Hi Maksim, On Wed, Jun 05, 2019 at 09:59:29AM +0200, Willy Tarreau wrote: > I'm still trying to fix this mess, it's not trivial at all and may > even require an extra -dev. So expect some updates later.
Just a quick note to let you know that Olivier just fixed a directly related issue in 2.0-dev and I'm fairly confident it fixes your case as well. In short, the low-level mux-h1 believed the connection was completed before the handshakes were finished. Thus, it signaled to the upper layers that they could start to work too early. For me it was causing some "SH--" fields instead of "sC--", and that's fixed now. Feel free to re-run a test. Willy