problem better):
* Re-assert the "RX data" interrupt after the rx data register is read and
  there is remaining data.
This also allows us to only signal the interrupt event for *enabled*
interrupts (which certainly seems logical).

hung threads are hanging while attempting to lock a destroyed semaphore.

only wake-up interrupt_thread() when *asserting* an interrupt.

PlatformSDK.

fields (i.e. to/from/subject)

KnK (v4.4) loops on FOSSIL function 0x0A until the input buffer is empty,
causing an infinite loop at the "Monthly Status" hit any key prompt without
this function implemented.

there's no data waiting.
KnK v4.4 (using QBSERIAL 3.10) apparently keeps reading the base I/O register
for input data until it reads 0.

"auto-savnum" feature isn't working 100%. Deuce: If you monitor api->savnum
going in and out of options from xtrn_cfg() you'll see what I mean.

debug output.
Added optional debugging of FOSSIL functions (counts number of calls to each).

debug output.

(and it's yield interval, in milliseconds, now).

assuming that select()ing for write can detect the condition before a
sendsocket() if more than a single byte, attempt to fix the wedged session
bug.  I give this about a 27% probability of being a fix.

apparently some Linux's clib doesn't define CLK_TCK. <shrug>

potential forced yield. This "tames" BRE down to single-digit CPU utilization!

(if it exists) over-ride those settings.

(the "exec" directory) - used if there's no sbbsexec.ini in the current working
directory (e.g. xtrn/somedoor).

properly checked on dorectory requests which didn't end with a /

(Pops up a calendar for selecting now)
Removed useless glade project file from CVS.

Only thing left is the properties window (whee!)

Enable BBS->Preview and BBS->Edit and Preview menu options.

(or stdin if no file specified)

app...

fine.