Martin wrote:
>  MFC: Plug a memory leak (shows up with +/@, i. 10).
>  All official 4.05 versions had this leak.  

Thank you for this interesting history!  My first J version was 4.06d (and
it's still one of my favorites; I miss the MDI IDE), so I wasn't around to
see this discussed on the Forums.  

My (previous) impression was that the interpreter essentially prohibited
such leaks from arising, because Roger once said "There are no memory leaks
in the J interpreter.
(Never has from day 0.)" and then went on to describe how the interpreter
assured that [1].  But my C is so old and rusty that I cannot even say
whether such a thing is possible.

>  handful of alignment bugs.

These, I've definitely run into.  Their most annoying characteristic is that
they can be latent -- they don't trigger a crash until the wound is touched,
which can be many lines later, or as a result of some side-effect (such as
displaying the array).

>  (Again, because Roger did good work on the testsuite, too:

I wish the test suite were made public; it would be very instructive.  Roger
has said that in a way, this suite specifies the language even more
precisely than the DoJ [2].  Also, it would be a boon to potential J
reimplementers (including me).

>  there's a reason when a test array has shape 3 5 13 -- all primes.

Clever!  Doesn't surprise me.

>  But for J programmers who want to implement 24/7 services, 
>  such small leaks do matter.

My primary professional J system was client/server, but I recycled the
server every night.  I'd be very interested in hearing about 24x7 J systems,
if anyone's aware of one.

-Dan


[1]  http://www.jsoftware.com/pipermail/programming/2007-March/005706.html 

[2]  http://www.jsoftware.com/pipermail/general/2005-March/021219.html and
     http://www.jsoftware.com/pipermail/general/2006-September/027885.html 


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to