On Thu, 24 May 2018, Sean Miller wrote:
Your educated guess was correct. Log with patch applied to 7.55.1 is
attached. Original observed delay is gone.
What delays remain are probably network latency.
Awesome! I think the fix could be slightly more polished (since it now sets
the socket to
On Fri, 18 May 2018, Sean Miller wrote:
Does it seem like that could be what happens in your case? If so, I'll
write up a patch you can test...
I'd be happy to test a patch. I've been poking around in the code a little
myself, but I'm unfamiliar with it.
I've reproduced this on
> I think you've stepped on a libcurl bug!
Uh oh. I'll have to clean off my shoe.
> Are you doing HTTPS through a HTTP proxy by any chance? That's what the
> PROTOCONNECT state is for basically.
Yes, almost everything I do in this large private cloud goes through proxies.
> Does it seem like
On Thu, 17 May 2018, Sean Miller wrote:
1) In the WAITCONNECT state, libcurl calls multi_timer_cb(200) and registers
interest in writing.
2) The FSM transitions first to SENDPROTOCONNECT and then to PROTOCONNECT.
3) Libcurl calls remsock() through sock_cb().
4) ... Nothing happens until the
I added simple logging and a debug function to the following Boost ASIO-based
example program:
https://curl.haxx.se/libcurl/c/asiohiper.html
Summary:
1) In the WAITCONNECT state, libcurl calls multi_timer_cb(200) and registers
interest in writing.
2) The FSM transitions first to