On Fri, 2 May 2025 04:56:51 GMT, Shaojin Wen <s...@openjdk.org> wrote:
>> As another step toward the removal of the old generics infrastructure, I >> propose to remove the usages of generic parsing utilities and use the >> facilities provided by BytecodeDescriptor, already used by >> MethodType.fromDescriptorString. This also prevents extra validation cost in >> use sites to defend against generic types. >> >> In this patch, BytecodeDescriptor and Wrapper see minor updates, mainly for >> better exception messages - previously, an unparseable char in the >> descriptor string just reports that char, and now the whole descriptor >> string is reported. >> >> These behaviors are already covered by the tests added in JDK-8350704 #23788. >> >> Testing: reflect/annotation/Class, running tier 1+2 > > src/java.base/share/classes/java/lang/Class.java line 1565: > >> 1563: Class<?>[] parameterClasses = >> types.toArray(EMPTY_CLASS_ARRAY); >> 1564: >> 1565: final Class<?> enclosingCandidate = >> enclosingInfo.getEnclosingClass(); > > Suggestion: > > Constructor<?>[] candidates = enclosingInfo.getEnclosingClass() > > .privateGetDeclaredConstructors(false); > > Here the local variable `enclosingInfo` is only used once, the above code may > be simpler This patch intentionally did not change this part - and this part can get much more in-depth optimization; for example, arrayContentsEq can be shared for all `Class` arrays, copy of reflective objects can be simplified, etc. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/24978#discussion_r2071104255
