[NEEDSINFO] JavaScript OOM causes segfault on Linux

[crosstalk]

When starting a multiplayer game, the client crashes. It reports a JavaScript OOM error then segfaults.

The attachments were created joining another player's game, but both instances crashed.

[crosstalk]

Backtrace, obtained via GDB

[historic_bruno]

Which version of the game are you using, Alpha 13 or building from SVN? Which OS, 32-bit or 64-bit? Does this happen with every multiplayer game you try?

Also the GDB bt isn't really useful since it's missing debug symbols. If you're using a release package, maybe you can install a debug package to get symbols and a more useful bt (for future reference; in this case it doesn't matter so much, since the OOM crash is easily reproduced on *nix).

[crosstalk]

Alpha 13, from the Ubuntu 13.04 repositories. It is a 64-bit Linux Mint install. It does not not occur with every multiplayer game -- just this one scenario.

The map was Deep Forest and one player was playing as the Persians -- after changing those two settings, we managed to successfully start a game (I assume the map is more likely to affect the AI players than another player's civilization).

I figured the backtrace wouldn't be useful, but I thought I'd post it anyway. It doesn't appear that there are pre-built Ubuntu packages with debug/GDB symbols available -- some time I might learn how to build my own packages, but it'll probably be a while before I have enough time to do so.

[historic_bruno]

Yeah, deep forest is probably more vulnerable to this than other maps, due to the huge number of entities it can have. The good news is in SVN we've increased simulation runtime size from 16MB to 32MB, so I doubt you will have the same error in A14. It's not a fix for the crash, but it means you're less likely to have OOM in the first place :)

(By the way, if you do have time to test, it's pretty easy to check out SVN or git and build, see BuildInstructions - A14 should be out in weeks)

[historic_bruno]

See also r13794.

I'm bumping this to A15 and keeping the ticket open because the segfault from lack of handling OOM conditions is the worse problem, no separate ticket exists for that (the closest is #742).

[Stan]

Is it still accurate ? We updated spidermonkey so I guess such memory leak are invalid. Closing it as needinfo feel free to reopen it.