Mac_BSD_resources underestimates RAM size

[aaaaalbert]

Mac_BSD_resources.py uses sysctl hw.physmem to find the size of RAM. Unfortunately, the info is off by a factor of 4 on a MacBook Pro 8,2 running Mac OS X 10.6.8 on which I'm testing this. (At least we are underestimating the size, so this only hurts the Seattle install, but not the resource donor.)

sysctl hw.memsize gives the correct memory size, 8 GB here.

See also https://discussions.apple.com/thread/1753088?start=0&tstart=0

We should test what the different hw. keys return on a couple of machines, and then decide on a fix.

[aaaaalbert]

@fmetzger reports that on his 10.10.5 box, hw.physmem doesn't even exist as a key in sysctl.

[XuefengHuang]

I tested on my MacBookPro11,1 running Mac OS X 10.10.5.

172-16-29-190:~ Xuefeng$ sysctl hw.memsize
hw.memsize: 8589934592
172-16-29-190:~ Xuefeng$ sysctl hw.physmem
hw.physmem: 2147483648
172-16-29-190:~ Xuefeng$ uname -v
Darwin Kernel Version 14.5.0: Wed Jul 29 02:26:53 PDT 2015; root:xnu-2782.40.9~1/RELEASE_X86_64

[lukpueh]

$ uname -v; sysctl hw.physmem; sysctl hw.memsize;
Darwin Kernel Version 13.4.0: Wed Mar 18 16:20:14 PDT 2015; root:xnu-2422.115.14~1/RELEASE_X86_64
hw.physmem: 2147483648
hw.memsize: 4294967296

[michaelkux]

$ uname -v; sysctl hw.physmem; sysctl hw.memsize;
Darwin Kernel Version 13.4.0: Wed Mar 18 16:20:14 PDT 2015; root:xnu-2422.115.14~1/RELEASE_X86_64
hw.physmem: 2147483648
hw.memsize: 17179869184

[pooledge]

$ uname -v; sysctl hw.physmem; sysctl hw.memsize;
Darwin Kernel Version 12.6.0: Wed Mar 18 16:23:48 PDT 2015; root:xnu-2050.48.19~1/RELEASE_X86_64
hw.physmem: 2147483648
hw.memsize: 17179869184

[aaaaalbert]

Per PR #6, my proposed fix is to use sysctl hw.memsize.

Thank you all for testing and contibuting!

[durbrow]

In case no one figured this out (which they did)...

In sysctl.h,
hw.physmem is documented to be an int (i.e. 32 bit signed int).
hw.memsize is documented to be a uint64_t (i.e. 64 bit unsigned int).

The value you are getting for hw.physmem is the maximum value a 32-bit signed int can hold.

[JustinCappos]

Always nice to get more clarity, even (way) after the fact. :)