Re: FWD: Re: xfce on the Hurd
There have been some other reports of flakiness in pflocal, which is the
server the implements pipes and unix-domain sockets. That is a definite
possibility. The thing to do is watch the pflocal process with ps. If it
dies and a new one gets started, then all old pipe/socket file descriptors
will become invalid. You can run "gdb /hurd/pflocal" and attach to the
pflocal process (just make sure your gdb session isn't doing anything that
wants to use pipes or unix-domain sockets); then cont and hopefully you
will see something helpful in gdb when pflocal crashes (if that is what's
happening).
If pflocal is hitting an assert or something, then it might try to print
something out on stderr that would be helpful to see. If you manually
restart the translator with settrans -a, then it will have a stderr
(whatever settrans's stderr was) so you can watch for it. Run:
settrans -ag /servers/socket/1 /hurd/pflocal
Reply to: