@anonymouspage reported it broken in !212.

So it can be reused for other projects more easily

1) Keep a rectangle updated per-screen rather than regenerate each time
2) Strip palette info when putting pixels into rectangles rather than
   during scaling
3) Tighten up the screen locks a bit
4) Don't require a full resend of both screens on an update request
5) Only force a redraw for cursor movement when the cursor is visible
   (And force it whenever the cursor changes)
6) Avoid doubles in interpolation
7) Heavily optimize interpolate_height()
   interpolate_width() likely doesn't need it because it's generally not
   used and also it reads from the next pixel in memory making the
   prefetchers job easier.
8) Fix some memory-leak-on-error issues
9) For ARGB8 XImages, manipulate the data directly rather than through
   XPutPixel()

At this point, the scaling and X11 output time is heavily dominated by
cache misses.  The only really effective way to reduce this hit is to
spread the work across all the L3 caches in the system or move it into
the GPU.

With the latest updates, at the SyncTERM menu, over 90% of the time is
spent in the rendering pipeline, and over 90% of that time is spent
thrashing the caches... the only real easy win left is vectorizing, but
that's highly compiler specific.

To that end, I've switched to -O3 for release builds.  There was a comment
that -finline-functions broke Baja "badly", but that's clearly false since
-f-inline-functions has been part of -O2 for quite a while now, and Baja
doesn't seem any more broken that it ever was.

Fixes *nix builds of Win32 programs.

Rather than depending on _WIN32 which is silly, make it clear what
exactly is being done and why.

Note that the spy socket stuff and some external stuff on *nix have
had the select() implementation removed.  To support using select()
on !Win32 will require de-refactoring that's not really worth doing.
Instead, just use #error and reference the commit where the old
code can be found.

MinGW32 is getting long in the tooth and is missing a lot of modern
Windows features as well as having broken headers.  Most people will
be using MinGW-w64 at this point, so add support for it.  Once I
ensure SyncTERM works properly with it, MinGW32 support will be
discontinued.

I suspect this will impact exactly zero people since the reason this
exists is to build the Win32 versions of SyncTERM on FreeBSD.

Changes:
- Explicitly request 32-bit Windows output
- Detect the string "mingw32" anywhere in the hardware description
- Explicitly link with libuuid
- Add a terrible hack to syncterm.c to block wincrypt.h

Hope to address error reported by rjwboys via IRC when using
gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.12)
 from ansiterm.cpp:22   /user/include/c++/5/bits/c++0x_warning.h:32:2
error #error this file requires compiler and library support for the ISO C++ 2011 standard. This support must be enabled with the -std=c++11 or -std=gnu++11 compiler options.

Needed for CIOLIB builds.

so define HAS_STDINT_H

The Linux lock() implementation in xpdev/filewrap.c requires that
_FILE_OFFSET_BITS is set to 64 or else OFD (sane) locks are not
used on Linux, defaulting back to the crazy per-process file region
locking of Linux of old. This was done in this commit:
http://cvs.synchro.net/commitlog.ssjs?99999#39639
because:
https://patchwork.kernel.org/patch/9289177/

but on 64-bit Linux, large file support is assumed and you don't
actually need to define _FILE_OFFFSET_BITS, we're peforming this
check for 32-bit Linuxes. Perhaps a check to see if sizeof(off_t)
== 8 would have been better. However, this change enables OFD
Locks for *all* flavors of Linux (not just 64-bit) - so that's
better.

Why do we care? Because using region locks on the same file in
a multi-threaded program doesn't really work right on Linux without
using OFD Locks. I saw this problem with receiving multiple
simultaneous emails in the mailserver and getting smb_locksmbhdr()
errors/failures when trying to save the message data or headers
of the one or more messages every time GitLab would send email
out to those whoe have opted-in to received email notifications
upon gitlab.synchro.net activity.

about redefining _XOPEN_SOURCE on every command-line (because pkt-config
is also defining it, to 600).

SDL audio should work from any thread, and on the platforms that care,
SDL video is coddled appropriately.
The only question is around SDL_Quit() in an audio-only build, but we
can punt on that since the default builds will never do that... the only
thing I'm aware of that could be broken that way is tone, which nobody
likes anyway.

any files), so use "del /q" instead.

deprecated

Prepare for upcoming "hash" library.

about undeclared functions (vasprintf and pthread_setname_np) when building
sbbs3.

(points to lib and include files in 3rdp/win32.release/sdl).

SBBSCTRL is set. Also changed the library versioning from 3.16 to 3.17.

in WinSock (wsock32.lib), but might add the IP Helper lib or other cfg
properties later.

on Windows XP now.  At the very least, it will now require MSVCR80.DLL.

warnings when a string or character literal contains invalid UTF-8.

is 'clang'

This will also enable SVID stuff, but that appears to be unavoidable.

See the following URIs:
https://lwn.net/Articles/590381/
https://sourceware.org/glibc/wiki/Release/2.20#Deprecation_of__BSD_SOURCE_and__SVID_SOURCE_feature_macros
http://www.gnu.org/software/libc/manual/html_node/Feature-Test-Macros.html#index-_005fDEFAULT_005fSOURCE

be set and read using the POSIX macros and related (currently Linux-only).

of the compiler (icc, gcc, and clang supported).

headers are rife with.  Re-enables cross-building with mingw32 (at least for
SyncTERM).

POSIX for stdio and more.
BSD because BSD is awesome.
X/Open for timezone.

except for the ISO C ones are disabled.  Since we use POSIX, BSD, and SYSV
APIs, define _GNU_SOURCE which re-enables them all.

Only for Linux so far.

Yay!