On 05/04/2023 17:46, Rémy Maucherat wrote:
On Wed, Apr 5, 2023 at 5:32 PM Mark Thomas <ma...@apache.org> wrote:
On 05/04/2023 13:01, Mark Thomas wrote:
Good news and bad news.
The good news is that the issues were with the test. The test has to use
statics to pass data to Endpoint instances and they weren't being reset
properly between tests. With that fixed there were 2 failures that are
both fixed by the proposed patch.
The bad news is that the proposed patch fixes one possible cause of the
deadlock but introduces another.
I'm working on an updated patch.
I have a fix but...
The proposed patch for BZ 66508 (both the original and the updated
patch) depend on the change from synchronizing on socketWrapper to using
a ReentrantLock. This change was made to 10.1.x onwards to support Loom
experiments but was not back-ported to 9.0.x and earlier.
https://github.com/apache/tomcat/commit/0a9480158874ea910a4d629d24f31d69d6cc5f96
What do folks think about back-porting this change to 9.0.x and 8.5.x?
Time to backport it then.
Thanks for the review.
After I wrote the above email I started to think about ways of
back-porting the fix that didn't require the synchronized ->
ReeentrantLock change.
If we need to back-port that change so be it, but if I can find a way to
back-port the original fix and keep synchronized that isn't any more of
an ugly hack than the original fix then I think I'll go that route as I
don't want to change existing API (even internal API) unless I have to.
Mark
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
