Factorio Forums

The game itself has been saved for the last few hours, without my participation.
looked two previous save are present. Tried to save manually two more times, with the same error.
Rebooted the computer and loaded the previous save. After it, the game has been saved several times normally.
I sincerely hope this is an isolated case of failure.
But the question is this: I had to check what was there in the coordinates (-28.85)?
upd: at the coordinates (-28.85) I have the edge of the railway, which I have not changed for a million years.

For an unsigned integer -28 with the first bit flipped becomes 2147483620, so quite possible you just had a random bit flip. It happens, you'll see reports of it on the forums decently often.

So this is not an error, but quite a common occurrence The main thing is that it does not happen so that none of the last three saves will load
I already sometimes double-check what I have written, but for me it is somewhat tedious, because the full cycle (saving with subsequent loading) is 20 minutes + 20 minutes.
Thanks for clarifying. Hopefully this doesn't happen again.

azesmbog wrote: ↑

Mon Jun 21, 2021 8:05 pm

So this is not an error, but quite a common occurrence

Yep, bit flips are quite common. That's why servers use ECC memory.

bormand wrote: ↑

Mon Jun 21, 2021 8:43 pm

azesmbog wrote: ↑

Mon Jun 21, 2021 8:05 pm

So this is not an error, but quite a common occurrence

Yep, bit flips are quite common. That's why servers use ECC memory.

I'm quite interested when I see sentence like this - how common are they?
I spent quite a lot of time with factorio and lots of time gaming in general and I don't recall having any visible signs of those. And thats without ECC memory.

I know that one of games had a built in check added to actually detect corruption in runtime (Fortnight?) and it was used to detect a lot of strange corruption. And I do see bug reports here with rendom crashes from time to time.
But is it really that common?

orzelek wrote: ↑

Mon Jun 21, 2021 9:17 pm

I spent quite a lot of time with factorio and lots of time gaming in general and I don't recall having any visible signs of those. And thats without ECC memory.

It probably varies largely by computer.

orzelek wrote: ↑

Mon Jun 21, 2021 9:17 pm

But is it really that common?

https://stackoverflow.com/questions/2580933

As for the games, most of them are graphic heavy, so you wouldn't notice a slightly changed pixel or a bugged vertex.

bormand wrote: ↑

Mon Jun 21, 2021 10:30 pm

orzelek wrote: ↑

Mon Jun 21, 2021 9:17 pm

But is it really that common?

https://stackoverflow.com/questions/2580933

As for the games, most of them are graphic heavy, so you wouldn't notice a slightly changed pixel or a bugged vertex.

Thanks for the link it's quite informative. I guess it's much easier to get a corruption in data that won't be noticed that on executable code that would crash instantly.

This time I hit much stronger: ((
Usually I make backups every 2-3 days, this time I only had a ten-day copy dated June 26 (working and proven)
This is a few hundred hours of play
I don't know if it's the same error or not, but this time the map didn't load at all, I tried all three last copies(saving every hour). I attach the logs, suddenly someone is interested))
Moreover, the game worked to the last, without shutdowns ...
Well cho, it is his own fault, it is necessary to be saved more often ... The first rule of the system administrator.

You might want to check your system for any HW issues, and if there's any overclocking done on the system, remove it. Bad tile data can also be a result of bit corruption, or a HW problem.

Good advice, of course.
but in the current configuration I have been playing for at least the last six months and several thousand hours, and the problems appeared on version 1.1.35, and as you can see from the forum at least two more players, so I tend to degrade the game in this version: ((
I would gladly return to version 1.1.30, but alas, this is impossible in this paradigm

There have been reports of bit flips since long before the current version, before 1.0 even - that's why the first thing I checked on your post was if it was likely a bit flip, because they're often the cause. Even your recent ones could be the same thing - any single bit value can become 0 with a single flip, and a single bit can make a 9 become a 73.

'It worked yesterday' is never a guarantee it will work tomorrow.