[7ca0fdf] Disable munmap() if it causes VM map holes
jasone at canonware.com
Mon Apr 16 14:40:52 PDT 2012
On Apr 16, 2012, at 2:19 PM, Mike Hommey wrote:
> Hoped for 0x7f82e5237000, got 0x7f82e5237000
Huh. The only explanation I can think of is that the dynamic loader is calling mmap in the midst of the mmap/munmap calls, thus resetting the "start from" address to use for address scanning.
> But anyways, why do you want to do a configure run test when you know
> the result you want?
I wanted it to "just work", but you've convinced me that it's hard to make a reliable test. I'll add --disable-munmap.
More information about the jemalloc-discuss