P4DTI issue job000334

TitleIntermittent socket errors from TeamTrack 5.0 when using a database upgraded from TeamTrack 4.5
Assigned userGareth Rees
DescriptionWhen using a database that was created in TeamTrack 4.5 and then upgraded to TeamTrack 5.0, there are socket failures. The P4DTI hangs for a minute or so, then gets a socket error from the TeamShare API. The failures are intermittent: a test may succeed several times, then fail.
The error message returned from the TeamShare API is as follows:
SOCKET_READ_ERROR: Socket error.
One reason might be database needs up upgrading.
Check event view for complete error message.
AnalysisEach failure is preceded by a pause of around a minute or so. If I stop the execution during this pause, I get a stack trace like this:
NTDLL! 77f682db()
MSAFD! 776642b6()
WS2_32! 776b8308()
CSocket::Receive(void * 0x017449c0, int 8196, int 0) line 157
TSSocket::FillBuffer() line 583 + 29 bytes
TSSocket::ReadChar() line 545 + 8 bytes
TSSocket::ReceiveInt(int * 0x0012dcbc) line 600 + 8 bytes
TSServer::Send(TSSocket * 0x09a4f670, const char * 0x0182b2c0, int 13465) line 485 + 12 bytes
TSServer::UpdateRecord(TSRecord * 0x03e6c5f0, TSRecord * 0x03e6c5f0) line 4198 + 23 bytes
teamtrack_record_update(_object * 0x08285300, _object * 0x0082312c) line 316 + 28 bytes
PYTHON20! 002428e8()
TEAMTRACK! struct _typeobject teamtrack_record_type address 0x0150c178

The API always hangs in CSocket::Receive, but it isn't always TSServer::UpdateRecord at the bottom; the problem can also occur in TSServer::ReadRecordListWithWhere.
When CSocket::Receive eventually returns, it returns SOCKET_ERROR. The error code returned by WASGetLastError() is WASECONNRESET (connection reset). This strongly suggests that the problem is to do with the server: this could happen if the server failed in some way so that there was nothing on the other end of the socket.

This occurs when using a database created in TeamTrack 4.5 and then converted to TeamTrack 5. The offending database is <http://www.ravenbrook.com/project/p4dt...c/2001-05-31/job000326/tmtrack-4509.mdb>. This does not occur when using a database newly created in TeamTrack 5. So perhaps the database is corrupt in some way.
GDR 2001-06-30: TeamShare found a defect in the API that may be the cause of this [1].
How foundmanual_test
Evidence[1] <http://info.ravenbrook.com/mail/2001/06/29/17-51-25/0.txt>
Observed in1.1.1
Created byGareth Rees
Created on2001-06-12 15:15:11
Last modified byGareth Rees
Last modified on2001-12-10 19:45:32
History2001-06-12 GDR Created.
2001-06-30 GDR Referred to defect found by TeamShare.


Change Effect Date User Description
13767 closed 2001-07-03 12:36:09 Gareth Rees Added support for TeamTrack 5.0:
Split code/tsapi/... into code/tsapi-4.5/... and code/tsapi-5.0/... and split the Developer Studio project teamtrack.dsp into teamtrack-4.5.dsp and teamtrack-5.0.dspso that I can build two copies of the Python interface to TeamTrack, one for TeamTrack 4.5 and one for TeamTrack 5.0.
Updated teamtrack-module.cpp so that it has two entry points (one for each module).
Added teamtrack_version configuration parameter.
Updated design, Administrator Guide and release notes.