Add shutdown trace functions

Add shutdown trace functions


BUG=chromium-os:14737
TEST=None

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=84662

[willchan no longer on Chromium, 2011-05-04 21:06:49]

LGTM

http://codereview.chromium.org/6943001/diff/1/chrome/browser/browser_main_pos...
File chrome/browser/browser_main_posix.cc (right):

http://codereview.chromium.org/6943001/diff/1/chrome/browser/browser_main_pos...
chrome/browser/browser_main_posix.cc:98: sleep(UINT_MAX);
I think you need <limits.h> for this constant and <unistd.h> for sleep (or you
could use PlatformThread::Sleep(), which might make this easier to port to
Windows if we ever do that).

Note that the noinline attribute is probably gcc only and this will probably
cause the clang bot to break. You should probably add this to
compiler_specific.h.

Also, I'm not completely sure this works, please note
http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html says: 
=====
noinline
This function attribute prevents a function from being considered for inlining.
If the function does not have side-effects, there are optimizations other than
inlining that causes function calls to be optimized away, although the function
call is live. To keep such calls from being optimized away, put
          asm ("");
     
(see Extended Asm) in the called function, to serve as a special side-effect.
=====

I think sleep() counts as a side effect, so it probably works as is, but I
thought I'd point that out.

[DaveMoore, 2011-05-09 17:41:42]

http://codereview.chromium.org/6943001/diff/1/chrome/browser/browser_main_pos...
File chrome/browser/browser_main_posix.cc (right):

http://codereview.chromium.org/6943001/diff/1/chrome/browser/browser_main_pos...
chrome/browser/browser_main_posix.cc:98: sleep(UINT_MAX);
On 2011/05/04 21:06:49, willchan wrote:
> I think you need <limits.h> for this constant and <unistd.h> for sleep (or you
> could use PlatformThread::Sleep(), which might make this easier to port to
> Windows if we ever do that).
> 
> Note that the noinline attribute is probably gcc only and this will probably
> cause the clang bot to break. You should probably add this to
> compiler_specific.h.
> 
> Also, I'm not completely sure this works, please note
> http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html says: 
> =====
> noinline
> This function attribute prevents a function from being considered for
inlining.
> If the function does not have side-effects, there are optimizations other than
> inlining that causes function calls to be optimized away, although the
function
> call is live. To keep such calls from being optimized away, put
>           asm ("");
>      
> (see Extended Asm) in the called function, to serve as a special side-effect.
> =====
> 
> I think sleep() counts as a side effect, so it probably works as is, but I
> thought I'd point that out.

Done. I kept the sleep() call, added the asm("") and included the necessary
files.

[Evan Martin, 2011-05-09 22:49:44]

http://codereview.chromium.org/6943001/diff/9001/base/compiler_specific.h
File base/compiler_specific.h (right):

http://codereview.chromium.org/6943001/diff/9001/base/compiler_specific.h#new...
base/compiler_specific.h:77: #define NOINLINE __attribute__((noinline))
Sorry to nitpick, but the comment above doesn't match this.  It looks like the
file style is to put each of these attrs in its own #if defined block with a
comment above them.