Re: Problem report #5

[Andreas Schwab <schwab@suse.de>]

"Stuart D. Herring" <herring@lanl.gov> writes:

>> 3140 	  prev = NULL;
>> 3141 	  for (tail = buf->overlays_before; tail; prev = tail, tail = next)
>> 3142 	    {
>> 3143 	      next = tail->next;
>> 3144 	      XSETMISC (overlay, tail);
>> 3145
>> 3146 	      /* If the overlay is not valid, get rid of it.  */
>> 3147 	      if (!OVERLAY_VALID (overlay))
>> [...]
>> 3161
>> 3162 	      beg = OVERLAY_START (overlay);
>> 3163 	      end = OVERLAY_END (overlay);
>> 3164
>> 3165 	      if (OVERLAY_POSITION (end) > pos)
>> 3166 		{
>> 3167 		  /* OVERLAY needs to be moved.  */
>> 3168 		  int where = OVERLAY_POSITION (beg);
>> 3169 		  struct Lisp_Overlay *other, *other_prev;
>> 3170
>> 3171 		  /* Splice the cons cell TAIL out of overlays_before.  */
>>
>> Event dead_error_condition: On this path, the condition "prev != 0" could
>> not be true
>> Also see events: [dead_error_line][assignment][const][const][assignment]
>>
>> 3172 		  if (prev)
>
> It's true that tail is never 0 within the loop (loop test), and that on
> every iteration but the first prev has been assigned tail, which wasn't 0
> at the time of assignment.  But on the first iteration, this test is
> guaranteed to pass!  No bug.

The point is that prev can never get non-NULL because later prev (while
still being NULL) is assigned to tail, or else the loop is exited right
away in the false arm of the conditional in line 3165.  But the comment
before the loop explains why it is written this way:

  /* We don't strictly need prev in this loop; it should always be nil.
     But we use it for symmetry and in case that should cease to be true
     with some future change.  */

So this is on purpose.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."