I think that is a bug in term. Its io_select fails with EINVAL when asked
to test for the "exception" (SELECT_URG) condition. This is clearly
wrong--it's always valid to inquire about a condition that will never be
true. The implementation of select in libc treats any descriptor whose
io_select RPC returns an error as having all three flavors of condition
ready, with the idea that if the program responds to this by making some
i/o call on that fd, it should see whatever the pathological error was.
That latter behavior might be a bit questionable, though I'm not sure what
else select might more usefully do instead. But it only arises at all
because of term's intransigence.
I'm checking in a change to term. In the meantime, it should be reasonably
safe as a workaround to just pass NULL in place of &efds to select in ppp.
_______________________________________________
Bug-hurd mailing list
[EMAIL PROTECTED]
http://mail.gnu.org/mailman/listinfo/bug-hurd
