[dupe] Unknown version, Windows 7-64: Unspecified headless server crash

[biel]

Compiled using: VC++: 1600 on: Nov 23 2012 00:38:33 platform: Windows endianness: little
The problem is that the headless server crashes in the middle of a game, kicks everyone and restarts. I don't remember the error message but it wasn't the typical "async error".

I hope that this help to slove the bug!

Sorry for my bad english (I'm catalan)

[tomreyn]

Sadly it does not help to report bugs without reporting technical details.

However, based on the compilation time (you did not report the version number) I assume you are reporting a bug in 3.7.1. I think that most bugs in 3.7.1 have, by now, been identified and reported. The issue you are running into is probably a known (and fixed in the subversion repository) memory allocation bug which primarily affects 64-bit Windows systems.

Since this is already known and fixed and because 3.7.1 is kind of old now there is not really a need to provide more information in this case. And unless you wish to discuss this more I suggest we just close this report as [dupe] since the supposedly underlying issue has been previously reported. Are you ok with this?

[biel]

Yes, i'm agree with that, but i thought that the latest stable version always had all bugs (identified) fixed. What does i have to do to fix this problem in my headless?

[tomreyn]

Well, ideally software doesn't have bugs. But factually, unless it's written by D. J. Bernstein or an OpenBSD developer, it actually does. As a result, developers usually try to understand, reproduce, analyze and fix these bugs, then release a new version. The same thing is going to happen here, it's just that there will be a newer version with both bug fixes and new features. And this can take a little longer than just a bugfix release. Still, releasing always involves a good bit of work, and the more often there need to be bugfix releases, the more does it slow down the games' actual development (of new features).

That's why we always ask people to test pre-releases a LOT before the final release, and to provide qualified bug reports, so that bugs in the pre-release version can be ironed out and the final release will thus be more stable, and fewer bugfix releases will be necessary. Sadly, often there is not enough contribution to testing - especially by Windows users, so it can happen that there are releases which work fine on Linux but not so well on Windows. This can be changed when more Windows players help testing pre-releases (snapshots, which are always available, as well as alphas + betas when they are made available).

For the current situation, you can choose from (and combine) several options:

• Write a script which automatically restarts the server should it fail.
• Wait for the next alpha, beta or stable release and upgrade to it.
• Run a development snapshot - but be aware that if you do this it will only be compatible with builds which are made from the same subversion revision - other players (such as those running MG 3.7.1) will either run into an error message or may even crash; also note while the developers try to keep the in-development code mostly stable, there can be no guarantees.

There are loose plans to provide a new pre-release soon (probably within the next couple of weeks/months). This would contain the fix for the problem you reported here.


And Biel responded: Okay, i'll wait for a stable release.

I split her post off this topic since it was mostly about how to get more people to help test development builds.