Page 1 of 2
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-19 15:22
by matty1053
Agreed.
It doesn't crash often on INS.
Maybe the Devs implanted a code that crashes servers if they play a map too often??
#sarcasm
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-23 16:07
by matty1053
Dr_Death wrote:I wonder WHAT Makes maps crashes.
IIRC, it's something with Python.
I remember someone told me in my APC squad like 2 weeks ago on Ramiel that server crashes are caused by a game error. It for example, someone goes to request a AP kit, and it spawns another factions on accident.
But what do I know.
Posted: 2014-08-24 00:13
by Microwaife
Because its a different kind of work done by different people. Most of the time.
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-24 16:41
by Mineral
We don't have a single clue what's causing most of them. And it very likely there are multiple reasons for thme. And we don't have any real way to find out.
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-24 17:52
by Psyrus
Dr_Death wrote:Well..... that's interesting, isn't there a pattern? some kind of asset used in between most of the maps that crashes? a faction? a building of some sorts? an object?
If there is a link, in a good 3 years of looking we haven't been able to find anything. With 100 people doing things at any given time and with literally thousands of lines of python code going [like >40000 lines across all python files] it's extremely hard to track down, especially if it ends up being like an extremely obscure materials issue that only happens once in a blue moon.
Like for example the server crashed on Shijia Valley yesterday
just as the Chinese CAS chopper started firing some rounds at the barn @ main base, but without knowing exactly what the other 97 players were doing at the time, it's very hard to know what the cause was.
The only possible way that I thought we could get down to the bottom of it would be an extremely hacked out piping system whereby the python files get piped to an external process (or just debug buffered on another process) and only once that has happened they run on the engine, and those logs [the last 20 python commands for example] get flushed to a central database/tracker whenever the server restarts unexpectedly. A huge amount of work that after Ancient and DB leave, no one on the team has the skillset to accomplish (actually maybe PiePie and Mats do... I certainly don't)
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-25 03:20
by Psyrus
Dr_Death wrote:What about having a recorded datalog of EVERY python line being active during a match, and when a crash occurs, see what happened with the code line and what made the mistake? of course, this would need a changelog that could describe even as much as every gun fired and every time a reload ocurred or when X entered/leaved a vehicle, but i dont think it has to be done with a full server, if we can make that kind of datalog, and it does records the error, then it could be fixed, if it does not appear, it means the error is so deep inside the code its just useless to fix it.
You just described the same thing that I did in my previous post, but your way wouldn't work because of this issue:
Flushing Buffers - The GNU C Library
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-25 09:38
by Psyrus
Dr_Death wrote:So basically for someone with no programming experience like me that translates into:
"When the game/server crashes all logs are completelly deleted and there is no way to change it"?
The log wouldn't be deleted - if you were handling the logging in the same process as the game (directly in the python scripts) it would execute the function that causes the crash, then before it can flush the buffer to the log, the process would crash and thus you wouldn't see the actual culprit. There are of course ways around this, and I'm not really sure why the programming team hasn't implemented them (besides being a heap of work.. maybe lack of manpower?) but to be honest I'm not an expert here and there may be other technical issues to deal with that I'm unaware of.
Edit: I'll give you an example of what could cause a crash but is like impossible to track down
http://prntscr.com/4g8w8o
I had just joined a game of Fallujah that had been going on for quite a while... and after spawning in and just leaving base (not shooting once) I checked the scoreboard and saw that I had been attributed a kill. I hadn't joined the server earlier that day at all and presumably it had crashed since I last played there several days prior... so there wouldn't have been any residual trace of my player on the server. Yet I was attributed a kill... very odd.
This leads me to believe that sometimes the server erroneously tries to assign points/kills to players that aren't there, and it throws an ArrayOutOfBounds exception, and that might not be handled, which would basically be a segmentation fault ... causing a crash.
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-08-27 07:00
by bren
[R-CON]Psyrus wrote:This leads me to believe that sometimes the server erroneously tries to assign points/kills to players that aren't there, and it throws an ArrayOutOfBounds exception, and that might not be handled, which would basically be a segmentation fault ... causing a crash.
Sounds about right.
Re: Ramiel AAS crashes are getting out of hand
Posted: 2014-11-10 07:58
by Arab
'[R-CON wrote:Psyrus;2030715']If there is a link, in a good 3 years of looking we haven't been able to find anything. With 100 people doing things at any given time and with literally thousands of lines of python code going [like >40000 lines across all python files] it's extremely hard to track down, especially if it ends up being like an extremely obscure materials issue that only happens once in a blue moon.
Like for example the server crashed on Shijia Valley yesterday just as the Chinese CAS chopper started firing some rounds at the barn @ main base, but without knowing exactly what the other 97 players were doing at the time, it's very hard to know what the cause was.
The only possible way that I thought we could get down to the bottom of it would be an extremely hacked out piping system whereby the python files get piped to an external process (or just debug buffered on another process) and only once that has happened they run on the engine, and those logs [the last 20 python commands for example] get flushed to a central database/tracker whenever the server restarts unexpectedly. A huge amount of work that after Ancient and DB leave, no one on the team has the skillset to accomplish (actually maybe PiePie and Mats do... I certainly don't)
Why not this?
Getting Any Kind Of !@#$% Blasted Error Output -- How? - Python - Official BF Editor Forums
Re: Ramiel AAS crashes are getting out of hand
Posted: 2015-04-01 02:43
by Tibbar
i just crashed on this was in a littlebird and held down forward and held down down arrow. is this where we report crashes?
Re: Ramiel AAS crashes are getting out of hand
Posted: 2015-04-01 17:14
by Tibbar
not server my littlebird