Follow-up Comment #1, bug #49278 (project hurd):
I suspect this is caused by a no-more-senders notification on the receive
right of the filesystem root port, to which the client sent the message. If
you add sleep(2) after the mach_msg call in the client, so that it doesn't
deallocate the send right before mach_msg times out in the server, then the
EINTR doesn't happen on the first call. (I didn't test more calls.) The
no-more-senders notification would be requested because trivfs_S_fsys_getroot
calls ports_get_right. The notification is handled in
ports_do_mach_notify_no_senders.
If the EINTR only happens on the first call, that seems wrong.
_______________________________________________________
Reply to this item at:
<http://savannah.gnu.org/bugs/?49278>
_______________________________________________
Message sent via/by Savannah
http://savannah.gnu.org/
