On Sun, 3 Apr 2005, Gisle Vanem wrote:

> Looks like some race condition... Then again I don't see how calling
> Curl_close() will end up calling destroy_thread_data(). Will it?

No, that's what I noticed as well. It should be a call to that function from
within Curl_disconnect(), since that is the function that closes a single
connection and frees data associated to a connectdata struct.

As I understand it, it should be called instead of the two Curl_safefree()
calls on lib/url.c:1509-1510 if libcurl is built with the threaded resolver.

Do you agree?

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html