Re: [reSIProcate] regression: repro working on x86_64, not i686 though?
Daniel sent a log offlist, here are the relevant error lines:
STACK | 20120412-144017.206 | repro | RESIP:TRANSACTION | 3038710640 | TransactionState.cxx:168 | No matching transaction for SipResp: 407 tid=d5d50ed3B6212E18 cseq=1 REGISTER / 1 from(tu)
INFO | 20120412-144017.206 | repro | RESIP:TRANSACTION | 3038710640 | TransactionState.cxx:364 | discarding stray response: SipResp: 407 tid=d5d50ed3B6212E18 cseq=1 REGISTER / 1 from(tu)
...shutdown...
WARNING | 20120412-144032.337 | repro | RESIP:TRANSACTION | 3064710864 | TransactionController.cxx:72 | On shutdown, there are Server TransactionStates remaining!
DEBUG | 20120412-144032.337 | repro | RESIP:TRANSACTION | 3064710864 | TransactionMap.cxx:18 | d5d50ed3B6212E18 -> 0x8457380: tid=d5d50ed3B6212E18 [ ServerNonInvite/Trying reliable target=[ V4
188.62.165.55:33860 TLS target domain=unspecified mFlowKey=23 ]]
INFO | 20120412-144032.338 | repro | RESIP:TRANSACTION | 3064710864 | TransactionMap.cxx:71 | Couldn't find d5d50ed3B6212E18 to remove
It seems that there are issues finding the transaction in transaction map. I suspect your problem is somehow related to the new Data::caseInsensitiveTokenHash implementation that was pulled in the Tekelec performance branch merge, and how it is used in TransactionMap.hxx - since it seems that TransactionMap::find is failing. Something you could try out is rutil/test/testData.cxx, since it has some new test cases for the new token hash stuff.
Scott
On Thu, Apr 12, 2012 at 8:54 AM, Scott Godin
<sgodin@xxxxxxxxxxxxxxx> wrote:
Hi Daniel,
Can you post more of the resip logs for the 32-bit machine at STACK level?
Scott
On Thu, Apr 12, 2012 at 7:11 AM, Daniel Pocock
<daniel@xxxxxxxxxxxxx> wrote:
I found my problem with repro does not occur on a 64 bit machine (both
machines are Debian squeeze, same libs, etc)
I've then tried the following:
- r9447 from SVN (pre-autotools) - same problem
- r9447 from SVN, using resip-ares - same problem
- resip 1.7 release tarball, using resip-ares - working fine on i386
Therefore, I believe there is a regression between 1.7 and r9447, and it
is not autotools related.
The phones are configured identically: I use a TCP port redirection from
the 32 bit machine when I want to test on the 64 bit machine.
On 64 bit, the 407 response is sent back successfully over the TLS
connection:
DEBUG | 20120412-032858.427 | repro | RESIP | 140333551826688 |
SipStack.cxx:550 | SEND: SipResp: 407 tid=e7b71622E71A9A47 cseq=1
REGISTER / 1 from(tu)
INFO | 20120412-032858.427 | repro | RESIP:DUM | 140333551826688 |
ServerAuthManager.cxx:45 | ServerAuth challenged request SipReq:
REGISTER pbx.example.org tid=e7b71622E71A9A47 cseq=1 REGISTER
contact=7001@192.168.1.125:57652 / 1 from(wire) tlsd=pbx.example.org
STACK | 20120412-032858.427 | repro | RESIP:TRANSACTION |
140333593790208 | TransactionState.cxx:706 | Found matching transaction
for SipResp: 407 tid=e7b71622E71A9A47 cseq=1 REGISTER / 1 from(tu) ->
tid=e7b71622E71A9A47 [ ServerNonInvite/Trying reliable target=[ V4
127.0.0.1:50317 TLS target domain=unspecified mFlowKey=23 ]]
On the 32 bit system, the 407 response is never sent:
DEBUG | 20120412-023212.223 | repro | RESIP | 2997517168 |
SipStack.cxx:550 | SEND: SipResp: 407 tid=8170d69d528F4676 cseq=1
REGISTER / 1 from(tu)
INFO | 20120412-023212.223 | repro | RESIP:DUM | 2997517168 |
ServerAuthManager.cxx:45 | ServerAuth challenged request SipReq:
REGISTER pbx.example.org tid=8170d69d528F4676 cseq=1 REGISTER
contact=test2@192.168.1.114:38983 / 1 from(wire) tlsd=pbx.example.org
STACK | 20120412-023212.223 | repro | RESIP:TRANSACTION | 3039480688 |
TransactionState.cxx:168 | No matching transaction for SipResp: 407
tid=8170d69d528F4676 cseq=1 REGISTER / 1 from(tu)
INFO | 20120412-023212.223 | repro | RESIP:TRANSACTION | 3039480688 |
TransactionState.cxx:364 | discarding stray response: SipResp: 407
tid=8170d69d528F4676 cseq=1 REGISTER / 1 from(tu)
_______________________________________________
resiprocate-devel mailing list
resiprocate-devel@xxxxxxxxxxxxxxx
https://list.resiprocate.org/mailman/listinfo/resiprocate-devel