Daniel Stenberg wrote:
> On Sat, 22 Aug 2009, Daniel Stenberg wrote:
>
>>> http://curl.haxx.se/mail/lib-2009-05/0173.html
>>
>> Oh. I can't recall I've seen that (and if I did I've forgotten about
>> it)... I'll check it closer within a few days.
>
> Luke, have you been running code with this patch since back then and
> things work fine?
>
> If so, what about the two FIXME comments in your patch? I'm especially
> interested in the second one, as that seems like perhaps it does a
> little too much. Does it ever trigger?

The patch is pretty similar to what I posted in ignorance of the patch,
it works fine, passes all tests, but it is still plain wrong.
Yes, it fixes behaviour for one case, but does not solve the real problem.

The real problem is the difference in splay queued timeouts and the real
timeout. *My* very own approach would be to unify splay & real-timeouts
instead of fixing the behaviour of the broken idea.

Curl's got rather tight release cycles, how to deal with major changes
linke this?

Markus
Received on 2009-08-27