There is an unadvertised 499.1203 version in the build/ dir (we're still testing). Try that if you haven't already. It seems to work better for our other tester.
Happened again on Eternia

http://eternia.byondpanel.com/Eternia.log

It doesn't create core files anymore despite the log saying it will...
Try the above build to see if it helps at all.
I will have to wait for Chance (Writing a new one) before I make any updates or shut his game down, this will also allow him as I to monitor it.

Really wish I could grab his attention @WritingANewOne ;)
In response to A.T.H.K
Please don't post the entire log file from Eternia; it contains some information the players would not want to be shared over a public forum.
In response to Doohl
Fixed.
When we updated the server to 499.1203, the game began corrupting peoples' savefiles as they logged in, rendered most of our maptext inoperable, and some other odd stuff that I don't know the full extent of.
We updated to 499.1203 but experienced some odd interface and spell corruption that forced us to downgrade. Some players were unable to view certain spell descriptions and couldn't drag said spells onto the hotbar.

We can post a bug report but considering that this isn't a public version, I'm not sure if you want us to do that. Also, I can probably the corrupted savefiles if you want to take a look at them.
That's kinda disturbing. I have doubts this is actually being caused by the new version, as 1203 basically just 1) disables the string rebalance check, 2) has some code to allow string integrity to be checked (for the purpose of trying to catch corruption), and 3) has some sanity checks to guard against string corruption. Then again, it's pretty suspicious this happened as soon as you updated, though it seems odd that Eternia would have a problem that SS13 wouldn't.
I dont actually think anyone has properly tried 1203 with SS13 yet, I might be able to bug SoS into testing it for a round or so though
I wish I could confirm that this was an issue with the game itself, but as soon as we reverted back to the previous BYOND version everything seemed to work fine and dandy.
In response to Lummox JR
Lummox JR wrote:
That's kinda disturbing. I have doubts this is actually being caused by the new version, as 1203 basically just 1) disables the string rebalance check, 2) has some code to allow string integrity to be checked (for the purpose of trying to catch corruption), and 3) has some sanity checks to guard against string corruption. Then again, it's pretty suspicious this happened as soon as you updated, though it seems odd that Eternia would have a problem that SS13 wouldn't.

Would e-mailing the corrupted savefiles be of any use? We're fairly certain it's due to the BYOND version considering that it occurred straight after updating (without changing Eternia's version), and since downgrading the problem is no longer an issue.
I would assume that Eternia is being compiled with the latest version of DreamMaker? Or are you using an older version of the compiler? Could this be an issue?

I would ideally like this to take priority over other things many people use Linux boxes to host... But I can't boss you around.. Damn!

FYI Eternia is now on 498.1158
In response to Writing A New One
Writing A New One wrote:
We updated to 499.1203 but experienced some odd interface and spell corruption that forced us to downgrade. Some players were unable to view certain spell descriptions and couldn't drag said spells onto the hotbar.

Huh. Does this happen in 1202 as well? Something is obviously really messed up here. Also, ATHK's comment about the compiler is relevant (it shouldn't make a difference, but perhaps it is coming into play here due to the new higher limits).
In response to Tom
Recently when we upgrade to 1202, it just crashed pretty much straight away, I was not told nor aware of any corruption but we ended up going back to 1197 which played well for a while but as you can see from prior posts ended up failing a few times.
I compiled using 1202. I could throw up a test server that's compiled with 1203 if you think that could prevent whatever is causing the issue.
By the looks of the download page 1203 is Linux only, so unless you want to boot up a VM with a Linux os installed.

I would assume 1203 is only DD changes, DM would most likely be the same as < 1202.
If you can get me a copy of the code to compile, and some savefiles to test on, I can see what I can find running in Windows. I really want to get to the bottom of this. I suspect this is too big to email, but you can send a link to me at [email protected]. I'll be able to look into it later tonight.
On a hunch, I have made another build of 1203 (using a different compilation environment). I doubt this will make a difference, but give it a shot: http://www.byond.com/download/build/499/ 499.1203_2_byond_linux.zip

Most important is to followup with Lummox JR to get to the bottom of this though. We've spent some time with another user on a related problem, but he isn't having nearly the issues with 1203.
As a followup, for any users having this issue, please contact me. I'd like to get together on AIM (you can install Pidgin if you don't have AIM, and just setup an AIM account). Just shoot me a pager message with your AIM name/email so I can add you to my buddy list, since it's private.

Ideally I'd like to know:

- Which operating system you're using
- Which build of glibc you're using
- Which BYOND builds have performed better/worse for you and how

What we know so far:

- The heap corruption has existed since at least 498. It may be exacerbated by recent builds, or that could simply be an artifact of the system allocating things slightly differently.

- Linux appears to be a common thread, specifically Ubuntu. (It's not truly clear if it's only Ubuntu, nor if only certain Ubuntu builds have the issue.)

- Crashes due to this heap corruption may occur much sooner on 1201-2 because of a string rebalancing routine that was added in 1201; it was removed in 1203 in favor of more sanity checks, which are cleaner and safer. The rebalancing isn't bad per se, but it was clouding the issue and it was more likely to choke on damage that might otherwise go undetected for a long running time.

- Two possible sources of heap corruption were fixed in build 499.1202 but we know corruption still exists there.

It's not really a lot to go on. Those who can, I'd suggest running valgrind if at all possible and trying to get some logins going once it starts up. Valgrind will slow the program down, but it will freak out on any invalid writes, highlighting the heap corruption as soon as it occurs. This isn't a viable solution for everyone, but bear in mind a log showing just some issues on a server that barely got off the ground was able to lead us to 1202's fixes, so this is the best way of finding the a smoking gun that I can think of.
Page: 1 2 3 ... 5 6 7 8