🚀 go-pugleaf

In article <857297254snz@wash.demon.co.uk>
           john@wash.demon.co.uk "John Washington" writes:
: Anyhow, perhaps this thread could usefully wander off to
: more general matters.  I wonder if other developers could
: share their advice on testing/debugging?  I find that modern
: debuggers, altho brilliant, are not well suited to
: investigating time-dependant programs such as comms.  I've

You could say. *shudder*

: not tried running ka9q under td386 (would it work?) because
: I reckon that any attempt to use breakpoints or single step
: would pretty soon drop the line or at least alter the
: picture.

'Pretty soon'? Instantly. And w/o breakpoints or tracing, a debugger
is useless.
It'd probably work under td386; however the millisecond timer may be
mangled (it depends what TDH386.SYS does in Ring 0) as under Win3.1
(but not under OS/2 nyahh nyahh nyahh.)

: Generally I either call a function to log info to a file, or

I've got a tiny C++ class to do this without slowing everything down
too much (it's got a microcache in it); as I compile *everything* as
C++ it's fairly useful.

: write direct to video ram.  On this particular problem I've
: been writing to video ram at the top (border) of the
: relevant window.  It works, but it feels like pushing a rock
: uphill with my nose.

I'd say. Writing to a file is better IMO because you can write
lots of trace info (i.e. more than one line) without each line
overwriting earlier ones.

A trace() function which calls vsprintf() does the job effectively.

: Anyone, any cures for a sore nose?

A scarf?

(I like the one from the Hacker's Dictionary myself - 'like kicking
dead whales down the beach' - very evocative...)

-- 
The weird stuff in my Reply-To: line *is* a real address; it just looks
like a message-ID to hinder junkmail scanners.
Coming soon: automatic message-ID transference!