Jump to content
=VG= ciro

Server Issues (Fly South Bug FIXED!)

Recommended Posts

1 minute ago, =VG= Fastjack said:

You laughing but you right. The airvehicles of BF2/PR use a codeline that is called Wing and that codeline is responseable that the Vehicle can take off.

lol. then lets try out if a tank can crash the server if he got "wings"^^

Share this post


Link to post
Share on other sites
18 hours ago, =VG= Melon Muncher said:

So gonna close this issue. It's been found to be a BF2 engine issue. Sadly there's nothing that can be done against it.

If you don't believe me you can find a BF2 copy and install a 64 player mod and create local servers. The issue will occur after a few map changes.

I'll look into some changes to try and make it occur less frequently. (removing osprey on 2km maps, jets on bamyam and silent eagle removal/base swapping etc)

As for what happens when someone deliberately crashes the server. Make a note of the map/player and we can check the logs and ban. Accidents will happen, but with a bit of work we can try and limit the issue.

 

Big thanks to everyone for the help. Bad  outcome though

I've made a final report on the Server Admin COOP forums of the PR website.  I suppose it's out of our hands for now, unless there is something that can be done within the engine itself to fix it.  Again, thanks to everyone who helped!

Share this post


Link to post
Share on other sites
2 hours ago, =VG= SemlerPDX said:

I suppose it's out of our hands for now,

Negativ. its only in our hands i think.

I do not think that it really interests someone at deployment (but cant blame anyone for it)... it seems there are not too big problems with the bug, otherwise you would read something in the forum. and JTG has mostly 0-2 players ...nobody cares. 

2 hours ago, =VG= SemlerPDX said:

unless there is something that can be done within the engine itself to fix it. 

 we should continue to analyze his behavior and try to find a way to neutralize or move around him.

there is different behavour! local crashes a lot less than online. deployment apparently too. maybe we should play a few more rounds bf2 again ... etc.

So it can really be that we can never do anything about it, but it can also be that Kavelenko tomorrow by accident finds out that they do not crash if something else cross the border in the south before or whatever...

 

Share this post


Link to post
Share on other sites
On 10/10/2019 at 5:07 PM, 0100011000101 said:
On 10/10/2019 at 2:19 PM, =VG= SemlerPDX said:

I suppose it's out of our hands for now,

Negativ. its only in our hands i think.

I do not think that it really interests someone at deployment (but cant blame anyone for it)... it seems there are not too big problems with the bug, otherwise you would read something in the forum. and JTG has mostly 0-2 players ...nobody cares. 

On 10/10/2019 at 2:19 PM, =VG= SemlerPDX said:

unless there is something that can be done within the engine itself to fix it. 

 we should continue to analyze his behavior and try to find a way to neutralize or move around him.

there is different behavour! local crashes a lot less than online. deployment apparently too. maybe we should play a few more rounds bf2 again ... etc.

So it can really be that we can never do anything about it, but it can also be that Kavelenko tomorrow by accident finds out that they do not crash if something else cross the border in the south before or whatever...


Regarding the first thing you said, while it's true that the focus of PR is Deployment, and it has been expressed that if COOP development was holding that back and no one existed to work on COOP, that the game mode would be dropped - that does NOT mean that the R-Dev's are okay with letting bugs persist, etc.  Bugs are reported and eventually fixed all the time for both game modes.  This one was unique in that it took a time to even know that it was a real bug, then time and drama happened, and even now I just learned that some miscommunications happened, too.

 

I was under the impression that crash dump reports had been forwarded to the R-Dev's a long time ago.  The other evening, I worked with Kavelenko to capture a crash dump report with procdump that was directly related to a Fly South server crash.  I forwarded this info to the R-Dev's, and along with the other info in our report, they've identified it and will attempt to work it out before the next major release (don't spread rumors, don't get hopes up, but a fix for this is coming if not in the next update, then soon).

 

TLDNR;   This bug has now been officially identified (I asked if I could directly quote the R-Dev from the private forums, will post if able) - and it will be fixed in an upcoming update/patch, hopefully the next one, if not, the one after that.

  • Like 1
  • Upvote 2

Share this post


Link to post
Share on other sites

 

Quote

AlonTavor wrote ...

Thanks, I have found the exact cause.


Someone forgot to check bounds on the heightmap array in the AI module. Going east or west just reads the next row of the array. Going north or south overflows/underflows the array and reads junk data. If memory layout is such that before/after the array the page is unalloacted, accessing that memory segfaults.

Interestingly, this seems to happen only with south? I guess the initialization order could guarantee that the next pages are allocated.
Since memory allocation methods depend on the OS, on some systems it can happen more often.

I'll try to fix it for 1.6

  • Like 2
  • Upvote 2

Share this post


Link to post
Share on other sites
8 hours ago, =VG= SemlerPDX said:

TLDNR;   This bug has now been officially identified (I asked if I could directly quote the R-Dev from the private forums, will post if able) - and it will be fixed in an upcoming update/patch, hopefully the next one, if not, the one after that.

WOW.

tears of joy. that would greatly enhance the quality of the game and I think our server would be fuller. we lose a lot of players with every crash ...

  • Like 1
  • Upvote 2

Share this post


Link to post
Share on other sites

This issue should be gone now.  The fix is in place and the server has gone for the last 8 hours or so without a crash.

Please actively try to crash the server by flying south of the map border, and report if it actually crashes the server.  We think it will not.

If the server does crash, please note the time (in ZULU/PRT/GMT) and the map (and if known, the next map that was going to be loading).  The crash dump system is still running, so I can cross reference any crashes with dump logs and provide them to Alon if needed.  As the issue involves memory allocation, it may be difficult to reproduce even if it is not fixed.  But we're assuming it is.

 

Thanks again to everyone for your patience and understanding throughout this whole thing!!  Cheers! :drinks:

  • Like 4
  • Upvote 3

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.

×