On Mon, 30 Jul 2007, Greg Morse wrote:

> I have another fix that may be more general than the other one:
>
> infof(data, "Keep sending data to get tossed away!\n");if (conn->writesockfd
> != CURL_SOCKET_BAD) k->keepon |= KEEP_WRITE;
>
> Explanation: turning write on with a bad write sockfd will cause an infinite
> loop when the read finishes since the write has already completed.

That is certainly a small and neat fix, but it does give me the feeling that
this is just covering the effects of a bug rather than the source of it. I
mean, if the rewinding is done it is because it wants to resend data and how
come writesockfd can be bad in this place in the code if it wants to send
data?

It feels like libcurl should've bailed out already when the socket turned
bad...

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