Hi Gerardo, On Thu, 16 Sep 2004, Gerardo Ganis wrote: > > > Hi Matthias, > > Unfortunately I am not able to reproduce this occurence by myself > > ********************** > > In netx/src/TXNetConn.cxx: > > ==9343== 48 bytes in 2 blocks are definitely lost in loss record 360 of 563 > ==9343== at 0x4002BD79: __builtin_new (vg_replace_malloc.c:172) > ==9343== by 0x63C19B4D: TXPhyConnection::BuildXMessage(ESendRecvOptions, bool, > bool) (in > /afs/slac.stanford.edu/package/cernroot/vol21/cvs040908_1418/Linux24RH72_i386_gcc2953/lib/libNetx.so) > ==9343== by 0x63C10789: TXConnectionMgr::ReadMsg(short, ESendRecvOptions) (in > /afs/slac.stanford.edu/package/cernroot/vol21/cvs040908_1418/Linux24RH72_i386_gcc2953/lib/libNetx.so) > ==9343== by 0x63C1360E: TXNetConn::ReadPartialAnswer(XReqErrorType &, unsigned > int &, ClientRequest *, bool, void **, TXNetConn::EThreeStateReadHandler &) (in > /afs/slac.stanford.edu/package/cernroot/vol21/cvs040908_1418/Linux24RH72_i386_gcc2953/lib/libNetx.so) > ==9343== by 0x63C12737: TXNetConn::ClientServerCmd(ClientRequest *, void const > *, void **, void *, bool) (in > /afs/slac.stanford.edu/package/cernroot/vol21/cvs040908_1418/Linux24RH72_i386_gcc2953/lib/libNetx.so) > ==9343== by 0x63C12868: TXNetConn::SendGenCommand(ClientRequest *, void const *, > void **, void *, bool, char *, ServerResponseHeader *) (in > /afs/slac.stanford.edu/package/cernroot/vol21/cvs040908_1418/Linux24RH72_i386_gcc2953/lib/libNetx.so) > > > ************************** > > Is this something new to the ROOT version or was it there also with > XTNetFile (this part of the code was basically unchanged) ? > I did not check if it's there if I use XTNetFile instead of TXNetFile. I'll see if I can find out in the next hour (I'll have to leave then for today). Matthias > > >The SafeDelete(cmdrespMex) in SendGenCommand does not always do its job, or > > it's not always reached. > > Yes, it must be something like that, but by looking at the code it is not > obvious to see why the cleanup should not always be complete ... It must be > some tricky situation. > > Fabrizio: any idea? > > Cheers, Gerri >