Rob Swindell authored 2 months ago and Deucе committed 2 months ago

Fix off-by-one error in checking TimeZoneComboBox->ItemIndex again tz_val
array. When choosing the last (Other/Automatic) item, no change to the
sys_timezone should be made (was setting it to 0/UTC).

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

This was evident by logging in with PETSCII/64col mode and the last ']' of the
"Loading message pointers..." progress indicator was left in the first right
column of the screen.

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

Increase the minmimum free disk space to 1G by default (most modern computer
sysops will probably want to increase this even more).

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

This means that when writing user.birthdate (in JS) for example, we will
always save the field in the proper (YYYYMMDD) format. This fixes issue #855.

Remove use of SM_EURODATE from userdat.c.

This commit includes new @-code MINSPACE (minimum diskspace required to allow
uploads) too (unrelated change).

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

... that includes GJ's (now well known/seen) "SYNCH" splash ANSI.

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

To hopefully address:
!ERROR 2 (No such file or directory) renaming s:\sbbs\data\file/somenum.rep
 (-1 bytes) to s:\sbbs\data\file/somenum.rep.6777ed7c.bad

Clearly flength() is returning -1 (file doesn't exist) at this point, so
perhaps the call to fremove() just above updated a cached directory?

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

Fix some newly introduced MSVC warnings:
  conversion from 'time_t' to 'uint32_t', possible loss of data

Should these local vars be time32_t instead of uint32_t?

I'm not clear why the 2 step conversions now (?):
	time_t -> uint32_t
	uint32_t -> time32_t

So I can poke it and break things.

0 is a valid value, results are undefined.

Validate the size of the message at each step and throw an error if
something goes wrong.

Should fix the issue where an IRC spammers DNS was detected as local,
so the anti-spammer bot banned my server address intead of the
spammers address.

I'm not sure if I can do this with events, or just issues, and I
don't want to directly annotate this issue because that could lead
to missing an issue after a future code change.

The return value of getch() will never be const though.

Since Coverity treates the return value of write() as tainted
(valid for negative values, not so valid for positive ones), do an
upper bound check on the result as well as lower bound to clear
the tainted flag.

Have zmodem_recv_nibble() explicitly range-check all values, and
handle negative values.  If this remains an issue, it can be
untainted with range checks on each nibble, but let's not go there
to start with.

This is purely to shut up Coverity since there's no way the value
could be outside the range.

Coverity seems to hate it mostly in the mutex, so change it to
mostly outside of the mutex and see if that's fine.

It doesn't really matter where it's changed as it's an atomic bool.

We don't want ch == -2 derping through the rest of the loop.
Luckily, ch is never currently == -2 as far as I know.

Initially set 1MB as the max macro size.  I can't imagine wanting
one bigger than this, but I can bump it up if needed.

This likely fixes a vulerability where the BBS could at least crash
SyncTERM, and at most take control of the users system.  This
likely isn't the only place that needs bounds checks like this.

If a write() to the sound device fails (for example USB device goes
away), this would loop forever (in it's own thread and whatnot,
so you only notice because sounds stops and a core gets used up).

Not that there's any need to, it appears that it was written for
the web server, then discarded for that purpose and now is only
called from the JS Socket.sendfile() method, which always passes
NULL and 0 for the last two parameters, and has a copy/pasted
implementation for TLS sockets.  The only consumer of that in
the tree appears to be gopher_service.js.

It was apparently to use as a wrapper for the high-performance
FreeBSD sendfile(), but that code behaved differently than all
the other platforms, and was disabled (behind USE_SENDFILE, which
isn't defined anywhere).

This should really just be folded into either js_socket_sendfilesocket()
or js_sendfile() with the extra knobs broken off and all the TODO
comments I'm adding here addressed.

Though, since Socket.sendfile() returns a bool where true indicates
that either the size of the file at the start of the function
was sent, or an EOF was reached, and false indicates that
"something else happened", some of the TODO comments don't really
need to be addressed.

Basically, this is a crap function and it's only used by an even
more crap JS wrapper, it should be tucked into a backward
compatibility function, removed from the JSDocs, and forgotten
about.

If Coverity keeps complaining about this, I'll wait until after
the next release and drag this out back and shoot it.  Otherwise,
I'll just forget it ever existed.

This may be the first time in my life that I've used difftime()...
which is strictly one of only two things you're actually allowed
to do with a time_t per the C standard (the other is convert into
a broken-down time).

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

This reverts commit 9b2c9c08.

Don't need this now that we use the property sheet method

Rob Swindell authored 3 months ago and Deucе committed 2 months ago

This is the (better) way. See, Deuce knows the coolness of property sheets!