On Sat, 12 Dec 2020 18:59:39 GMT, Phil Race <p...@openjdk.org> wrote:
>>> Wouldn't it have been nice to have a new file having same JNF macros so it >>> will not result in changing any source code but just instead of mapping to >>> apple provided JNF layer, we will map to our macro through JNF wrapper and >>> link/unlink apple JNF lib based on version or so in makefile? >> >> No. Not really. We should not keep the JNF namespace. Nor is there any >> desitre to have that 1:1 mapping. >> This is NOT a "copy of JNF". It is a conscious attempt to be rid of it and >> all the unecessary baggage. >> So I am not interested in that approach even one little bit > > I updated this fix with the additional changes for the accessibility files. > There's a little bit of clean up as some declarations were unused and also > these files shared a single defintion of some classes across files which were > mostly unncecessary. One class I made static and declared per-file which is > cleaner, > Also the only A11Y reference in AWTView was moved into its only call site. > With these changes I think the desktop module is now 100% free of the > JNFCall* pattern and the *CACHE* pattern > I re-ran our full automated headful jtreg test suite and everything passed. I > also ran VoiceOver and ran through SwingSet2 without seeing any issues > although I am not sure what to expect for this so it would be best if > @azuev-java checked it out. Updated with some fixes for the A11Y support. Adding some logging that will report any problems looking up a JNI method or field or class. ------------- PR: https://git.openjdk.java.net/jdk/pull/1679