Hi, the behavior of m->ReadRaw is supposed, in case of error, to build a message with the error information inside, to be passed to the higher level for proper processing. So, the only error expected from ReadRaw should be the impossibility to create a Message object to be enqueued, which seems not the case. May i have the piece of logfile explaining the error that the lower levels encountered? This would help me in finding the problem. Fabrizio Peter Elmer wrote: > Hi Fabrizio, > > It looks like this was never dealt with. Is there a useful way to check > the return code and recover? > > Pete > > On Wed, Dec 15, 2004 at 03:10:35PM -0500, Gregory J. Sharp wrote: > >>The latest version of xrootd (20041214) is causing us some headaches. >>We are getting repeated disconnections twixt client and server. This >>may be our fault, but I don't think so. Anyway, in the process it >>revealed a serious problem that is not our fault. >> >>In XrdClientPhyConnection.cc the method BuildMessage has the call (at >>approx. line 402) >> >> m->ReadRaw(this); >> >>ReadRaw returns an error code (throwing an exception might have been >>better) if there was a socket error of some sort. This error is not >>checked for and so it loops in(de)finitely, printing error messages in >>the logs, but not actually dying or trying to reconnect, or anything >>user friendly. >> >>I will get back to you about the problem of the disconnects if I figure >>out why they are happening. >> >>(Does anybody read Henry Spencer's "The 10 Commandments of C >>Programming" anymore? Yea! Verily!!) >> > > > > > ------------------------------------------------------------------------- > Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) 767-4644 > Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland > -------------------------------------------------------------------------