Tracy,

one update from current "research" - be conservative when giving vfp memory. At the moment I am building a pattern of 1150-errors thrown after a longrunning app depending on physical memory and sys(3050) settings.

The errors crop up in the runtimes of vfp8SP1, vfp9Sp1 and vfp9Sp0 running a vfp8Sp1 compiled exe (with small provisions for the already scheduled uplift to vfp9 via version() ).
Base settings are 1 GB of RAM and at least 1 and 2 GB of swap disk - they have a startup setting of 694,25MB for vfp.
I encounter errors a sys(3050,x,2**20*256) while sys(3050,x,2**20*192) works. I'll identify the border more exactly (also checking the memory used by OS install and other apps) before plotting the error occurrence on lower memory settings (and later on with higher settings: new machines to play and measure are just around the corner).
But that takes time, since the app runs about half a day and working of saved milestones doesn't show the error. It might be a freak ocurrence (the app is decidedly not the typical vfp-program, even if I routinely force garbage collection and even close down the tables perodically), but if you encounter any 1150's - lower your sys(3050).

my 0.01 EUR (still gaining insight <g>)

thomas