Blacklist low scale factors.

chrome/browser/ui/libgtk2ui/gtk2_ui.cc

Index: chrome/browser/ui/libgtk2ui/gtk2_ui.cc
diff --git a/chrome/browser/ui/libgtk2ui/gtk2_ui.cc b/chrome/browser/ui/libgtk2ui/gtk2_ui.cc
index ca74730bcb5f7e1e3944e145ee472f128f7eb7a3..21e7ce31e942bf51bac8b875fe5febf76ec7592b 100644
--- a/chrome/browser/ui/libgtk2ui/gtk2_ui.cc
+++ b/chrome/browser/ui/libgtk2ui/gtk2_ui.cc
@@ -1438,9 +1438,11 @@ float Gtk2UI::GetDeviceScaleFactor() const {
   if (gfx::Display::HasForceDeviceScaleFactor())
     return gfx::Display::GetForcedDeviceScaleFactor();
   const int kCSSDefaultDPI = 96;
-  float scale = GetDPI() / kCSSDefaultDPI;
+  const float scale = GetDPI() / kCSSDefaultDPI;
   // Round to 1 decimal, e.g. to 1.4.
-  return roundf(scale * 10) / 10;
+  const float rounded = roundf(scale * 10) / 10;
+  // See crbug.com/484400
+  return rounded < 1.3 ? 1.0 : rounded;

[mario.endlessm, 2016/01/12 15:07:12]

Came across this patch set today and found that this comparison is not working
as intended: As 'rounded' is a float variable and the 1.3 literal is a double,
the function is returning the 1.0 "adjusted" value in those cases where
'rounded' gets set to the 1.30000 (float) value.

Thus, a quick follow-up patch to fix this to be 'rounded < 1.3f ? 1.0f :
rounded' would be nice. I can do it myself, but as it's quite trivial I thought
I would just comment here first, just in case.

[Evan Stade, 2016/01/13 04:40:13]

Good catch.

Also, I'm not sure if this is intentional, but the CL description is slightly
inaccurate:

"For any device-scaling-factor < 1.3, a device-scaling-factor of 1 will
be enforced."

In fact, a scale factor of 1.25 will return 1.3 rather than enforcing 1 (modulo
the bug Mario points out).

[mario.endlessm, 2016/01/13 08:58:23]

I thought the same thing too and then understood it as if it was an error in the
description (which would perhaps mean "_rounded_ device-scaling-factor"), but
perhaps it's better to use floor() instead of round() here.

 }
 
 }  // namespace libgtk2ui