You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
JCK 8 have (for linux) existing text extractor, which simply lok to the failed test, and find the class, read all depndencies, and pack them to tarball which you can extract and immediately run.
This was removed in jck9, because it was not perfect - the native part of test was missing. However native tests are less then 1% of all tests, so we had implemented our own via compiler api: https://github.com/zzambers/jck-test-extractor (will be moved to https://github.com/rh-openjdk/ before implementation). In addition there is one, by accident missing peace which we forgot internally, the search for main method and conservation of additional javas (-D, -X...) and test's arguments. This will be added slightly later.
Generally when there is no native code involved, this is (especially on secondary architectures) saving quite a lot of time, as the (working) reproducer is part of jck run artifices, and can be just unpacked and directly run. If this will be implemented, there must be check, that only eg first 10 reproducers are generated, as in case of borked run with all tests down the full generation would be unbearable by both time and space.
There were unofficial rumours about official reproducer generator appearing again in jck21 with better test definition, but it seems it did not happened.
Wdyt, worthy to include?
The text was updated successfully, but these errors were encountered:
JCK 8 have (for linux) existing text extractor, which simply lok to the failed test, and find the class, read all depndencies, and pack them to tarball which you can extract and immediately run.
This was removed in jck9, because it was not perfect - the native part of test was missing. However native tests are less then 1% of all tests, so we had implemented our own via compiler api: https://github.com/zzambers/jck-test-extractor (will be moved to https://github.com/rh-openjdk/ before implementation). In addition there is one, by accident missing peace which we forgot internally, the search for main method and conservation of additional javas (-D, -X...) and test's arguments. This will be added slightly later.
Generally when there is no native code involved, this is (especially on secondary architectures) saving quite a lot of time, as the (working) reproducer is part of jck run artifices, and can be just unpacked and directly run. If this will be implemented, there must be check, that only eg first 10 reproducers are generated, as in case of borked run with all tests down the full generation would be unbearable by both time and space.
There were unofficial rumours about official reproducer generator appearing again in jck21 with better test definition, but it seems it did not happened.
Wdyt, worthy to include?
The text was updated successfully, but these errors were encountered: