On Thu, 22 Oct 2020 14:29:28 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:
>> You mentioned the possibility that the OOME is not thrown because it is >> another thread that consumes all memory than the one calling >> forceEarlyReturn() which is supposed to fail with OOME. TLAB could be an >> issue then. In general GC could have a heuristic in place to raise OOME in >> greedy threads when another thread would still be able to allocate >> successfully. I think I could change the debugger part to call >> consumeAllMemory() in the debuggee. This should be executed in the same jdwp >> agent thread as the later forceEarlyReturn. >> >> But honestly I don't think it is worth it and I cannot even test if it fixes >> the issues. I'd prefer to skip the 3 test cases if running with ZGC. Please >> let me know what you prefer. > >> But honestly I don't think it is worth it and I cannot even test if it fixes >> the issues. I'd prefer to skip the 3 test cases if running with ZGC. Please >> let me know what you prefer. > > That's one option if this only happens with ZGC. You also mentioned doing > retries. I would only suggest that if you think you can limit the chances of > the OOME not happening to be so unlikely that we are no likely to ever see it > happen. With the last commit the combined change is - The 3 problematic test cases are skipped if ZGC is selected. - They are also skipped if no OOME during object reallocation can be expected because allocations are not eliminated. - In consumeAllMemory, as a last step, empty LinkedList nodes are created without long array to fill up small blocks of free memory. - EATests.java is removed from the problem list for ZGC. ------------- PR: https://git.openjdk.java.net/jdk/pull/775
