On Sun, Apr 28, 2019 at 03:14:47PM +0200, Tim Düsterhus wrote: > I'm afraid I still can't reproduce a crash, but I can confirm that I saw > some issues within valgrind in ssl_sock.c (so CC'ing Emeric here): > > ==22812== Conditional jump or move depends on uninitialised value(s) > ==22812== at 0x4172CE: ssl_sock_from_buf (ssl_sock.c:5848) > ==22812== by 0x538967: mux_pt_snd_buf (mux_pt.c:282) > ==22812== by 0x4FA0CA: si_cs_send (stream_interface.c:688) > ==22812== by 0x4FA85E: si_cs_process (stream_interface.c:570) > ==22812== by 0x539C63: mux_pt_wake (mux_pt.c:119) (...)
Note that valgrind says this all the time for traffic coming from SSL, it's even more visible when trying to debug H2 and was one of my motives for having H2 work over clear connections. I suspect that it sees that data were encoded using some keys derived from some randomness but this is a pure guess. Willy
