On Wed, 1 Sep 2004 12:24:02 -0700, Michael Stenner wrote:
> On Wed, Sep 01, 2004 at 12:13:07PM -0700, Paul Dickson wrote:
> > In my case, my network is connected via a PPP dialup connection. As long
> > as the pppd is running, my system doesn't receive ICMP messages about the
> > lack of a connection (pppd queues packets until a connection is made).
> > Whenever I reconnect, my ISP gives me a new IP address.
> >
> > Whenever the dialup connection drops, yum transfers end up hanging,
> > sometimes for hours until I interrupt them.
> >
> > This is why I added the wget option to my copy of yum. Wget works around
> > the problems I have with downloading.
>
> I see. That's an icky problem. Do you know how wget deals with it?
> Does it just have some sort of high-level timeout? (ie, not tcp
> timeouts, but its own)
>
I believe this might be how wget does it (from wget's connect.c):
| /* Read at most LEN bytes from FD, storing them to BUF. This is
| virtually the same as read(), but takes care of EINTR braindamage
| and uses select() to timeout the stale connections (a connection is
| stale if more than OPT.READ_TIMEOUT time is spent in select() or
| read()). */
|
| int
| iread (int fd, char *buf, int len)
| {
| int res;
|
| #ifdef HAVE_SELECT
| if (opt.read_timeout)
| if (select_fd (fd, opt.read_timeout, 0) <= 0)
| return -1;
| #endif
| do
| res = READ (fd, buf, len);
| while (res == -1 && errno == EINTR);
|
| return res;
| }
The file connect.c has a lot of functions with timeout options.
Most seem to use the a function called run_with_timeout at the end of
util.c, see the lengthy comment just before the function.
-Paul
