On Thu, 22 May 2025 18:56:29 GMT, Daniel Jeliński <djelin...@openjdk.org> wrote:
>> We observed a case where the instants returned by `TimeSource.now()` were >> returned in non-monotonic order. The reason was that sometimes we were using >> a delay calculated with one `localSource` as an input to a different >> (updated on another thread) `localSource`. This was confirmed by putting >> `assert firstNanos + delay == nanos;` under `instant(long, long)`. >> >> The fix ensures that we won't accidentally use the incorrect delay by >> removing the `instant(long, long)` overload, and calculating the delay in >> the method where it is used. >> >> No new test; instrumenting this class for testing would likely double its >> size. Tier2 tests continue to pass. > > Daniel Jeliński has updated the pull request incrementally with one > additional commit since the last revision: > > Restore instant(long,long) src/java.net.http/share/classes/jdk/internal/net/http/common/TimeSource.java line 100: > 98: long nanos = System.nanoTime(); > 99: long delay = source.delay(nanos); > 100: // use localSource if possible to avoid a volatile read Hello Daniel, the changes look OK to me. However, this pre-existing comment is now misplaced. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/25390#discussion_r2103819574
