Merge lp://qastaging/~smspillaz/unity/unity.fix_943456 into lp://qastaging/unity

This is the same as my fix, which I was about to propose. Except that this one fixes many more key combos. Not just Alt+F1/Alt+F2.

Oh, nice. This should fix bug 943223 too:

25 panel_controller_->StartFirstMenuShow();
26 - return false;
27 + return true;

My previous comment is a *maybe*.

Somehow seems to cause an issue when re-focusing a terminal/sometimes other windows - the terminal gets pushed to an edge, either top or bottom in my experience, confirmed through neil also if you need testers/can't reproduce.

A bit of debugging i noticed that if i removed the PluginAdapter::restoreInputFocus() call from HUDs hide method the issue no longer happens. of course the previously focused window no longer recieves focus when the hud closes if this method is removed.

I also get the same as Gord, but additionally sometimes the global menu gets stuck in "showing" mode, so even switching apps, I can still see the global menu on the panel (File Edit View ...)

I locally merged this branch with current trunk, compiled and ran it. This is the outcome of the test...

Bugs actually fixed by this branch:
 #943456
 #943239
 #943194

Bugs still occuring running this branch:
 #943223

> Somehow seems to cause an issue when re-focusing a terminal/sometimes other
> windows - the terminal gets pushed to an edge, either top or bottom in my
> experience, confirmed through neil also if you need testers/can't reproduce.
>
> A bit of debugging i noticed that if i removed the
> PluginAdapter::restoreInputFocus() call from HUDs hide method the issue no
> longer happens. of course the previously focused window no longer recieves
> focus when the hud closes if this method is removed.

This isn't an issue with this branch in particular, but with some other part of the focus code, so it needs to be fixed separately.

OK. I've tested this branch too.

I've unlinked the Alt+F10 bug that this branch doesn't fix.

Also showing the menus works in gnome-terminal, but not in Qt, XUL apps, or chromium for me now. I did notice a compiz-core branch from Daniel over the weekend that may well impact those apps too. We should check ASAP.

I think this branch is worth landing as is, as it fixes expected interaction between compiz and unity. Yes there may well be focus issues, but that is separate, and shouldn't block this landing.

I am wondering if we'll still need the forceKeyboardUngrab for the hud reveal with the compiz fix.

forceKeyboardUngrab is to work around bug 943194. I think we do need it for now, at least until a fix is done for compiz-core (none yet). I will be having another look at bug 943194 and all the Alt issues this week.

WARNING:
I suspect the forceKeyboardUngrab hack might cause a potentially serious regression. It looks like holding Alt for more than 250 milliseconds will break the grab and disable tap/non-tap detection. So if someone takes more than 250ms to press Alt+something, then that will be detected as a spurious Alt tap and trigger the HUD.

I hope I'm wrong...

Nevermind. I just noticed this which makes the hack safe:

  if (current_time - last_hud_show_time_ > (local::ALT_TAP_DURATION * 1000))
  {
    LOG_DEBUG(logger) << "Tap too long";
    return false;
  }

175 + g_timeout_add(local::ALT_TAP_DURATION, forceKeyboardUngrab, (gpointer) this);

This is causing the regression described by bug #948227 :/

FYI: Fixed in lp:~3v1n0/unity/alt+tab-escape-regression-fix ;)