Mac: clang build

Mac: clang build

Fixes the last few warnings / errors necessary to build chrome/mac with clang.

See http://code.google.com/p/chromium/wiki/Clang

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=60131














Dependencies:

skia newer than http://code.google.com/p/skia/source/detail?r=583
http://codereview.chromium.org/522020
( http://llvm.org/bugs/show_bug.cgi?id=7445 for release builds)
( http://codereview.chromium.org/2859012 )
http://codereview.chromium.org/2857015 for unit_tests

http://llvm.org/bugs/show_bug.cgi?id=7461
http://llvm.org/bugs/show_bug.cgi?id=7465 (for nacl)
-or-
http://llvm.org/bugs/show_bug.cgi?id=7487


Done:

http://lists.cs.uiuc.edu/pipermail/cfe-dev/2010-June/009374.html
http://lists.cs.uiuc.edu/pipermail/cfe-dev/2010-June/009375.html
http://codereview.chromium.org/2769014
http://codereview.chromium.org/2730015
http://codereview.chromium.org/2836002
http://llvm.org/bugs/show_bug.cgi?id=7390
http://trac.webkit.org/changeset/61360
http://llvm.org/bugs/show_bug.cgi?id=7431
http://codereview.chromium.org/2801014
http://codereview.chromium.org/2805026
http://codereview.chromium.org/2876002
http://codereview.chromium.org/2840017
http://codereview.appspot.com/1706045
http://lists.cs.uiuc.edu/pipermail/cfe-dev/2010-June/009570.html 
v8 newer than http://code.google.com/p/v8/source/detail?r=4888
http://codereview.chromium.org/3089004
http://codereview.chromium.org/3012045
https://bugs.webkit.org/show_bug.cgi?id=43491

http://llvm.org/bugs/show_bug.cgi?id=7353
http://llvm.org/bugs/show_bug.cgi?id=7358
http://llvm.org/bugs/show_bug.cgi?id=7355
http://llvm.org/bugs/show_bug.cgi?id=7386

( http://llvm.org/bugs/show_bug.cgi?id=7389 )

…and many more.

[Nico, 2010-09-21 23:17:20]

Turns out clang/linux builds with way more warnings disabled…I guess some of
these flags can be removed once this CL lands?

[Nico, 2010-09-21 23:29:59]

Done.

[Nico, 2010-09-21 23:37:58]

With this, target "chrome" builds successfully with clang. And the resulting
Chromium.app even seems to work.

(Except for

 	PhaseScriptExecution "Postbuild \"Verify global text symbol order\""
/Users/thakis/src/chrome-git/src/clang/obj/chrome.build/Debug/chrome_dll.build/Script-1EEBF8FF12C450C824B00DF6.sh

, but who cares about that, right?)

[viettrungluu, 2010-09-21 23:49:08]

LGTM with nits fixed.

http://codereview.chromium.org/2762014/diff/66002/84001
File build/common.gypi (right):

http://codereview.chromium.org/2762014/diff/66002/84001#newcode1240
build/common.gypi:1240: # This is required soleley for
base/third_party/dmg_fp/dtoa.cc
s/ley/ly/

Which conveniently gives you room for a period at the end of the sentence.

http://codereview.chromium.org/2762014/diff/66002/84002
File chrome/browser/autofill/autofill_text_field_mac.mm (right):

http://codereview.chromium.org/2762014/diff/66002/84002#newcode29
chrome/browser/autofill/autofill_text_field_mac.mm:29: // |id<NSCopying>|, but
we want to call an NSObject method (-isKindOfClass:)
*we*?

http://codereview.chromium.org/2762014/diff/66002/84002#newcode32
chrome/browser/autofill/autofill_text_field_mac.mm:32: NSObject* object =
static_cast<NSObject*>(anObject);
Maybe this should be NSObject<NSCopying>*? (And then you could use |object|
everywhere.) But this would be more verbose.

(Or, you could even say id<NSObject,NSCopying>* for super-extra correctness.)

[Evan Martin, 2010-09-22 00:03:46]

Please update the description with what the actual change contains before you
submit :)

http://codereview.chromium.org/2762014/diff/85001/70008
File ipc/ipc_channel_posix.cc (right):

http://codereview.chromium.org/2762014/diff/85001/70008#newcode138
ipc/ipc_channel_posix.cc:138: const size_t kMaxPipeNameLength =
sizeof(((sockaddr_un*)0)->sun_path);
Is this really the best way to accomplish this?  There's gotta be some fancy
C++ism you could use here...

[Nico, 2010-09-22 00:06:25]

http://codereview.chromium.org/2762014/diff/85001/70008
File ipc/ipc_channel_posix.cc (right):

http://codereview.chromium.org/2762014/diff/85001/70008#newcode138
ipc/ipc_channel_posix.cc:138: const size_t kMaxPipeNameLength =
sizeof(((sockaddr_un*)0)->sun_path);
On 2010/09/22 00:03:47, Evan Martin wrote:
> Is this really the best way to accomplish this?  There's gotta be some fancy
> C++ism you could use here...

I asked on cfe-dev, and this is what The Chris Lattner recommended. Someone said
that sizeof(sockaddr_un::sun_path) works in c++0x, but…

[Mark Mentovai, 2010-09-22 16:38:35]

LG

http://codereview.chromium.org/2762014/diff/85001/70002
File build/common.gypi (right):

http://codereview.chromium.org/2762014/diff/85001/70002#newcode1237
build/common.gypi:1237: ['clang==1', {
Do we need this for anything else in the Mac build, or is it just this one -Wno?

http://codereview.chromium.org/2762014/diff/85001/70004
File chrome/browser/cocoa/authorization_util.mm (right):

http://codereview.chromium.org/2762014/diff/85001/70004#newcode51
chrome/browser/cocoa/authorization_util.mm:51: NSString* prompt_ns =
const_cast<NSString*>(
Double casts are worse, readability-wise, than a C-style cast.

[Nico, 2010-09-22 16:41:07]

http://codereview.chromium.org/2762014/diff/85001/70002
File build/common.gypi (right):

http://codereview.chromium.org/2762014/diff/85001/70002#newcode1237
build/common.gypi:1237: ['clang==1', {
On 2010/09/22 16:38:36, Mark Mentovai wrote:
> Do we need this for anything else in the Mac build, or is it just this one
-Wno?

There's more: gtest.gyp and skia.gyp check this and do minor tweaks.