instead of a bit-mask representing all the log levels to actually "see", use a
single numeric (or nmemonic) level value (ERROR, WARNING, NOTICE, etc.)
and less severe log levels (levels with a higher numeric value) will be filtered
(e.g. not "seen" or logged). It's the basic syslog() concept.

(Where it is created)
Fixes double-destroy and lock-after-destroy error on Win32.

outbuf_write_initialized before locking/unlocking/destroying outbuf mutex.

if the filename to be edited is the same as the editor's temp file name.

Don't forget... listSemTryWait return values are reversed from sem_trywait()

Micros~1 shortened paths in the drop files. Another fix for installations to
e.g. "C:\Synchronet"

to be moved into a hashlib (todo).

some XSDK doors to possibly work when run from C:\Synchronet\...

Also fix incorrect type lassed to iniFreeStringList.

(Should this be a separate type then?  ini_list_t or something?)

Fix sizeof(char*) leak when reading a webctrl.ini file.

None of these added FREE_AND_NULL() calls should actually do anything.

Still need to comment the lists and ring buffer.

needing to be able to hijack an SSH session.

Currently implemented with socketpair() which is non-Win32 and using
UNIX-domain sockets only since socketpair() on many platforms doesn't
support PF_INET with socketpair().

Something needs to be done about this.

Also, the two extra threads are annoying.

storms as much as possible.

!!!NOTICE!!!
At least Cryptlib 3.2.3a and 3.2.3 (The latest) crashes Synchronet when you
attempt to connect using OpenSSH (SyncTERM with experiment SSH works though)
DO NOT USE THIS UNLESS YOU WANT YOUR BBS TO CRASH!!!

use inkey()/ungetkey() and don't work with down-arrow (ctrl-J). This should
be okay now that input_thread auto-converts Telnet's CR/LF to CR.

error and is fairly common on Linux.

zmodem_recv_file_data() - the offset value was used in rev 1.34 (back when
this change was made) since zmodem_recv_file_frame() accepted the current
file offset as an argument (but doesn't any longer). Mystery solved.

when passed a NaN value (Mozilla bug 284032) - using JSVAL_IS_NUM().

Fix mouse mis-selection bug on non full-sized lists.
Fix possible segfault when no *bar is passed.

Fix bug.
This effects all range requests (podcasts etc)

This fixes a bug where query strings used with PathInfoIndex indexes were
ignored.

Though, now that I think about it, set_cookie() should be a method of the
http_reply object imho.  I may fix this some day.

comments.

is defined.