Doomsday Engine

I tried this on OS X with two local servers, but it didn't trigger any problems. Next I'll have to set up suitable remote servers and try it on Windows.

BTW, were you using any resource packs in these games?

Tried local servers on Windows (without resource packs), didn't get a crash.

I wasn't using any resource packs and was trying to join a non-local server.

Specifically, various ones of Kuri Kai.

It could be a version mis-match (and that causes a seg fault when one tries to join a server with a different version from a game other than the one the MP game is for).

During my Windows debugging I discovered and fixed #1960, which may have been the primary culprit here, since it was possible to trigger it also with key/mouse release events.

Needs to be tested again with Friday's build (or later). I'll keep the MP test servers running for convenience (occasionally if not all the time).

the above gdb output is the server and the client on the local machine, so both are build 1500.(not the nz/au servers)

vermil's test from above are most likely to the nz/au. whcich he is correct that they are version mismatches as they are the latest stable build

That does show a crash occurring, however without debug symbols it doesn't help much. Any chance you could do a debug build? That would be extremely helpful.

I'll try local servers on Linux myself and see if I can reproduce it.

attached now is the output from a debug build

config_user.pri contained

CONFIG += deng_extassimp
CONFIG += deng_debug
PREFIX=/opt/deng1500

From the .2 log:

Reading symbols from doomsday...(no debugging symbols found)...done.
Executable: Doomsday Engine 1.15.0 (Unstable 64-bit) Feb 9 2015 18:18:11

Did you rerun qmake after modifying config_user.pri and do a clean rebuild? This should say "Unstable 64-bit +D +R", and gdb should be finding debug symbols.

CONFIG+=deng_debug

This should be just "CONFIG+=debug".

debug build output uploaded

Thanks! That indicates the crash is caused by the UI widget, I'll look into it.

I've fixed the likely cause of the crash in the UI widget. A new build will be available later today for testing.

I discovered and fixed another crash. Just as vermil suggested, first having a version conflict with a server and then joining another was causing invalid network packets to be buffered on the client. I've now fixed this for the next build (not in 1502).