2011/3/9 Michel Dänzer <daen...@debian.org> > On Die, 2011-03-08 at 22:00 +0000, Thue Janus Kristensen wrote: > > > > With both drmDropMaster breakpointed in the second server, > > and drmSetMaster breakpointed in the first server, drmDropMaster is > > called first in the second server, and the first server doesn't reach > > drmSetMaster until after I continue the second server from the > > drmDropMaster breakpoint. No crash. > > > > > > With only drmSetMaster breakpointed in the first server, but the > > second server still running with gdb attached for consistency, there > > is no crash when I continue the first server from the drmSetMaster > > breakpoint. > > > > > > With both servers running inside gdb, but no breakpoints, it crashes. > > So, apparently the problem is that drmSetMaster runs 'too early'... > > How about if you only set a breakpoint on drmDropMaster in the second > server, and on hitting it just 'finish' the function before continuing. > Does the first server hit drmSetMaster before drmDropMaster has finished > in the second one? >
Will try when I get home. > BTW, did you also get any SIGUSR1 or other signals in either server > while doing this? If so, did you do the same for them as for SIGPIPE, to > prevent them from influencing the timing between drmDrop/SetMaster? Yes, there were a number of other signals I had to stop gdb from breaking on. Regards, Thue