which broken handling EAGAIN/EINTR

rather than SSH_OPEN_UNKNOWN_CHANNEL_TYPE

from Andrzej Szombierski

--HG--
branch : nocircbuffer

and /**/ comments

--HG--
branch : fastopen

--HG--
branch : fastopen

if (connecting || ptys || x11) tos = LOWDELAY;
else if (tcp_forwards) tos = 0;
else tos = BULK;

TCP forwards could be either lowdelay or bulk, hence the default priority.

connection backlog, avoid check_close() for channels that haven't had IO

be more precise about maximum channel sizes

we're waiting for a key exchange.

This should hopefully fix the situation where "~." doesn't work to terminate a
client session when a laptop wakes up. The client will be stuck waiting for a
key exchange on a dead connection, so won't have read the escape character

Fixes the situation
$ ./dbclient root@1.2.3.4 'cat; echo foo'
^D
<no output>

Long-standing bug probably stemming from the awkwardly named
delete_channel() versus remove_channel()

- Fix build if ENABLE_SVR_REMOTETCPFWD is disabled but ENABLE_SVR_LOCALTCPFWD
  is enabled

  with a ChanType->closehandler()

--HG--
extra : convert_revision : 997e53cec7a9efb7413ac6e17b6be60a5597bd2e

code works

--HG--
extra : convert_revision : ff66e05ae040561110af70114bf83f11ed528f05

--HG--
extra : convert_revision : c92c7b4fef739b505d9842b45a393e1089097e3b

already sent a SSH_MSG_CHANNEL_CLOSE. This fixes a bug triggered by more
recent versions of putty with tcp fowarding and window adjustment request
messages.

--HG--
extra : convert_revision : 10b2f286b9886364db39dfbb4f8f46e49e345d87