Re: [patch] proposal tiny cleanup for window code

[martin rudalics <rudalics@gmx.at>]

 > I've found that too_small_ok member of struct window isn't used anywhere
 > except size_window,
 > so it looks like there are no reasons to have too_small_ok within struct
 > window...

I fail to understand the use of that but the current code reads as

   if (old_size < min_size && nodelete_p != 2)
     w->too_small_ok = Qt;

   /* Move the following test here since otherwise the
      preceding test doesn't make sense.  martin. */
   if (nodelete_p == 2)
     nodelete_p = 0;

   /* Maybe delete WINDOW if it's too small.  */
   if (nodelete_p != 1 && !NILP (w->parent))
     {
       if (!MINI_WINDOW_P (w) && !NILP (w->too_small_ok))

hence too_small_ok is never set to t when nodelete_p initially equals 2.
With your change this becomes

   /* Move the following test here since otherwise the
      preceding test doesn't make sense.  martin. */
   if (nodelete_p == 2)
     nodelete_p = 0;

   /* Maybe delete WINDOW if it's too small.  */
   if (nodelete_p != 1 && !NILP (w->parent))
     {
       if (!MINI_WINDOW_P (w) && old_size < min_size && nodelete_p != 2)

hence (the "implied") too_small_ok can be set to t when nodelete_p
initially equals 2 and your version is not semantically equivalent to
the previous one.  Please verify again.