Hi Andy, On Mon, Aug 15, 2005 at 07:57:38PM -0700, Andrew Hanushevsky wrote: > My prediliction is that it's a timing issue since the timeouts are > relatively short. It's also possible that valgrind has issues with a > heavily threaded program. I know that gdb fails more often than not when > confronted with a running xrootd. The client won't be heavily threaded, though, at least not like the server. One of us should give it a try to see if we can repeat it. > Now, as for the attached file. Could you tar the file up and then attach > it? The reason is that the saved attachment gets linend characters > appended to each line which makes diff and most anythig else I try to use > on the file quite ineffective. Thanks. I imagine it is the M$ MUA that you are using that screws up the attachment somehow for you. I've copied it here at SLAC: ~elmer/gs_20050815_XrdOucPthread.cc Does that look okay? Pete > On Mon, 15 Aug 2005, Gregory J. Sharp wrote: > > > Running valgrind on our data analysis application while accessing > > xrootd found some problems in the error handling in XrdClient and a > > few minor nits in XrdOuc. This message contains the suggested patches > > to XrdOuc/XrdOucPthread.cc. The next email will contain a raft of > > patches for XrdClient. > > > > One thing that troubles me is that by using valgrind, the protocol > > fails completely, apparently because of timing issues. Connections > > get torn down while we are in the poll call, waiting for a response > > from the xrootd server. My best guess is that this is either a bug in > > the protocol or a timeout issue. (My second-best guess is that > > valgrind is broken, but since it works well enough most of the time, > > I lean more towards my best guess.) > > > > ------------------------------------------------------------------------- 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 -------------------------------------------------------------------------