Add a simple about:tcmalloc page that just displays its canned stats....

Add a simple about:tcmalloc page that just displays its canned stats.
Add USE_TCMALLOC to control its inclusion.  Make browser depend on
tcmalloc to get the PERFTOOLS_DLL_DECL definition.
BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=26816

[Mike Belshe, 2009-09-21 15:14:15]

lgtm!

http://codereview.chromium.org/216033/diff/1/2
File chrome/browser/browser_about_handler.cc (right):

http://codereview.chromium.org/216033/diff/1/2#newcode43
Line 43: #include "third_party/tcmalloc/google/malloc_extension.h"
Does this work ok on linux/mac?  Not sure if we need to start defining a
USE_TCMALLOC macro into chrome?

http://codereview.chromium.org/216033/diff/1/2#newcode188
Line 188: std::string data;
This should definitely be off for those platforms not using it.

[sgk, 2009-09-21 17:34:54]

http://codereview.chromium.org/216033/diff/1/2
File chrome/browser/browser_about_handler.cc (right):

http://codereview.chromium.org/216033/diff/1/2#newcode43
Line 43: #include "third_party/tcmalloc/google/malloc_extension.h"
On 2009/09/21 15:14:16, Mike Belshe wrote:
> Does this work ok on linux/mac?  Not sure if we need to start defining a
> USE_TCMALLOC macro into chrome?

No, it doesn't work on Mac and Linux (hooray for trybots!).  A USE_TCMALLOC
sounds right.  I'll parameterize that and resubmit a real CL.

[sgk, 2009-09-21 21:10:02]

This adds an explicit USE_TCMALLOC.

Please double-check the by-hand #define PERFTOOLS_DLL_DECL in
browser_about_handler.cc.  Without this, it gets LNK4217 (locally defined
symbol) errors, but setting it by hand feels wrong.

I tried #include "third_party/tcmalloc/config.h" but that ended up not being
able to pull in tcmalloc/windows/port.h and I wasn't sure whether to keep
following that thread or clean it up in some other way.

[Mike Belshe, 2009-09-22 04:52:30]

Let's figure out what the deal is with PERFTOOLS_DLL_DECL.  We shouldn't need it
- maybe it is related to why we're seeing symbols exported from chrome.dll in
release builds.

http://codereview.chromium.org/216033/diff/7001/8001
File chrome/browser/browser_about_handler.cc (right):

http://codereview.chromium.org/216033/diff/7001/8001#newcode56
Line 56: //#include "third_party/tcmalloc/config.h"
did you want this commented out line?

http://codereview.chromium.org/216033/diff/7001/8001#newcode58
Line 58: #define PERFTOOLS_DLL_DECL
Something fishy is going on with this macro.  You should not need it.

[sgk, 2009-09-22 12:37:33]

http://codereview.chromium.org/216033/diff/7001/8001
File chrome/browser/browser_about_handler.cc (right):

http://codereview.chromium.org/216033/diff/7001/8001#newcode56
Line 56: //#include "third_party/tcmalloc/config.h"
On 2009/09/22 04:52:31, Mike Belshe wrote:
> did you want this commented out line?

Done.

http://codereview.chromium.org/216033/diff/7001/8001#newcode58
Line 58: #define PERFTOOLS_DLL_DECL
On 2009/09/22 04:52:31, Mike Belshe wrote:
> Something fishy is going on with this macro.  You should not need it.

Right you are.  The browser target in chrome.gyp needed a dependency on tcmalloc
instead.

[Mike Belshe, 2009-09-22 15:46:15]

lgtm

On 2009/09/22 12:37:33, sgk wrote:
> http://codereview.chromium.org/216033/diff/7001/8001
> File chrome/browser/browser_about_handler.cc (right):
> 
> http://codereview.chromium.org/216033/diff/7001/8001#newcode56
> Line 56: //#include "third_party/tcmalloc/config.h"
> On 2009/09/22 04:52:31, Mike Belshe wrote:
> > did you want this commented out line?
> 
> Done.
> 
> http://codereview.chromium.org/216033/diff/7001/8001#newcode58
> Line 58: #define PERFTOOLS_DLL_DECL
> On 2009/09/22 04:52:31, Mike Belshe wrote:
> > Something fishy is going on with this macro.  You should not need it.
> 
> Right you are.  The browser target in chrome.gyp needed a dependency on
tcmalloc
> instead.