> 1. There's one change in `MethodHandleProxies` to explicitly use platform >> class loader instead of `null` for boot class loader. Tests work fine. >> 2. The null check test uses the same set of classes as CorpusTest to make it >> suitable for tier 1. >> 3. The test is not comprehensive; currently it is permissive toward IAE, >> because sometimes some substitution arguments are invalid, especially >> primitive ClassDesc or Opcode with wrong types. > > test/jdk/jdk/classfile/NullHostileTest.java line 1: > >> 1: /* > > I'm a bit worried about maintainability of such complex parametrized test > with hard-listed method signatures in a text form. I would recommend to find > better location/category and tier for the test. Perhaps the parameters could be annotated with a **JDK**‑internal `@Nullable` annotation, which would then be used by the test to determine whether the parameter is supposed to allow `null`s. -------------------------------------------------------------------------------- This would also ease future migration to [Null‑Restricted and Nullable Types]. [Null‑Restricted and Nullable Types]: https://openjdk.org/jeps/8303099 ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/21458#discussion_r1796699729
