iwyu: Cleanup in the following files:

iwyu: Cleanup in the following files:
* at_exit.cc
* atomicops.h
* base_paths.h
* bzip2_error_handler.cc
* callback_internal.h
* command_line.cc
* cpu.cc
* environment.h
* event_recorder.cc
* file_descriptor_shuffle.cc
* file_path.cc

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=80340

[James Hawkins, 2011-03-30 16:41:49]

Brett, I'm taking a look at include-what-you-use against the Chrome code base.
For the most part the fixes are trivial and have revealed some hidden
dependencies that caused problems.

I took a stab at a few files in base, wanted to get your feedback on the types
of changes made.

In particular, I have a style question about size_t and stddef.h vs cstddef.

There are 9 uses of std::size_t (as opposed to leaving std:: prefix off).  This
implies usage of the C++ stdlib instead of stddef.h; however, cstddef explicitly
declares `using std::size_t`.  Thoughts on using std::size_t vs size_t? 
Thoughts on including cstddef vs stddef.h?

[brettw, 2011-03-30 17:32:02]

http://codereview.chromium.org/6759017/diff/1001/base/base_paths_linux.cc
File base/base_paths_linux.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/base_paths_linux.cc#new...
base/base_paths_linux.cc:7: #if defined(OS_FREEBSD)
Can you move this to the bottom? We've standardized on conditional includes
last. Currently, this depends on build_config which isn't included until later.

http://codereview.chromium.org/6759017/diff/1001/base/command_line.cc
File base/command_line.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/command_line.cc#newcode19
base/command_line.cc:19: #include <algorithm>
I don't really like having conditional includes for cross-platform includes. It
seems much more readable to just have these in the shared section.

http://codereview.chromium.org/6759017/diff/1001/base/cpu.cc
File base/cpu.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/cpu.cc#newcode8
base/cpu.cc:8: #include <string.h>
Ditto.

[James Hawkins, 2011-03-30 20:16:02]

Any comments on the size_t issue?

http://codereview.chromium.org/6759017/diff/1001/base/base_paths_linux.cc
File base/base_paths_linux.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/base_paths_linux.cc#new...
base/base_paths_linux.cc:7: #if defined(OS_FREEBSD)
On 2011/03/30 17:32:02, brettw wrote:
> Can you move this to the bottom? We've standardized on conditional includes
> last. Currently, this depends on build_config which isn't included until
later.

Done.

http://codereview.chromium.org/6759017/diff/1001/base/command_line.cc
File base/command_line.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/command_line.cc#newcode19
base/command_line.cc:19: #include <algorithm>
On 2011/03/30 17:32:02, brettw wrote:
> I don't really like having conditional includes for cross-platform includes.
It
> seems much more readable to just have these in the shared section.

OK. I agree, but it's also unfortunate. What really needs to happen here is a
better refactoring of platform-specific implementations.

http://codereview.chromium.org/6759017/diff/1001/base/cpu.cc
File base/cpu.cc (right):

http://codereview.chromium.org/6759017/diff/1001/base/cpu.cc#newcode8
base/cpu.cc:8: #include <string.h>
On 2011/03/30 17:32:02, brettw wrote:
> Ditto.

Done.

[brettw, 2011-03-30 20:57:51]

LGTM.

std::size_t seems weird to me, and I can't ever remember seeing code that uses
it. I think we should stick to plain size_t.

I guess I don't care too much about cstddef vs. stddef.h