[0.17.74] Crash while saving: "lua_traceandabort"

[Krypt]

Crash during game saving, progress bar stuck at beginning unusually long and then game crashed.
639 Kb long temp file was generated


About paths: For some reason Factorio able to see through ntfs junctions:

D:\Program Files (x86)\Steam\steamapps\common\Factorio -> F:/games/Factorio
%appdata%\Factorio -> F:\Factorio Data\Factorio

[Rseding91]

Thanks for the report. Do you have any way to reproduce the crash? it crashed in a place that shouldn't be possible. That makes me think there might be some hardware issue on your end.

Can you try running memtest86+?

[Krypt]

> Can you try running memtest86+?
Will do it

[Krypt]

1 pass - no errors (I probably will leave it to run over night later for multiple passes)
My CPU is overclocked (but I have no stability issues) 3.3 -> 4.1 GHz, and system seen some shenanigans in general.

> Do you have any way to reproduce the crash? i
No, it was a single time


Which kind of impossible it was? It was a savegame imported from 0.16, later used a mod to regenerate surface features of explored map cells (WorldGen Update), and it did work less then perfect (overlapping cliffs, at least).

[TruePikachu]

I'd have to pull the 0.17.74 symbols to be sure of the exact crash site (since the error message doesn't appear to be in the log for whatever reason), but the reason why it's "impossible" is because the crash is inside the Lua API. Going by my copy of vanilla 5.2.1 sources, assuming no Lua functions were inlined, either a table wasn't pushed to the stack like it should have been (likely impossible to be the case) or the stack overflowed (unlikely to be the case). Either way, if there was a bug, it probably would have been far more common than a single report.

[Krypt]

Can it be caused by system running out of memory?

[TruePikachu]

Almost certainly not, that would be a completely different error and backtrace. However, a RAM failure could be uncovered in that scenario, if it only exists within an unfrequently-used section of the sticks.

[Krypt]

Well, not yet retested memory. I had a faulty stick of this brand (replaced by store with one of current ones, and, apparently all three sticks are from the same batch). Errors occurred only on 3rd pass (single bit had invalid value in pattern writing/reading test), but system was unstable. But it was couple of years ago and PC works fine.
So... yeah. I need to perform longer test just in case.

[Krypt]

RAM is ok. Does not exclude something like hdd cache corruption, but RAM itself is ok