Account for minimum tabstrip width in opaque browser frame.

chrome/browser/ui/views/frame/opaque_browser_frame_view.cc

Index: chrome/browser/ui/views/frame/opaque_browser_frame_view.cc
===================================================================
--- chrome/browser/ui/views/frame/opaque_browser_frame_view.cc	(revision 104341)
+++ chrome/browser/ui/views/frame/opaque_browser_frame_view.cc	(working copy)
@@ -302,6 +302,17 @@
       close_button_->GetMinimumSize().width();
 #endif
   min_size.set_width(std::max(min_size.width(), min_titlebar_width));
+
+  // Ensure that the minimum width is enough to hold a tab strip with minimum
+  // width at its usual insets.
+  if (browser_view_->IsTabStripVisible()) {
+    AbstractTabStripView* tabstrip = browser_view_->tabstrip();
+    int min_tabstrip_width = tabstrip->GetMinimumSize().width();
+    int min_tabstrip_area_width =
+        width() - GetBoundsForTabStrip(tabstrip).width() + min_tabstrip_width;

[Peter Kasting, 2011/10/06 20:31:33]

This worries me because it is flexible.  For example, it might allow a Chrome
window with no avatar showing to shrink to a small size, but then if we show an
avatar for multiprofile or incognito, the window will be too small.  It seems
like we can solve this by
(a) Moving this calculation up into the code just above that's already
accounting for the other things in the frame here -- we just need to know how
the tabstrip is spaced relative to the caption buttons, and
(b) Adding something about the avatar to that calculation as well.

+    min_size.set_width(std::max(min_tabstrip_area_width, min_size.width()));
+  }
+
   return min_size;
 }