Side-effect of intelligent screen refresh code?

[YONETANI Tomokazu]

Hello.

Since this commit (and a few more commits related to it)
  e8d36bf10: Refresh cleverly to improve scrolling speed.

when I typed a multibyte character (which is also twice as wide as an ASCII
one), it's drawn as a blank and won't show up until the screen is redrawn.
With the recent code, this only happens when the screen is split, either
horizontally or vertically.  This is consistent across multiple OSes and
a few terminal emulators from which I usually attach to the screen, so I
poked here and there in the source code and found a minimal change which
apparently fix this issue (it's in LPutChar):

diff --git a/src/layer.c b/src/layer.c
index 4ae587f..e29313b 100644
--- a/src/layer.c
+++ b/src/layer.c
@@ -337,7 +337,7 @@ int x, y;
   if (l->l_pause.d)
     LayPauseUpdateRegion(l, x, x, y, y);
 
-  FOR_EACH_UNPAUSED_CANVAS(l,
+  for (cv = l->l_cvlist; cv; cv = cv->c_lnext)
     {
       display = cv->c_display;
       if (D_blocked)
@@ -354,7 +354,6 @@ int x, y;
          break;
        }
     }
-  );
 }
 
 void


Anyone else seeing this?