Beautify a lot of the remaining strange closing brace style.
Use FREE_AND_NULL() where appropriate.
Log SMB error strings for most common smblib function failures.
Log debug output in the areas.bbs alter function (help isolate Joe's crash).

requests - should handle more unexpected echolist lines (e.g. extra long, blank
or whatever). Hopefully resolves crashes in strdup() Joe is seeing.

with an invalid file descriptor (e.g. -1), don't attempt to close it.

SAFECOPY() ain't safe for pointers... use strdup() since that's what we
want anyway.

variable (with various race conditions). This resolves a long LONG standing
bug that caused crashes when shutting down or recycling servers/services
while one or more servers/services were in use.
This change requires every thread which calls into libjs to have its own JS
runtime.

skipped/ignored: import_netmail() moves the file position to the end of the
message but returns non-zero (-2) when encountering an AreaFix message
and the calling code assumes a non-zero return value indicates an error and
the file position is somewhere *before* the end of the message (the NUL char).
Resolved by saving the current file position before calling import_netmail()
and restoring it after, so it matters not what improt_netmail does with the
file pointer or what the return value is (in this case).
Thanks for the report, Joe!

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.

Whatever made us do that seems to not be happening with --std=c99.

(alphabetically) since pulling/merging from multiple lists (add'l echolists
and the areas.bbs) often creates a confusing mix of sorted and unsorted area
names.

Fix %LIST AreaFix response for ELIST_ONLY mode.
Fix %UNLINKED AreaFix response for non-ELIST_ONLY mode (include echolists).
Fix alter_areas() crash introduced in rev. 236 (j was not valid here).

rev 236 (this this one, access===D).

it resolves Joe's errors with areas.bbs modifications via AreaFix on Windows.

Yay!

Better/more log output when receiving AreaFix requests for echo/area lists.

ELIST_ONLY option: include unlinked echoes from "additional echolists" in the
response (rather than just "None.").
Created and used area_is_linked() function for the commonly copy/pasted code
for checking if a specific node address is already linked with a specific area.

application" error message (when run on XP) and reportedly will allow target
executables to run on non-SEE (pre-Pentium III) CPUs. Seriously, anyone really
running Windows XP on a Pentium II today? I guess it's possible <shrug> and we
really get little benefit from SSE in Synchronet.

I couldn't get Windows 2000 running in Hyper-V, so I guess Windows 2000 is
now officially unsupported by Synchronet (and long unsupported by Microsoft).

Thanks to Android8675 for the bug report and Rushfan for the solution tip.

(don't assume SSE support in target CPU).

number of hash sets (don't let these files grow indefinitely).

build error) - conflict with definition form vadefs.h.

loop immediately upon socket receive error (e.g. 54, ECONNRESET).

better and has a higher alchohol content.

Use a linked list to track runtimes, thereby allowing an unlimited number
of them.

These functions are still needed since Spidermonkey crashes if more than
one runtime is created simultaneously.

initialization due to no service sockets being successfully bound. The crash
was in cleanup(), while checking the 'theads_pending_start' protected-int,
which had not been created/initialized yet. The fix is to create/init the
threads_pending_start protected-int before any possible call to cleanup().

also clear the associated counts (sizes) for these arrays, eliminating some
potential crashes while dereferencing these arrays after freeing (e.g. during
shutdown or recycle of sbbs).

so the runtime "pool" concept might need to go away. For now, we'll just change
the RT_TYPE to RT_UNIQUE (no sharing of runtimes, using more memory) and
double the size of the RT pool from 128 to 256. Without sharing, the pool may
not make sense unless there is a significant time penalty when first creating
RTs (in which case reusing RTs will be faster).