Remove static function pointer from LazyInstance

Remove static function pointer

This save ~1720B from the size of debug chrome build.
Also make the traits Delete method typesafe.

BUG=None
TEST=lazy_instance unittests


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=107170

[joth, 2011-10-25 14:22:57]

This doesn't remove the static initializer, but I believe it does reduce the
quantity of linker initialized static data as LeakyLazyInstanceTraits::Delete no
longer requires any storage. (and nor does the new kRegisterOnExit member as
it's const).

http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h
File base/lazy_instance.h (left):

http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h#oldcode184
base/lazy_instance.h:184: // threaded, so don't use atomic operations.
this comment looked incorrect, as it is using an atomic operation
(Release_Store). I made a rough-stab at correcting it.

[willchan no longer on Chromium, 2011-10-25 15:57:25]

LGTM, thanks!

http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h
File base/lazy_instance.h (left):

http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h#oldcode184
base/lazy_instance.h:184: // threaded, so don't use atomic operations.
On 2011/10/25 14:22:57, joth wrote:
> this comment looked incorrect, as it is using an atomic operation
> (Release_Store). I made a rough-stab at correcting it.

I think what's happening here is we *shouldn't* need the atomic. But when we
have bugs with non-joinable threads we get races here. This helps shut up TSAN,
although the code will still probably crash :P Meh. I don't have any opinion
about changing this comment since I'm not sure how to fix it correctly.

[joth, 2011-10-25 17:00:01]

On 2011/10/25 15:57:25, willchan wrote:
> LGTM, thanks!
> 
Cool, thanks. I'll see what CQ thinks of it.

> http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h
> File base/lazy_instance.h (left):
> 
>
http://codereview.chromium.org/8393002/diff/2001/base/lazy_instance.h#oldcode184
> base/lazy_instance.h:184: // threaded, so don't use atomic operations.
> On 2011/10/25 14:22:57, joth wrote:
> > this comment looked incorrect, as it is using an atomic operation
> > (Release_Store). I made a rough-stab at correcting it.
> 
> I think what's happening here is we *shouldn't* need the atomic. But when we
> have bugs with non-joinable threads we get races here. This helps shut up
TSAN,
> although the code will still probably crash :P Meh. I don't have any opinion
> about changing this comment since I'm not sure how to fix it correctly.

Thought it was something along these lines. My reasoning was, there's no
requirement to have a "DELETING" state, c.f. the CREATING thread.
Hopefully this is sufficient.

[commit-bot: I haz the power, 2011-10-25 17:00:14]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/joth@chromium.org/8393002/2001

[commit-bot: I haz the power, 2011-10-25 19:04:49]

Change committed as 107170