[raiguard][1.1.91][linux] "stopped responding" when saving bar was full. (non-blocking save)

[matjojo]

Got a "Factorio stopped responding" message just when a save was almost done. I attached the log file and a screenshot of the screen to show the saving bar.
Running on Manjaro 23.0.1, Gnome, non-steam factorio, non-blocking saves, running the Seablock modpack.

The game did not stop at the first kill request (via system monitor, I believe it sends a sigterm). But did die after the second. I waited around 15 seconds.

Annoyingly I didn't really see anything in the logs.

[matjojo]

Okay so I loaded the game again today, and the quicksave was in the "continue" button and also successfully loaded. Surprised me. Maybe relevant for debugging.

I've attached autosave 1, which is the most recent. It is, weirdly, smaller than autosave 3, the one before that, which I've also attached.

[raiguard]

It appears to have gotten stuck when saving the blueprint library. Could you attach your blueprint-storage.dat file so I can take a look?

[matjojo]

Most certainly, I'm truly not sure if I've made changed to it since then, but this is as it is now.

I could not attach a `.dat` file (invalid file extension error) so I removed the extension, okay, also not allowed to do that. I added `.log`.

Hope that this helps.

[Illiander42]

Similar bug here. Factorio stops responding occasionally. CPU usage drops to zero.

Bug appeared somewhere in the last ~10 updates (Current as of this post: 1.1.92)

Loading the last autosave did not restore game to state when crashed. Timing could be right for the lockup happening on autosave, but not certain.

Game stopped at first kill request.

Gentoo Linux, Steam Factorio, Mate desktop.

What should I grab next time it happens?

[raiguard]

What should I grab next time it happens?

Next time it happens, please upload the save file, your blueprint library, and log file. I haven't reproduced this yet so any extra data would be helpful.

[Illiander42]

It seems to have stopped happening now.

At least I haven't had it happen in a while. Will try to remember to grab relevent files if it does.

[Illiander42]

Happened again on a brand new game.

Hope these are the right files.

[Loewchen]

Hope these are the right files.

Post the save to sync mods.

[Illiander42]

Save. (Crash happened on the first autosave interval on a new game, so this is mostly for the mod list.)

[Illiander42]

Crash right at the end of an async autosave.

2 factorio processes at the time, one was zombie.

[Illiander42]

And an instance without jankey-quality on latest-experimental: Factorio process hanging on "unix_stream_read_generic"

Included everything I could see that might help.

[raiguard]

Thanks for the files. I was unable to reproduce the issue on my PC, but I spent several hours studying the async saving code and found several flaws that I have fixed for the next release. One of them - not closing redundant pipe file descriptors on each process - is specifically touted as being the source of hard-to-reproduce freezes.

Because I don't 100% know if this is fixed, I will wait until I hear from you before I move this to resolved.

Thanks!

[raiguard]

Version 1.1.102 has released and includes the aforementioned changes. Please let me know if the issue happens again. If I don't hear anything for a few weeks I will move this to resolved.

[raiguard]

Moving this to pending until I get confirmation as to whether this has been fixed or not.

[nagapito]

I actually started having issues a couple weeks ago, so maybe around release 1.1.102.
I thought it was actually something on my system and never occurred to me it could actually be an issue on Factorio since it worked fine for months...
Might have not noticed the update to correlate the cause/effect

From what happened today, game started, did first autosave after 10mins, 2nd autosave never triggered and after a couple mins (24-25...) game froze... so maybe something is still hanging and not terminating correctly...

I dont have the laptop with me to try reproduce it but I will try tomorrow and pay more attention so I can give more meaningful data.

[nagapito]

I tried to reproduce the issue today and couldn't...
Until yesterday I could not play for more then 20-30 mins without the crash, to the point I gave up playing while on Linux laptop.
Yesterday, I updated the client, opened a save that was from 1.1.103 on 1.1.104 client, after 20 ish minutes had the crash and stop playing.

Today, continued the only autosave it did, already on 1.1.104 and played for like 5 hours in 2 different sessions with zero issues...

Dont know if the version of the save matters...
Maybe the file contained something incorrect that would cause the crash and 1.1.104 does not save that incoherent state to the file?
The only thing different from previous weeks to today is that I actually loaded a 1.1.104 save on a 1.1.104 client...

[raiguard]

Thanks for your findings, could you let me know what distro/wm/hardware etc you are running? I have never been able to reproduce the problem on my machine.

I'll keep looking for potential deadlocks in the code.

[nagapito]

I am running a Fedora 39, Kernel 6.7.3-200.fc39.x86_64 on a Asus ExpertBook B1 B1400CEAE-71EHDPP1

[raiguard]

The next time anyone reproduces this issue, please make note of your RAM usage and provide your system journal output.

Also, if you can, hook GDB to the frozen processes and print the stacktraces.