Re: Finding Memory Leak ?

[Pavel Romashkin <promashkin(at)cmdl.noaa.gov>]

Echoing what David said, my chase after the ghostfont object immediately
came to my mind. I used Randall's suggestion to verify it was a bug. The
advice was to do the following.
As heap variables and pointers are created, the pointers (or orefs) are
created sequentially. If you use Help, /heap before you kill the
program, you will get the list of all existing heap references.
Copy-paste that to your favorite MS Word processor. Then, exit the
widget program and get heap info on the leaking pointers. Compare
pointer numbers with those you had in the program (using word processor
allows faster search if you had a lot of them, as I did). Thus, you can
locate when did they get created, and from that, who actually made them
(you or RSI :-).
The ghostfont bug, actually, was not caused by any object alone. It was
caused by something in .Reset_session that was not doing what it should
have to the graphics system; I am not too good at that, maybe somebody
from RSI could tell us better. The bug only appeared after using .Reset_session.
Cheers,
Pavel