Support for interactive set-chrome-as-default in Windows.

chrome/browser/shell_integration_win.cc

Index: chrome/browser/shell_integration_win.cc
diff --git a/chrome/browser/shell_integration_win.cc b/chrome/browser/shell_integration_win.cc
index f4c8609f43ec88bc8c84e97eb715f7ec009cafb8..6665e4dc083420dfdfc3c877b125a3010ad05f07 100644
--- a/chrome/browser/shell_integration_win.cc
+++ b/chrome/browser/shell_integration_win.cc
@@ -425,8 +425,15 @@ bool ActivateApplication(const string16& app_id) {
 
 }  // namespace
 
-bool ShellIntegration::CanSetAsDefaultBrowser() {
-  return BrowserDistribution::GetDistribution()->CanSetAsDefault();
+ShellIntegration::DefaultSettingsChangePermission
+    ShellIntegration::CanSetAsDefaultBrowser() {
+  if (!BrowserDistribution::GetDistribution()->CanSetAsDefault())
+    return CHANGE_DEFAULT_NOT_ALLOWED;
+
+  if (base::win::GetVersion() >= base::win::VERSION_WIN8)
+    return CHANGE_DEFAULT_INTERACTIVE;
+  else
+    return CHANGE_DEFAULT_UNATTENDED;
 }
 
 bool ShellIntegration::SetAsDefaultBrowser() {
@@ -471,6 +478,23 @@ bool ShellIntegration::SetAsDefaultProtocolClient(const std::string& protocol) {
   return true;
 }
 
+bool ShellIntegration::SetAsDefaultBrowserInteractive() {
+  FilePath chrome_exe;
+  if (!PathService::Get(base::FILE_EXE, &chrome_exe)) {
+    NOTREACHED() << "Error getting app exe path";
+    return false;
+  }
+
+  BrowserDistribution* dist = BrowserDistribution::GetDistribution();
+  if (!ShellUtil::ShowMakeChromeDefaultSystemUI(dist, chrome_exe.value())) {
+    LOG(ERROR) << "Failed to launch the set-default-browser Windows UI.";

[gab, 2012/06/06 22:47:21]

LOG(DFATAL) is preferable here (as it will DCHECK on debug builds instead of
silently logging an error, but will still log an error on release builds).

[grt (UTC plus 2), 2012/06/07 02:50:36]

I'm not so sure about this.  Ideally, an integration test that checks to see if
the dialog appears when it's expected should be our way of detecting breakage
here rather than a DCHECK, which puts a burden on all Windows developers.  I say
this as a developer who has several DCHECKs disabled/removed in my local build
because they're well known and they get in my way.

[motek., 2012/06/07 14:58:50]

No code around here appears to do it in the same situation. Could you please
explain what makes this particular situation different? Intuitively, there is
nothing 'FATAL' to failing this operation, just like in any other case here.

[motek., 2012/06/07 14:58:50]

Ack.

[gab, 2012/06/07 15:18:54]

My thinking is that it's not 'FATAL' (and that's why we don't want to crash on
Release builds), but it's nice to know on Debug builds (say when Windows 9 comes
out) if we can't pop this anymore, i.e. we would hit a DCHECK as opposed to
having to dig to find the root of some weird bug.

On the other hand it might be more handy when Win9 comes out to just look at the
'ERROR's in the log and know all that we have to fix instead of fixing a DCHECK
only to fall on the next DCHECK.

I don't feel strongly about it, it's up to you.

+    return false;
+  }
+
+  VLOG(1) << "Set-as-default Windows UI triggered.";
+  return true;
+}
+
 ShellIntegration::DefaultWebClientState ShellIntegration::IsDefaultBrowser() {
   // When we check for default browser we don't necessarily want to count file
   // type handlers and icons as having changed the default browser status,