Is CHAR_TO_BYTE considered to have no side-effects?

Is CHAR_TO_BYTE considered to have no side-effects?

[Juanma Barranquero]

The following patch removes two unused local variables from
current_column_1 and Fmove_to_column, but is is conditional on supposing
CHAR_TO_BYTE has no side effects.

In fact it can have, has CHAR_TO_BYTE hides a call to
buf_charpos_to_bytepos, which in turn can potentially make things like
creating (short-lived) markers for caching, etc.

So, can CHAR_TO_BYTE be considered, interface-wise, as side-effect free?


                                                           /L/e/k/t/u




Index: indent.c
===================================================================
RCS file: /cvs/emacs/src/indent.c,v
retrieving revision 1.148
diff -u -r1.148 indent.c
--- indent.c	17 Apr 2002 14:19:00 -0000	1.148
+++ indent.c	17 Apr 2002 14:20:46 -0000
@@ -512,7 +512,7 @@
   /* Start the scan at the beginning of this line with column number 0.  */
   register int col = 0;
   int scan, scan_byte;
-  int next_boundary, next_boundary_byte;
+  int next_boundary;
   int opoint = PT, opoint_byte = PT_BYTE;
 
   scan_newline (PT, PT_BYTE, BEGV, BEGV_BYTE, -1, 1);
@@ -520,7 +520,6 @@
   scan = PT, scan_byte = PT_BYTE;
   SET_PT_BOTH (opoint, opoint_byte);
   next_boundary = scan;
-  next_boundary_byte = scan_byte;
 
   if (tab_width <= 0 || tab_width > 1000) tab_width = 8;
 
@@ -540,7 +539,6 @@
 	    goto endloop;
 	  if (scan != old_scan)
 	    scan_byte = CHAR_TO_BYTE (scan);
-	  next_boundary_byte = CHAR_TO_BYTE (next_boundary);
 	}
 
       /* Check composition sequence.  */
@@ -937,8 +935,7 @@
   int prev_col = 0;
   int c = 0;
   int next_boundary;
-
-  int pos_byte, next_boundary_byte;
+  int pos_byte;
 
   if (tab_width <= 0 || tab_width > 1000) tab_width = 8;
   CHECK_NATNUM (column);
@@ -948,7 +945,6 @@
   pos_byte = PT_BYTE;
   end = ZV;
   next_boundary = pos;
-  next_boundary_byte = PT_BYTE;
 
   /* If we're starting past the desired column,
      back up to beginning of line and scan from there.  */
@@ -968,7 +964,6 @@
 	  pos = skip_invisible (pos, &next_boundary, end, Qnil);
 	  if (pos != prev)
 	    pos_byte = CHAR_TO_BYTE (pos);
-	  next_boundary_byte = CHAR_TO_BYTE (next_boundary);
 	  if (pos >= end)
 	    goto endloop;
 	}

Re: Is CHAR_TO_BYTE considered to have no side-effects?

[Richard Stallman]

    So, can CHAR_TO_BYTE be considered, interface-wise, as side-effect free?

For most purposes, yes.  Unless the surrounding code operates directly
on the marker data structures in a way that might be fooled by 
creating new markers, it is ok.