[asherkin]

Originally Posted by Dragokas Asherkin, nice! Good job. Can you please answer the first part of my question?

Your crash doesn't have any plugin code on the stack - there is no plugin directly involved.

You can see that it passed though SM in 2 places: the hook for the OnGameFrame forward, and the hool for the OnPlayerRunCmd forward, but the server has continued much deeper than those and is busy executing the games internal PlayerRunCommand code.

Now, the stack may not be 100% correct, generally things are more likely to be wrong the further away from #0 you look (which is always correct), and that one is extra questionable because of the odd recursive behaviour with CountdownTimer::Now at the top and the fact that all the frames have been found using stack scanning (L4D must be one of the only Valve games compiled with frame pointers omitted, and without debug symbols it has had to do some guesswork to reconstruct the stack, so it is very possible for it to go off on some wild tangent.)

A directly plugin-caused crash looks a lot more like this: https://crash.limetech.org/fqoikydcdxdc

Originally Posted by Dragokas offtop. Are there any settings in sm - my server doesn't create crash logs sometimes (but, I see RCON console show that crash happens). The case of infinite server rebooting (possibly, incompatiblity between 1.10 and my plugins, 1.9 - work ok).

On Windows, Accelerator can miss some crashes due to how it is implemented, but on Linux it will catch all abnormal terminations. There are cases in the game/engine where it exits cleanly rather than crashing, which may be what you're seeing. There are also cases where Breakpad cannot generate a crash report due to OS issues, but that is generally not intermittent. In any case, Accelerator doesn't upload any crash reports until the server has started normally again, but they're all saved even when it is restarting in a tight loop (as long as it is getting far enough to load Accelerator) and will be uploaded later.