Alex, That seems like it could be likely, but none of my code is spawning processes. However, when I run an strace on the process, I can see a child process getting created (assuming this is from ant?), and the deadlock from the full buffer could be happening there, though you would think this would be handled. One variance I thought of is the JVM version. Locally I am running Hotspot on OS X, but the deadlock occurs when running Open JDK on Ubuntu and CentOS. Thanks.
On Wed, Nov 9, 2011 at 10:02 AM, Alex Boisvert <[email protected]>wrote: > Wild guess: if you are somehow exec()'ing a process, you need to read the > stdout/stderr streams otherwise the child process will wait/hang until it > output buffers are consumed. (only happens when outstanding output exceeds > > X kilobytes) > > alex > > On Wed, Nov 9, 2011 at 4:48 PM, Allen Riddle <[email protected]> > wrote: > > > All, > > > > My build script periodically hangs on Linux systems. The last message I > > see on the log is "Trying to override old definition of datatype junit." > > This doesn't happen on OS X or Windows, but only in Ubuntu and CentOS. > I > > did an strace on it the process and what I see is a child process is > > getting created but nanosleep is repeatedly being called. I don't see > > anything my unit tests that could cause such an issue. Any ideas? > Thanks. > > >
