The big remaining issue is defining *_EXPORTS where needed, which
looks rough from a quick glance.

The configure script was using uname -m as the build platform, but
Bullseye on RPi uses a 64-bit kernel (aarch64 uname -m) with a 32-bit
userland (arm gcc build host), which resulted in a lot of bad decsions
around JIT stuff.

With this change, we explicitly pass the build/host/target values as
taken from the compilers in question, which fixes the issue for Bullseye
(But may break mingw32?  Well we'll see how the pipes go).

This works on Buster at least... update the other config.guess file
the same as the top-level one, and always define JaegerSpew() so the
JIT stuff isn't broken.

About ten seconds in extract/patch, two minutes in build.

Allows caching in gitlab.

This may or may not actually work, and will certainly need a few
runs to build up the cache first.

Search in both /usr/local (Homebrew) and /opt/local (MacPorts) for
libraries.

Also, have the name of GNU make be an in option in CI so we can use
the Apple-provided make.

Now if you don't specify MIN_MAC_OSX_VERSION, it will use the
"current" version of macOS as the target.  "Current" is defined as
X.0 where the major version (X) is greater than 10 and X.Y when
major version is 10.

This allows building for a specific target, but uses the newest
libraries if you don't specify.

For Synchronet "stuff", we can trivially change this, but some work
needs to be done to enable Spidermonkey and Cryptlib to pick up the
chosen value.

JS should be fairly simple since it's a configure argument.  Cryptlib
is a bit trickier since it's in a patch.

It builds and appears to run... not actually tested.

No real purpose to keeping the default min version at 10.9.

Oops, missed this with the cl32.dll commit.

<Deuce> Ok, Win32 will need a new cryptlib with these latest patches.

Deuceth commandeth, Robeth doeth

If the SSH_ANYAUTH option is set, the server will accept the "none"
authentication method, and not even suggest the client send a
password or public key.

The client must still send a user ID, but we just ignore it
completely and don't even log it.

readHeaderFixed() already assumes the same header pointer will be
passed in next call if the read is incomplete.

We've patched this thing enough to have a different identifier.

By default, Cryptlib uses the type of whatever channel happens to
be currently selected, so if the server has sent some data, that
will be shell.  Instead, look up the new channel and use that type
info.

Ugh.

Previously, it was read into a local variable, and in the case of a
partial read, everything would go to hell and SSH would hang.

"Luckily" this was very hard to trigger.

This should work exactly the same, but is more correct.  It's possible
that in the future, packet data could be added to the buffer after
receiveBufPos, and we could know that no more is required.

The patch that sets the read timeout every time for getData() was
flawed, because getData() itself will sometimes set the read timeout
for the next call.

The patch was intended to ensure that the write timeout never
clobbers the read timeout, so now we track if getData() explicitly
set the timeout and if it did, we don't reset it.