of default style elements.

- ini_key_len
- ini_key_prefix
- ini_section_separator
- ini_value_separator
- ini_bit_separator
- ini_literal_separator

and no actual observed problem.

malloc'd strings.

elements (member variables) would be changed to point to a string constant.
Don't do that - instead just use local stack variables for those style
element default values. This leads the way to constifying the style arguments
in the future and allows the caller to use non-NULL style elements as indicator
of something that should be freed if the caller allocated some memory for it.

I have a new use for it in the v3 Terminal Server, it now follows the
"NO_DOS" option in the [bbs] section of the sbbs.ini file. So if you have
multiple Win32 instances of sbbs (like I do), but only some of them
support DOS programs, you can now use this ARS keyword to enable/disable
access to a program based on the existance of DOS support. For example,
I have some 16-bit DOS message editors where I had the ARS set to
"WIN32", but on my 64-bit Windows instance of sbbs (vert.synchro.net),
16-bit DOS programs aren't supported while on my 32-bit Windows instance
(bbs.synchro.net) they are. So I needed a good run-time check for whether
DOS programs were supported or not and this old ARS keyword seemed the
right fit. For Linux builds, this keyword will only evaluate to true if
built with the USE_DOSEMU option.
This is only true of the Terminal Server - use of "DOS" ARS keyword in
other servers and services will still just evaluate to "false".

"short name" (e.g. derived from the "AREA tag" from FILEGATE.ZXX) is the same
as a pre-existing dir, just over-write that dir - don't create a new one.

previous commit: don't clear the line counter or call pause() - which will
do nothing - if the console abort flag is set.

would not always auto-redisplay menu files when the user hits 'Q' (or 'N'
or Ctrl-C) at auto-pause prompts:

When displaying a "message" (e.g. file), if the console abort flag is set
before a single line has been displayed, clear the abort flag. This most
likely means a pause (hit a key) prompt was automatically displayed due to
a clear screen code in the message/file and the user hit 'Q' or 'N' or
Ctrl-C at the prompt. In this special case, don't treat that as an
"abort message" command and keep displaying the message/file.

Once one or more lines in the message/file have been displayed, then the
behavior is the same as before: a console abort (even as a result of an
auto-pause prompt key) will stop the display of the current message/file.

 term 0087 TLS ERROR 'Unexpected <Unknown type> (24) packet, expected application_data (23)' (-1) popping data
 message repeated 492 times: [ term 0087 TLS ERROR 'Unexpected <Unknown type> (24) packet, expected application_data (23)' (-1) popping data]

When using TLS with a JS Socket object, if there was any kind of data error,
the recvline() method would return a blank string rather than null/undefined.
nntpservice.js just loops when it receives a blank string, so this caused an
infinite loop (with disk-filling error log messages).

First change: if no data has been received (i == 0) and there's any kind of
receive error or timeout or disconnection, just return null. And not undefined,
but null (!) like in v3.15 (before the great JS engine update of 2000-mumble).
Also, there appeared to be a JS_RESUMEREQUEST call missing in the TLS error
return case - so that's another bug fixed.
Commented on the magic return values for js_sock_read_check()
and js_socket_recv().
Simplified js_sock_read_check() return value a tad: let the caller decide if
they want to do something special based on the value of 'i'.
Added some comments to make this code more readable.

We are now no longer treating the different error return values (0 and -1) from
js_socket_recv() special in this function, but we dont' treat them special in
any of the other calls in this file/object either, so that seems to be the
norm.

Deuce really wanted to pass 1000 as a value here (use 1 instead). I don't
know if this was an observable problem or not, but it certainly *looks*
like a bug.

first white-space character encountered.

'content-disposition' filename parameter is not terminated with a double-
quote or semi-colon character.

Added version for Bill McGarrity: for each linked-node, a preferred
local address (AKA) can be specified to be used as the source address
when creating packets destined for that linked-node. If no local address
is specified, the previous (best match) logic is used: find an AKA that
has the same zone and net as the destination address, if none found, use
the first address with the same zone.

of the user record is going to fail anyway. *this* explains a lot of
instances of the user.dat file being open concurrently, at least one per
active thread with a JS context.

user data is thusly cached - set the 'cached' property member to TRUE. This
prevents an unnecessary re-read of the user file and the leaving the user file
(user.dat) open, at least for JS contexts that contain a "user" object. I don't
think this explains the "too many open files" errors, but it explains at least
*some* number of the user.dat open file descriptors.

to .msg file doesn't mean you should delete the file attachment! The .msg
file will still be pointing to that file and it needs to be there.
SBBSecho was silently failing to delete file attachments on my system, so
I didn't see this problem. Oops. Anyway, this explains Al's problem with
netmail attachments.

server log output/window.
Also added comment about an observed exception upon shutdown with regards to
the node_threads_running counter (protected int) being accessed after
destruction.

removed.

sbbs::lputs() will log the current user alias already, logline()->lputs() would
create log messages with the user alias twice.

attempted.

by Ragnarok@DOCKSUD.

mail "To" and "From" header fields of the form: user name <user@addr> or
"user name" <user@addr> - the email.can filtering would not work

getsmsg() is now based on the current console column (!=0), not the current
node action.

additional, optional, "history" argument - an array of strings that can
be recalled/scrolled using the up/down arrow keys. Currently, I just plan
to use this for netmail destination address recall, but it could be used
other places easily enough.

from a counter, which allowed 2 depths of nesting (any 2 hot key handlers)
to a bit-mask which disallows *any* hot key handler to be re-entrant, but
you can enter any of the *other* hot key handlers from any *other* hot
key handler.

Previously, for example, you could hit Ctrl-P to enter the private message
prompt and then Ctrl-P again to re-enter the private message prompt
(recusively) - confusingly, requiring 2 'Q'uits to get back to whatever
you were doing before you hit the first Ctrl-P. Been like this forever.
Now, while in the Ctrl-P (private message) handler, hitting Ctrl-P again
does nothing. But you can hit Ctrl-T, or Ctrl-K or any other global hot
key and it'll work but those in turn will also disallow recursion of their
own handlers.

the data dir here, we're using it to delete the data files! This explains
why sub data wasn't getting deleted when deleting a message group and
answering "Yes" to the "Delete All Data in Group" dialog.
Eliminated the "clearing ptrs" thing - not needed any more since the move
from ptrs/*.ixb to user/*.subs files.

- If an FTN Echo Tag is more than 25 chars, copy it to the newly created
  sub's "newsgroup" name
  (when exporting areas to echolists, use the newsgroup name for the FTN
  Area/Echo Tag field, if it's not blank)
- Copied the duplicate internal code logic from the importing file areas
  code - it works a lot better

With these changes I was able to successfully import 10247 areas from the
usenet group listing he sent me (with 10258 areas listed). The areas that
were over-written during import were due to the first 63 characters of the
newsgroup names being duplicated - sorry, that's the best I can do.

accommodate configurations of more than 10,000 things (e.g. sub-boards)
in a single menu (e.g. message group).
This brings this value over the 16-bit thresholds (0x7ffff and 0xffff)
where problems could happen if short's/int16's are used to capture the
return value of uifc.list(), but that's unlikely happening anywhere. Still
a good idea to make sure if you're using this library/headers.

name specified, use that instead of the short name as the "Echo Tag"
(Area Tag) value.