In article <07832b7652.tigger@xxxxxxxxxxxxxxxxxxxxxxxxxx>, Nick Roberts <tigger@xxxxxxxxxxxxxxxxxxxxx> wrote: > After running it for about 20 hours, heap growth is almost linear at > 40K per hour, although there is a very slight tail-off. > Assuming an initial size of around of 148K, that means that the the > heap will reach the 4MB limit (presumably an OS limit? - My Iyonix > won't let me increase the system heap beyond 4MB either) just over 4 > days later. > Although all the evidence points to !NewIcons, after a quick > inspection of the code I can't see anything obviously wrong - the > only DIM instructions are at the start or in PROC0, and the latter > only appears to be called once. > As far as I can see, the SWI OS_FSControl looks legal, and sensible > - it should return the canonicalised name in the buffer pointed to > by C% (256 bytes long, created in PROC0). Yes - these results are almost identical to the data I posted earlier. The rate of memory leak appears to depend on the pollidle delay, the default being 500. I think I have eliminated OS_FSControl. a. I have set up some simple test cases and there was no problem with memory leaks. b. I bypassed the results of this call in various ways, although it was still made. The memory leak depended on what other parts of the code was called. -- Chris Johnson --- To alter your preferences or leave the group, visit //www.freelists.org/list/armini-support List-related queries to support@xxxxxxxxxxxx