Yet Another Linux Crash Bug

[CODE][vinland] :: New login: New Client (2.22+)
[vinland] Version: GNW03014 (2.22)
[vinland] Account: EricLane

bit length overflow
code 3 bits 6->7

bit length overflow
code 4 bits 6->7

bit length overflow
code 16 bits 5->6
[vinland] :: Client disconnected: EricLane
[login] :: New login: New RC (2.22+)
[login] Version: GSERV024 (1.1)
[login] Account: animelovalol
[login] Attempted RC login by animelovalol.

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xb7d14b70 (LWP 20977)]
0x00000000 in ?? ()
(gdb) bt
#0 0x00000000 in ?? ()
#1 0x08095c0b in TServer::doMain (this=0x815d120) at …/…/server/src/TServer.cpp:232
#2 0x080954ec in TServer::operator() (this=0x815d120) at …/…/server/src/TServer.cpp:116
#3 0x080889b5 in boost::detail::thread_data<boost::reference_wrapper >::run (this=0x8182d30) at /usr/include/boost/thread/detail/thread.hpp:81
#4 0xb7fcfc45 in thread_proxy () from /usr/lib/libboost_thread.so.1.40.0
#5 0xb7d37585 in start_thread () from /lib/i686/cmov/libpthread.so.0
#6 0xb7e1926e in clone () from /lib/i686/cmov/libc.so.6
[/CODE]

Annie May, stop trying to log into Vinland as RC!!

Hmm. Can you reproduce this consistently? If so, does the latest SVN version I committed today fix it?

___Merged doublepost__________________

Also, did this gserver host multiple servers at once?

told you he hacks

1.1 I have no idea, I haven’t tried.
1.2 We’ll just have to wait and see. I have recompiled to the latest version on SVN.
2. Yes.

Well, we will see what happens. It could be an issue where one of the threads died. I’m not sure, though. #0 0x00000000 in ?? () doesn’t really tell me anything. =/

Yeah, well, that’s what I got. :stuck_out_tongue:

It happened again:

[CODE]bit length overflow
code 0 bits 6->7

bit length overflow
code 17 bits 6->7

bit length overflow
code 0 bits 6->7

bit length overflow
code 3 bits 6->7

bit length overflow
code 4 bits 6->7
code 0 bits 6->7

bit length overflow
code 17 bits 6->7

bit length overflow
code 10 bits 6->7
code 11 bits 6->7

bit length overflow
code 4 bits 6->7

bit length overflow
code 3 bits 6->7

bit length overflow
code 6 bits 6->7

bit length overflow
code 16 bits 5->6

bit length overflow
code 6 bits 7->6
code 16 bits 5->6

bit length overflow
code 17 bits 7->6
code 0 bits 5->6
[vinland] Client Eagle has timed out.
[vinland] :: Client disconnected: Eagle

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xb7513b70 (LWP 20990)]
0x4c71473a in ?? ()
(gdb) bt
#0 0x4c71473a in ?? ()
#1 0x08095bd7 in TServer::doMain (this=0x8188d70) at …/…/server/src/TServer.cpp:232
#2 0x080954b8 in TServer::operator() (this=0x8188d70) at …/…/server/src/TServer.cpp:116
#3 0x08088ac9 in boost::detail::thread_data<boost::reference_wrapper >::run (this=0x81c1db0) at /usr/include/boost/thread/detail/thread.hpp:81
#4 0xb7fcfc45 in thread_proxy () from /usr/lib/libboost_thread.so.1.40.0
#5 0xb7d37585 in start_thread () from /lib/i686/cmov/libpthread.so.0
#6 0xb7e1929e in clone () from /lib/i686/cmov/libc.so.6[/CODE]

I committed some new CSocket changes in hopes that it prevents the crash.

The weird thing is that this only happens when the server has been on for a while. what’s the bit length overflow about?

Smurf yeah!

The bit length overflow messages are from zlib. They are warning messages that get enabled when zlib is compiled with the DEBUG flag. They are harmless.

This is still happening. Apparently any kind of disconnection is now crashing the server.

O_o; Weird? That “any kind of disconnection” bug was fixed about 10 revisions ago.

I guess it’s back

or never was fixed… Nalin stated that the fixes could erase the problem. It was never confirmed that the problem was gone. Apparently it isn’t.

No no, that one specific bug was corrected. Must’ve been undone with the newest CSocket Patch.

I tested it on one of Joey’s server, that one crashed too from a normal disconnection.

Well, this specific problem has not ceased to exist during the whole time.

That’s a pretty vital bug if no one can even log out.
That’s why I had Bomber Arena specifically rolled back before (most of) the CSocket editing.
Only bug I’m afflicted with is sending a large file like an MP3 will sometimes disconnect people.