On Thu, 27 Jan 2011, �������� ������� wrote:

> Continuing the work for the custom resolver, I've made some simple test for
> the situation which nervous me: what the library does if the cancel and
> destroy happens for the request which is in the WAITRESOLVE state?

Why are you nervous for that? Shouldn't it just shut everything down nice and
cleanly?

> This test is not deterministic enough to include it into the test suite,
> just because of different network timeouts and bandwidth, but has
> predictable behavior looking to output generated by valgrind tool starting
> the test. The test code itself is applied. The test has meaning only for
> asynchronous resolver of different natures, either threading, or c-ares.

That's a great test! Yes I could reproduce the problem easily using that and
I've pushed a fix (and a follow-up fix as I also spotted a memory leak) for it
now. Thanks!

> That's what I meant when told about complexity ...

We don't seek nor defend complexity. We need functionality and and we write
code for that and yes we have bugs from time to time. If you can make the code
less complex and maintain the funcionality then I'm sure a bunch of us will
praise your work!

Are you suggesting your solution won't have bugs?

-- 
  / daniel.haxx.se

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2011-01-27