[Mac] Use an AppleEvent to tell the Finder to open downloaded items, rather than NSWorkspace.

[Mac] Use an AppleEvent to tell the Finder to open downloaded items, rather than NSWorkspace.

BUG=32921, 50263
TEST=Force a PDF to download. Quit Preview, if open. Open the downloaded PDF from the download shelf. Preview opens and becomes frontmost.
TEST=Download a file of a type that you do not have an application with which to open it. Open it from the download shelf. Finder bounces for your attention to choose an application to open it.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=56026

[Avi (use Gerrit), 2010-08-12 20:17:35]

http://codereview.chromium.org/3151011/diff/1/2
File base/base.gypi (right):

http://codereview.chromium.org/3151011/diff/1/2#newcode187
base/base.gypi:187: 'scoped_adesc.h',
adesc? aedesc, perhaps?

http://codereview.chromium.org/3151011/diff/1/3
File base/scoped_aedesc.h (right):

http://codereview.chromium.org/3151011/diff/1/3#newcode15
base/scoped_aedesc.h:15: // This class is paramtized for additional type safety
checks.  You can use the
parameterized

http://codereview.chromium.org/3151011/diff/1/3#newcode18
base/scoped_aedesc.h:18: class scoped_AEDesc {
This is the worst of both naming worlds. We have ScopedBstr and
ScopedNSAutoreleasePool yet scoped_nsobject. We certainly don't have anything
with both underline and intercaps. Talk to Mark the Style Maker.

http://codereview.chromium.org/3151011/diff/1/4
File chrome/browser/platform_util_mac.mm (right):

http://codereview.chromium.org/3151011/diff/1/4#newcode33
chrome/browser/platform_util_mac.mm:33: //  2. Silent no-op for unassociated
file types http://crbug.com/50263
Stick some delimiter after the text and before the URL; a colon?

http://codereview.chromium.org/3151011/diff/1/4#newcode47
chrome/browser/platform_util_mac.mm:47: status = AECreateDesc(typeApplSignature,
 // type
FYI: typeApplicationBundleID, all new-school and shiny (don't bother here tho).

http://codereview.chromium.org/3151011/diff/1/4#newcode87
chrome/browser/platform_util_mac.mm:87: typeFSRef,  // typeCode
typeFileURL? Avoids the FSRef conversion (see
http://developer.apple.com/mac/library/technotes/tn/tn2022.html). All the same
to me though.

http://codereview.chromium.org/3151011/diff/1/4#newcode112
chrome/browser/platform_util_mac.mm:112: kAENoReply,  // sendMode
+ kAEAlwaysInteract?

[Robert Sesek, 2010-08-12 20:58:25]

http://codereview.chromium.org/3151011/diff/1/2
File base/base.gypi (right):

http://codereview.chromium.org/3151011/diff/1/2#newcode187
base/base.gypi:187: 'scoped_adesc.h',
On 2010/08/12 20:17:35, Avi wrote:
> adesc? aedesc, perhaps?

Do'h.  Good catch.

http://codereview.chromium.org/3151011/diff/1/3
File base/scoped_aedesc.h (right):

http://codereview.chromium.org/3151011/diff/1/3#newcode15
base/scoped_aedesc.h:15: // This class is paramtized for additional type safety
checks.  You can use the
On 2010/08/12 20:17:35, Avi wrote:
> parameterized

Not my day for spelling.

http://codereview.chromium.org/3151011/diff/1/3#newcode18
base/scoped_aedesc.h:18: class scoped_AEDesc {
On 2010/08/12 20:17:35, Avi wrote:
> This is the worst of both naming worlds. We have ScopedBstr and
> ScopedNSAutoreleasePool yet scoped_nsobject. We certainly don't have anything
> with both underline and intercaps. Talk to Mark the Style Maker.

Using scoped_aedesc because it's templatized.

http://codereview.chromium.org/3151011/diff/1/4
File chrome/browser/platform_util_mac.mm (right):

http://codereview.chromium.org/3151011/diff/1/4#newcode33
chrome/browser/platform_util_mac.mm:33: //  2. Silent no-op for unassociated
file types http://crbug.com/50263
On 2010/08/12 20:17:35, Avi wrote:
> Stick some delimiter after the text and before the URL; a colon?

Done.

http://codereview.chromium.org/3151011/diff/1/4#newcode87
chrome/browser/platform_util_mac.mm:87: typeFSRef,  // typeCode
On 2010/08/12 20:17:35, Avi wrote:
> typeFileURL? Avoids the FSRef conversion (see
> http://developer.apple.com/mac/library/technotes/tn/tn2022.html). All the same
> to me though.

I saw that, but decided against it due to the need to pack the CFURL raw data
manually.  Seems like more code just to use a new API when the old one works
just fine.

http://codereview.chromium.org/3151011/diff/1/4#newcode112
chrome/browser/platform_util_mac.mm:112: kAENoReply,  // sendMode
On 2010/08/12 20:17:35, Avi wrote:
> + kAEAlwaysInteract?

Done.

[Avi (use Gerrit), 2010-08-12 21:10:58]

Yay! Next up for you: APIs that require Pascal strings.

LGTM