On Wed, Feb 3, 2016 at 3:01 PM, Yann Ylavic <ylavic....@gmail.com> wrote: > On Wed, Feb 3, 2016 at 12:34 PM, Yann Ylavic <ylavic....@gmail.com> wrote: >> >> So I don't see why the response is not flushed on the network before >> entering READ_REQUEST_LINE or CHECK_REQUEST_LINE_READABLE states. > > OK, I can now reproduce with mpm_event.
Actually no. I can reproduce the pending data not being flushed by a FLUSH bucket (which is expected with event), but still event flushes them by calling the core output filter directly, and it works here... > > Joachim, looking into this from my own environment (no need to apply > patches yourself now ;) So unfortunately there seem to be something special in your environment which prevents non-blocking writes to go out (not talking about the weird TLS handshake played by the client according to the pcap). Could you try Rüdiger's (logging) patch? And maybe take network traces on the server side? Thanks, Yann.
