Linux: Refactor X11DesktopHandler

chrome/browser/ui/startup/startup_browser_creator.cc

Index: chrome/browser/ui/startup/startup_browser_creator.cc
diff --git a/chrome/browser/ui/startup/startup_browser_creator.cc b/chrome/browser/ui/startup/startup_browser_creator.cc
index 639d5b2cb93b32b163a02154077598946e5bb364..b23b416af325ecba0a5c4974102758d32c88df30 100644
--- a/chrome/browser/ui/startup/startup_browser_creator.cc
+++ b/chrome/browser/ui/startup/startup_browser_creator.cc
@@ -103,7 +103,7 @@
 #endif
 
 #if defined(OS_LINUX) && !defined(OS_CHROMEOS)
-#include "ui/views/widget/desktop_aura/x11_desktop_handler.h"
+#include "ui/events/platform/x11/x11_event_source.h"
 #endif
 
 #if defined(ENABLE_PRINT_PREVIEW)
@@ -711,8 +711,9 @@ bool StartupBrowserCreator::ProcessCmdLineImpl(
     uint64_t time;
     std::string switch_value =
         command_line.GetSwitchValueASCII(switches::kWmUserTimeMs);
-    if (base::StringToUint64(switch_value, &time)) {
-      views::X11DesktopHandler::get()->set_wm_user_time_ms(
+    if (base::StringToUint64(switch_value, &time) &&
+        ui::X11EventSource::HasInstance()) {
+      ui::X11EventSource::GetInstance()->SetLastSeenServerTime(

[danakj, 2016/08/11 21:48:19]

I'm a bit unsure about why we're using the "window launch time" as the "last
seen server time". This would confuse things if we wanted to get a current time
from the server, cuz this is the time where the window was requested to be
created, which is different.

I'll continue reading the CL to learn more but I'm a bit skeptical this is what
we want here.

[Tom (Use chromium acct), 2016/08/15 18:42:45]

It's because metacity/mutter/muffin all deny activation requests if a timestamp
is old.
https://github.com/GNOME/mutter/blob/5ee0f24ab90d457f3b53a32e96f89db816a98f0a...

Actually, they discard the request if it's older than the _NET_WM_USER_TIME of
ANY other window.
The EWMH spec says "The timestamp is Client's last user activity timestamp (see
_NET_WM_USER_TIME) at the time of the request" about the _NET_ACTIVE_WINDOW
client message.  So I guess these WMs have a bug that all date back to 2005 when
this commit was made in metacity
https://github.com/GNOME/mutter/commit/d11681e5050b9ddf37d211f09c07557bef374277 
As a side note, compiz actually gets this right and checks only against the
window's _NET_WM_USER_TIME
https://github.com/compiz-reloaded/compiz/blob/a7a23334283cbe4572061a86842c83...

You bring up an interesting point in another comment:  we should add
X11EventSource::GetTimestamp() to return the time of the event that's currently
being dispatched, and fall back on UpdateLastSeenServerTime if there's no event
being processed, or if the current event doesn't have a timestamp.  This should
make these misbehaving WMs happy, and we can remove the SetLastSeenServerTime
logic.  I think this would belong in a separate refactoring CL.

           static_cast<Time>(time));
     }
   }