[Bug tree-optimization/119148] Inconsistent -Wstringop-truncation warning when using strncpy

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119148

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2025-03-07

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
The diagnostic says that the copy truncates "123456" which is true.  The
diagnostic is emitted via gimple_fold_builtin_strncpy during CCP.

The difference seems to be that in once case we succeed in hitting

  /* Look for dst[i] = '\0'; after the stxncpy() call and if found
     avoid the truncation warning.  */
  gsi_next_nondebug (&gsi);
  gimple *next_stmt = gsi_stmt (gsi);
  if (!next_stmt) 

but in the other case not because in one case we have

 arr[l_5] = 0;

but in the other

 _1 = ptr_7 + l_6;
 *_1 = 0;

the code should possibly skip a stmt computing an offset but then it
would need to do more analysis on the deref.

The issue is we fold this quite early and CCP only substitutes/folds
the following stmts _after_ it, so this code doesn't really work reliably.