On Thu, Dec 3, 2009 at 12:16 AM, Hrvoje Niksic wrote: > Actually, Wget tries hard to implement the timeout in all of DNS, > connect(), and read(). If that doesn't work, it's a bug. > Can you specify how you test this, exactly? An strace dump of the > running Wget process might also help in debugging the problem.
Having written an example web server to demonstrate the problem, I see now that my mistake was simply not specifying --tries=1 along with -T - so wget was in fact timing-out but retrying much longer than the (individual) timeout I'd specified. This wasn't apparent because I only use -T in a script and always suppress output in scripts, but have been incorrectly using --quiet instead of --no-verbose. I'd simply expected the 'master' timeout (-T) to apply to the run of wget rather than each individual attempt, when the number of attempts isn't explicitly specified, and having a lot of retries with a timeout is a slightly counterproductive default (unless you're smart enough to look at the output properly, unlike me). I have, however, noticed that -nv displays an error message for a read timeout but not for a bad host name. Thanks. -- Lee Haywood.
