On Thu, 22 Jun 2023 01:38:05 GMT, Chen Liang <li...@openjdk.org> wrote:
>> The API specification for descriptorString not being a strict inverse of >> Class::forName and MethodType::fromDescriptorString are not entirely correct. >> >> 1. Class::descriptorString was never an inverse of Class::forName, which >> takes a binary name instead. The note about different class loaders is moved >> to getName, as ClassDesc requires an explicit lookup for resolution already. >> 2. MethodType::toMethodDescriptorString ends with a meaningless sentence: >> "fromMethodDescriptorString, because the latter requires a suitable class >> loader argument.", and the "Note:" section can be replaced with an >> `@apiNote`. >> 3. Both of these didn't mention hidden classes (or other >> non-nominally-describable classes) as a reason that prevents the inversion >> operation, in addition to distinct class loaders. Added valid method type >> descriptor/binary name as a prerequisite for the distinct class loader >> explanation. >> >> A few user-defined anchor links are replaced with updated javadoc link tag >> format as well. The explicit html-style links in `@see` tags are unchanged >> in order to retain the non-code output. >> >> The rendered specifications: >> https://cr.openjdk.org/~liach/8309819/06/java.base/java/lang/Class.html >> https://cr.openjdk.org/~liach/8309819/06/java.base/java/lang/invoke/MethodType.html > > Chen Liang has updated the pull request incrementally with one additional > commit since the last revision: > > Missed recommendations from Alan src/java.base/share/classes/java/lang/invoke/MethodType.java line 1229: > 1227: * Note that this is not a strict inverse of {@link > #fromMethodDescriptorString fromMethodDescriptorString}. > 1228: * Two distinct classes which share a common name but have > different class loaders > 1229: * will appear identical when viewed within descriptor strings. Re-reading this, I think this api note reads fine. OTOH I find a little confusing how you rephrased it. "A method type produced by changing a component class object to...". I understand you tried to make it crystal clear two method types may have the identical descriptor strings even if its parameter type or return type are distinct Class objects. I think the original version is clear enough and less lengthy. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/14411#discussion_r1242627816
