ARS keywords. These require that user->cursub and user->curdir are initialized
with the proper internal code associated with the current sub/dir of the user.
These ARS keywords were previously only supported in the chk_ar internal to the
Telnet/RLogin server (chk_ar.cpp).

unbaja.c:1830: warning: pointer targets in passing argument 1
of ?decompile_ars? differ in signedness

(Semaphore checks could have resulted in hung threads)

This should get 'er Fidde.

defaults, but I can't find anywhere to frob this.

Check that the endian macros work correctly.

is a hack anyways so I don't care all that much.

support it.  Therefore, check for the appropriate header, and make the
/dev/speaker code dependant on it.

session thread from being stopped/restarted every time the select() times
out for the semaphore check.

(or packing for FTP download).

Completely and utterly untested.  None of my Linux boxes even have
/usr/include/alsa on them, so I simply *can't* test this without yet
another Linux install.  If anyone cares about this, has ALSA working
correctly on their Linux box, is willing to give me a remote account that
can play sound via ALSA, and is willing to sit in IRC beside the system in
question to tell me if sounds are coming out, I can probobly get this
working.

Until then, use OSS emulation and everythign should be fine.

HTTP server optimizations.

0 = linger disabled (no time-out value)
non-zero = time-out value (in seconds), linger enabled

Close this end too.

recv() error, pass socket by reference rather by value, rely on a much more
natural termination sequence:
1) When terminate_server is set, close_request() always ends the session.
2) The session will now complete when terminate_server is set.
3) The output thread waits until the session is complete.

One int -> SOCKET fix.
etc.

active_clients could probobly be used for this, but it seems to work
differently...

[re]init.

occurances of these calls and report them (they're bugs).
Use the TERMINATE function in the strncpy version of SAFECOPY()
(does it need the int-typecast Deuce?).

(bugs found by Deuce ala Valgrind).

on the passed path...
For fexist() simply return TRUE if access()!=-1 && !isdir()
For fexistcase() *try* the access()/isdir() method first, then do the full
search.  Should make a reasonable difference.  Profileing suggests that
iniFileName() is quite slow due to the fexistcase() calls.  As the general
case is that the passed path is correct, this should speed up most calls
to fexist*().