On Mon, 8 Aug 2016, Marc Espie wrote:
> Well, unless someone can repair connect, here's a very ugly hack that
> should prevent SIGWINCH from fucking ftp up.
Hmm, there's a bug in connect(), but not the one you want.
Like nanosleep(), poll(), and select(), the connect() call never restarts
after a signal. For connect() it's _supposed_to_be_ because the signal
doesn't actually interrupt the underlying connect operation. To quote
POSIX:
If connect() is interrupted by a signal that is caught while
blocked waiting to establish a connection, connect() shall
fail and set errno to [EINTR], but the connection request
shall not be aborted, and the connection shall be established
asynchronously.
...but we don't implement that, and interrupting connect() takes the
socket out of the connecting state. Looks like FreeBSD and NetBSD
switched behaviors some time ago, maybe now's a good time in the cycle for
us to switch...
So for now, for pkg_add, your solution of blocking the signal across the
call is the best choice. Once the kernel changes behavior you would have
the option of instead handling EINTR by looping in a poll() for
writability; not sure if that would be simpler but it would allow WINCH
processing to occur during the connect...
Philip Guenther
