2016-12-01 15:20 GMT+01:00 Mark Thomas <ma...@apache.org>:
> It is the start of a new month so I'm planning to tag 9.0.0.M14 and
> 8.5.9 soon(ish). As usual, I want to work through the open bug reports
> and any other known issues first.
IMO 8.5.8 is not really usable due to 60372 so +1 for a release.
> I suspect the actual tags will be next week at the earliest. Rémy
> mentioned Async timeouts + HTTP/2 as a possible root cause of a bug and,
> while the OP wasn't using HTTP/2, I took a quick look and there appears
> to be some work required in that area. On the basis anything async
> related is rarely quick and simple, my guess is that it will take into
> next week to fix this along with the open bugs.
> The stream had a big setSocketWrapper(null) so that's why I asked, but
actually it's the same at the end of a regular HTTP/1.1 request when
recycling the processor. So there's a race between that and the removal in
the waitingProcessor set which occurs during processing. So it could be
enough to skip over if the socketWrapper is null in AstractProcessor.