wojciech.kapcia@tigase.net opened 4 years ago
|
|
Reason is always the same:
I'd guess that something in this change https://github.com/tigase/tigase-server/commit/da3b2916c6c1a76c694d3ef821588e4ccac09c42 is throwing an exception (NPE?) that kills the connection, but at the same time, there is no relevant ST. Question is - what, and why only sometimes and randomly? I'll continue tomorrow... @andrzej.wojcik - anything comes to mind? |
|
Just a while guess, but what if we would use BouncyCastle as a TLS provider? Why? Because I'm not sure right now what is Jaxmpp using. It used to use JCA but there were some works with BouncyCastle to provide SASL (SCRAM-*-PLUS) and I'm not sure what is being used. It could be the some combination of BC and JCA is causing the issue ie. due to some error. Please also look at
Do we have here an invalid number of arguments? more arguments then we use in the template? The same issue is in line 165. Moreover, I do see a weird thing. In
It looks to me that for some reason
It looks to me that the root cause is Maybe vhost repository is not loaded or there was an issue with loading a vhost. We've made a few changes in the |
|
It's using JCA in TTS-NG. Probably need configuring BC (same as in Tigase)
Yes we do (I'll correct it) but this should not matter - it would case IndexOutOfBound exception if we had more references than items in array AFAIR (in current case it would simply not use last item)
I made this change to somewhat limit debug verbosity (information about vhost details were already available, in this case we only care about vhost name) though the change itself (even with above line to prevent accidental NPE) should not matter in this case.
Yes, this should not happen (having NULL here) and it's most likely not happening (haven't reviewed the logs yet) and before this commit (I assume) it was not happening so we could assume that there wasn't any NPE in this place. I just made a couple of changes that could prevent possible NPEs but tests are still failing. (it's kinda annoying that there is no detail with why the connection is broken)
Yes, but it was working just fine before so previous VHost changes should at least be manifested before. |
|
Looking at the summary page, it seems that the problem started with build
so the commit in question seems to be unrelated. There were two others:
Given that all those failures/skipped tests are related to failure in TLS during registration I'd say the second one would be the most likely cause. I reverted the commit from the jaxmpp and it seems number of this type of failures dropped significantly. I'm starting to suspect that this is some sort of concurrency issue in jaxmpp (can't reproduce it locally)... |
|
It seems it boils down changes in TTS-NG; commits up to 2a64a551 works fine so far. |
|
StackTrace:
|
|
After some work all tests are finally passing. Mostly done:
|
Type |
Bug
|
Priority |
Normal
|
Assignee | |
Version |
tigase-server-8.1.0
|