At the moment I have a pretty slow program directory for ABL. There is quite a bit of pain when ABL "checks files with stat" when running external programs in a tight loop. This can become a problem as the loops build up to 10's of thousands of iterations.
We run most of our ABL via "state-reset" appserver. I'm thinking of using the "-q" (quick) option, even in development. But I was hesitant because of a concern that it might not allow changes to be picked up.
If I use "-q", are my "state-reset" appservers supposed to auto-flush the r-code from memory when a connection is closed?
Or should I use the cool "r-code trick" in my connection procedures instead: http://knowledgebase.progress.com/articles/Article/P130577
I think I've done enough testing to convince myself that "-q" is always safe on development "state-reset" appservers. Despite the use of that parameter, ABL will still flush out the r-code from memory when client connections are closed. No need for the manual "r-code trick".
Now other types of appservers (eg. state-free) are probably a "whole nother story".