More memory stats code cleanup:...

More memory stats code cleanup:
Move GetSystemCommitCharge() into bsae\process_util*.
Kill PrintChromeMemoryUsageInfo(), which was only used by
reliability_tests.exe on Windows and whose stats are obsolete.
Delete the now-unnecessary chrome\test\perf\mem_usage* files.
BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=31423

[sgk, 2009-11-06 23:52:03]

Let me know if there's any reason GetSystemCommitCharge() really shouldn't be in
base.  I'd like it there for access from my next CL, but if it's a problem
there, I'll put it in chrome\test\chrome_process_utils*.

[sgk, 2009-11-08 15:36:33]

Just noticed darin on vacation, +mbelshe as primary reviewer.

[Mike Belshe, 2009-11-09 03:11:39]

lgtm with comments addressed.

http://codereview.chromium.org/371025/diff/54/1047
File base/process_util.h (right):

http://codereview.chromium.org/371025/diff/54/1047#newcode423
Line 423: // Returns the memory commited by the system.
nit:  units are bytes?
nit: may also want to document that if this routine returns 0 it couldn't
compute the commit charge.
recommend: maybe use KB instead of bytes?  This will overflow at 4GB for most
systems.

http://codereview.chromium.org/371025/diff/54/1048
File base/process_util_linux.cc (right):

http://codereview.chromium.org/371025/diff/54/1048#newcode479
Line 479: LOG(ERROR) << "Failed to parse /proc/meminfo.  Only found " <<
nit: You logged an error here, but not for the case on line 474.

http://codereview.chromium.org/371025/diff/54/1046
File base/process_util_mac.mm (right):

http://codereview.chromium.org/371025/diff/54/1046#newcode226
Line 226: return 0;
nit:  should these log errors like the linux version does?

http://codereview.chromium.org/371025/diff/54/1049
File base/process_util_win.cc (right):

http://codereview.chromium.org/371025/diff/54/1049#newcode809
Line 809: static BOOL InternalGetPerformanceInfo(
You might want to add a comment that this is racy if called from multiple
threads concurrently.

http://codereview.chromium.org/371025/diff/54/1049#newcode826
Line 826: 
nit: remove extra line

http://codereview.chromium.org/371025/diff/54/1049#newcode836
Line 836: return -1;
bug: you're returning -1 here, but the other platform implementations returned 0

http://codereview.chromium.org/371025/diff/54/1043
File chrome/test/chrome_process_util.h (right):

http://codereview.chromium.org/371025/diff/54/1043#newcode28
Line 28: // Returns -1 on failure.
bug: some of the implementations return 0.
recommend: I think this has to be KB, not bytes.  32bits is too small.

[sgk, 2009-11-09 05:30:18]

http://codereview.chromium.org/371025/diff/54/1047
File base/process_util.h (right):

http://codereview.chromium.org/371025/diff/54/1047#newcode423
Line 423: // Returns the memory commited by the system.
On 2009/11/09 03:11:39, Mike Belshe wrote:
> nit:  units are bytes?
> nit: may also want to document that if this routine returns 0 it couldn't
> compute the commit charge.
> recommend: maybe use KB instead of bytes?  This will overflow at 4GB for most
> systems.

Done.

http://codereview.chromium.org/371025/diff/54/1048
File base/process_util_linux.cc (right):

http://codereview.chromium.org/371025/diff/54/1048#newcode479
Line 479: LOG(ERROR) << "Failed to parse /proc/meminfo.  Only found " <<
On 2009/11/09 03:11:39, Mike Belshe wrote:
> nit: You logged an error here, but not for the case on line 474.

Done.

http://codereview.chromium.org/371025/diff/54/1046
File base/process_util_mac.mm (right):

http://codereview.chromium.org/371025/diff/54/1046#newcode226
Line 226: return 0;
On 2009/11/09 03:11:39, Mike Belshe wrote:
> nit:  should these log errors like the linux version does?

Done.

http://codereview.chromium.org/371025/diff/54/1049
File base/process_util_win.cc (right):

http://codereview.chromium.org/371025/diff/54/1049#newcode809
Line 809: static BOOL InternalGetPerformanceInfo(
On 2009/11/09 03:11:39, Mike Belshe wrote:
> You might want to add a comment that this is racy if called from multiple
> threads concurrently.

Done.

http://codereview.chromium.org/371025/diff/54/1049#newcode826
Line 826: 
On 2009/11/09 03:11:39, Mike Belshe wrote:
> nit: remove extra line

Done.

http://codereview.chromium.org/371025/diff/54/1049#newcode836
Line 836: return -1;
On 2009/11/09 03:11:39, Mike Belshe wrote:
> bug: you're returning -1 here, but the other platform implementations returned
0

Done.

http://codereview.chromium.org/371025/diff/54/1043
File chrome/test/chrome_process_util.h (right):

http://codereview.chromium.org/371025/diff/54/1043#newcode28
Line 28: // Returns -1 on failure.
On 2009/11/09 03:11:39, Mike Belshe wrote:
> bug: some of the implementations return 0.
> recommend: I think this has to be KB, not bytes.  32bits is too small.

Done.

(This declaration was leftover cut-and-paste.)

[vandebo (ex-Chrome), 2009-11-09 19:10:29]

http://codereview.chromium.org/371025/diff/83/95
File base/process_util_linux.cc (right):

http://codereview.chromium.org/371025/diff/83/95#newcode474
Line 474: LOG(ERROR) << "Failed to open /proc/meminfo.";
Should this be ERROR, or just WARNING?  It doesn't seem that we should kill the
browser if /proc isn't mounted.

http://codereview.chromium.org/371025/diff/83/96
File base/process_util_win.cc (right):

http://codereview.chromium.org/371025/diff/83/96#newcode835
Line 835: LOG(ERROR) << "Failed to fetch internal performance info.";
Error/Warning ?