(In reply to Dr. Werner Fink from comment #10) > (In reply to Franck Bui from comment #9) > > > Well I wrote a simple test case that allocates then frees 1000 objects of > > 4096MB. I could see the same behaviour: the resource seemed not to > > be released by glibc. However strace(1) showed me that it was (by shrinking > > the heap by calling brk(2)). > > Try to get the system swapping then the kernel will try harder to pick up > the freed memory areas which then results in changes of virtual memory size > and resident set size. ok I will try, thanks. Im the meantime, I'm closing the bug.