This might be due to the port numbers being used with the debugserver. Right now we are grabbing a random port number in the valid user space port range and passing that down to debugserver as an argument for it to connect back up to lldb. The better way to do this, for local debug sessions, would be to open a port by binding to port zero and finding the port number, then passing an option to debugserver with the port number for debugserver to connect back to lldb.
On Aug 8, 2012, at 2:14 AM, Filipe Cabecinhas <[email protected]> wrote: > Hi all, > > I'm getting some Heisenbugs in the test-suite, sometimes. > > This is what their trace looks like: > --------------------------------- > runCmd: run > output: Process 99371 launched: > '/Users/filcab/dev/lldb-pristine/test/functionalities/data-formatter/data-formatter-objc/a.out' > (i386) > > > runCmd: thread list > runCmd failed! > error: Process must be launched. > > > --------------------------------- > > > > It only happens sometimes. Even with a fresh checkout. But it doesn't happen > every time. We must have a race condition somewhere. > Is anyone seeing this too? > > I'm on Mac OS X Lion (the latest). > > Regards, > > Filipe > > > _______________________________________________ > lldb-dev mailing list > [email protected] > http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev _______________________________________________ lldb-dev mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
