2016-01-20 06:06:39 ERROR line 1836 renaming c:\fd\outbound\20060640.pk_ to
c:\fd\outbound\20060640.pkt
I think this is because the file 200606040.pkt already existed when the
.pk_ temp file was created and thus the rename will fail (can't rename over
an existing file) and then the packet is sent to the wrong link.
When generating a packet name, keep incrementing until both the .pkt and
.pk_ variants do not already exist. This should fix this scenario.

Currently, the (previously undocumented) top, left, width, dflt, and bar
parameters to uifc.list() are moved into this object and can no longer be
passed in the function call.

This is still only half-baked, so it will need a few weeks at least in CVS
before it congeals enough to be stable.

lock across things with timeouts in them, but since if another thread has
the file locked we shouldn't be fillding with it anyway, this should be
fine.

Fixes weird (and relatively rare) timing issues with node.dab accesses on a
many core host.

seen this in a regular build, so I'm not sure if this is a "real bug" or
not.

in the default user config, a scansubs(SCAN_ALL) will still not do a new
file scan.

This will create the entire path (e.g. of sub-directories) requested and is
used extensively in the scfg-related stuff.

exit_code value.

(based on FTS-5005 description) - both detection (won't pack/bundle outgoing
mail for a node that is "busy" with a mailer session) and exclusive-creation
(to inform the mailer not to allow a session for this node while outgoing mail
is being processed).
- Uses strListFind() from updated xpdev.
Debug-level log message for all file deletions.
Some log message update (more detail in some, removal of others).
Bug-fix: outbound FREQ netmail messages would not be deleted after being
converted to BSO *.req files. It appears no one has yet tested this feature?
Cleaned-up sbbsecho.h: use of C99-types, defined SBBSECHO_ARCTYPE_NONE magic
number macro.
Incremented version to 2.32.

This solves the problem of a JS-external program (e.g. SyncWX) being aborted
at a pause prompt with Ctrl-C or Q and that abort status propagating back to
the external program's menu. If there is a module which needs to propagate
the console-abort status up to the parent(s), then this could introduce a
problem. We'll see.

signedness

- Added a lot more detail for outbound echomail and netmail (msgs, packets,
  and bundles) to help identify the source of problems
- Removed a lot of unnecessary "status" type log output ("Scanning for ...")
  as it was back in 2005.
- Date format in log is more standard (year-month-day)
- Logs error level and import/export statistics when exiting
- Packs bundles *before* writing to FLO files (the only operational change).
- sbbsecho.log should actually be smaller (grow slower) though there is more
  helpful information than before.
- Incremented version to 2.31.

file or creates a netmail message (*.msg) file. This will hopefully help to
provide that evidence that either:
A. there is a bug in SBBSecho with regards to passing mail to downlinks, or:
B. Michiel van der Vlist is just a whimpering ass-hat

old line length was.  Some usage specifies a huge old line length (such as
1<<30) to indicate that it's already one line perparagraph.  In thise case,
word_unwrap() would allocate oldlen+1 bytes to store the paragraph data in
regardless of it never really being that long.

Also, check the return value of word_unwrap().  Not checking it caused a `
crash with triggered finding the above mentioned bug.

The JS-modeled user object's ip_address and host_name property values did not
reflect the current connection during logon events (JS modules executed during
logon, e.g. logon.js). Moved the initialization of the current user
(sbbs_t::useron) ipaddr and comp field to sbbs_t::logon() rather than when
control returns to sbbs_t::answer(). Removed the (now) redundant saves of these
user fields from answer().

The "Start with message number" prompt for the [F]ind text command would
always start searching at message #64 due to typo (i=64 rather than i=i64).
Maybe 'i64' isn't a good variable name.

File.crc32 contained the bit-inverted result of the CRC-32 calculation.

- it actually *is* an expected condition that a to_net.addr or from_net.addr
msg header field is populated with a valid address (string) even though the
to_net.type or from_net.type value is NET_NONE.

The nntpservice.js and newslink.js modules rely on this behavior to display
and export local NNTP-posters' email addresses in message headers, but still be
able to identify newslink-imported messages as *not* local (the from_net.type =
NET_INTERNET) and thus not export them back to the network.

This was not the original intention behind these header fields in SMB, but it
was retrofitted to support this condition 10+ years ago and I forgot.

area-fix requests or generating any netmail not from the mail base.

When using the SBBSCTRL:User->List menu option, it would take minutes for the
form to fully populate and display (1457 users). It turned out this was due to
a 200ms delay for each open of the data/user.dat file.

I never noticed this problem before, but I went ahead and optimized for this
situation since it was a bit silly to be opening
and closing the user.dat almost 1500 times to get a list of users. This made a
huge difference and the user list appears in about 4 seconds now.

Oddly, this problem does not happen when listing users in the terminal server
or when iterating through them in JavaSciprt (e.g. userlist.js).

- openuserdat()
- readuserdat()
- parseuserdat()

getuserdat() still exists and just calls those 3 functions. This allows other
code (e.g. the control panel) to optimize iterative calls to getuserdat() by
opening the file just once or even reading the entire contents at once and then
parsing each user record from memory.

Updated style of arguments in userdat.h (i.e. remove argument name if obvious
from the type).

Don't add TZUTC kludge lines to exported netmail and echomail messages when
there is already a TZUTC FidoNet control line in the message header (unexpected
but apparently, GoldEd+ does this).
If there are other duplicate FTN control/kludge lines created by SBBSecho
exporting messages added to SMBs by GoldEd+, we should address those too, even
though I believe in most cases, you should configure the editor to *not* add
these header fields in the first place.
Do not convert exisiting Origin Line's to effectively comments (replacing *
with # in " * Origin") if the current sub-board has been configured to not add
network tag/origin-lines.

Bug-fix: MSG_TO_NET would not expand properly (to a blank string) when the
to_net.type header value was NET_NONE.

Display translated to/from/reply net-type (from numeric to string).

Don't include the to/from network-arddress in the displayed header if the
net-type value is NET_NONE (not a normal/expected combination).

addresses (e.g. FidoNet and QWK) using the new smblib feature.

and RECIPIENTNETADDR fields in messages posted to sub-boards. Apparently
GoldEd+ does this and it triggered a chain of bugs (mostly cosmetic), but bugs
that bugged me, so I'm squashing them. Now messages being export to QWK will
not include the RECIPIENTNETTYPE/ADDR information unless it's netmail.
Ditto for importing. Also, the net_type is more accurately determined (i.e. if
there's no '@' in the address, as there not be any expected for QWK and Fido
net addresses).

All other command options remain case insensitive.

to stdout).

Previous behaviour was:
1) By default, send error messages to errfp (stderr), and other log messages to
   confp (stdout).
2) If -e and one of -o or -q are specified, also send a copy of error messages
   to statfp (defaults to stderr, changed to /dev/null by -n)
3) If -e is specified, also send a copy of error messages to confp (defaults to stdout)

If -e, -o, were both specified, this resulted in an error log entry going into the file
specified by -e, the file specified by -o, and confp (stdout by default).

New behaviour:
1) By default, send error messages to errfp (stderr), and other log messages to statfp
   (stderr).
2) If -e is specified, send error messages to the specified file and statfp (stderr).
3) Always send status messages to statfp (stderr).

Set up some more stuff for jsdoor...

Call gethostname() since there is no "SCFG value" for it.

Send "status messages" (log messages that aren't errors, and the banner) to
/dev/null by default (ie: force the "-n" argument)

Send "error messages" (log messages that are error according to log level) to
"error.log" by default (ie: force the "-eerror.log" argument)

Don't report a JavaScript warning ("Disconnected") until the OperationCallback
has been executed 10 times with the user offline. This gives the executing
script ample time to recognize the disconnected state and exit gracefully.
Script that do not check/notice that the user has disconnected will still be
auto-terminated (by default) and this warning will still be logged in those
cases. The value 10 was picked at random and appears to work fine.

version of configured time zone, e.g. "PST").