On Mon, 2010-12-06 at 16:10 +0000, David Woodhouse wrote:
> Am I missing some way that the completion of the getaddrinfo thread is
> supposed to poke us to call back into libcurl? Some hook we've failed to
> register? Or were we supposed to *repeat* that 1ms timer until told
> otherwise?

Repeating the timer (by returning TRUE from our g_timeout callback
function so it keeps triggering every 1ms) makes it work, of course, but
I'm fairly sure that's not what we're supposed to do...

Note that prior to that 'fix', it isn't even timing out after 30 seconds
and aborting as it should; libcurl isn't calling our timer function for
a second time after that first 1ms request.

If I call curl_multi_timeout() from the g_timeout callback, it returns
29995. But libcurl never called our timer func to *tell* us that. Wasn't
it supposed to?

And if we *are* expected to poll for the completion of the async lookup
thread, shouldn't the time be a bit lower than that?

-- 
dwmw2
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html