Http-bind room bug

When requesting the room list or joining a room, the server only responds after the wait time out.

Currently sparkweb is setting a wait value of only 2 seconds (which is very close to just polling) so these kind of errors don’‘t show up so easily when testing with sparkweb. I’‘ve altered my version of sparkweb to set a wait of 20 seconds (which is still much quicker than my live installation). If you do the same then many of the issues I’'m having will be apparent.

It is not possible to create a room on any wait setting as far as I can tell.

We’‘re seeing a similar bug where it just continues to sit and spin. We’'re running sparkweb behind an Apache proxy where we have /sparkweb and /http-bind forwarding to an internal port. Is this a known bug with the new SparkWeb beta?

It’'s not just Sparkweb.

It’'s something in the http-bind implementation in wildfire.

Hey guys,

I have changed the hold value to 10 on my local development version. Running the latest and greatest HTTP-Binding and SparkWeb and I am currently unable to reproduce this issue. There have been a few improvements to HTTP-Binding since the release of 3.2.0, and just before the release (igniterealtime.org has yet to be upgraded):

JM-955

JM-959

SparkWeb also needed to me modified to take advantage of the latest and greatest HTTP-binding implementation. It cheated a little before Wildfire could handle multiple connections it would tell it to hold one connection open even though it could only handle one connection at a time.

Thanks,

Alex