On Thu, 18 Nov 2021 10:02:45 GMT, Magnus Ihse Bursie <i...@openjdk.org> wrote:
>> @mlchung the way I looked at this was the filesMap is a HashMap, so the >> filesMap.keySet() iteration order is not defined (or necessarily >> deterministic), the expanded order of the file sets for each version could >> be ver1, ver2, ver3 or ver3, ver2, ver1... the next time you run it with the >> same parameters. >> LinkedHashMap will ensure the iterated order is as inserted (or as specified >> in the jar args list..) >> >> Although you say "filesMap is sorted by version", I don't see where that is >> done? > > I'm guessing Mandy asks not if the keys are sorted -- this is changed by your > patch, but if the value retrieved -- the String[], is sorted. It's a valid > question, I think. It's not apparent to me at least that this is the case. If > it is not sorted when it is inserted it needs to be sorted when extracted. @magicus ah right, got you. So the order of each version's String[] is as ordered on the jar cmd line args: https://github.com/openjdk/jdk/blob/3e87db19bb7e2359e4b1ceaac7c123542345ecb1/src/jdk.jartool/share/classes/sun/tools/jar/Main.java#L664 So for the same jar cmdline input, the output will be deterministic, as in the same order. If we think we need to be deterministic for different jar cmdline input for the same set of files, I can add a sort, but that wasn't my aim. eg: both these produce the same order? jar --create --file my.jar --release 9 v9classesdir1 v9classesdir2 jar --create --file my.jar --release 9 v9classesdir2 v9classesdir1 My aim was same input == same output ? ------------- PR: https://git.openjdk.java.net/jdk/pull/6395
