[Bug tree-optimization/103173] strncpy output may be truncated copying 32 bytes from a string of length 1439 (bogus) [-Werror=stringop-truncation]

["msebor at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103173

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |msebor at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-11-10
           Keywords|                            |diagnostic
             Blocks|                            |88781

--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
Confirmed.  The warning here is the result of the limits of the representation
GCC uses for the source argument.  The warning sees the following IL:

  strncpy (&dest, &MEM <char[32]> [(void *)&data], 32);

and uses the size of data as the upper bound on the length of the string stored
in it.  That obviously doesn't correspond to the source code which is more
closely represented by this IL:

  j = 0;
  _1 = &data[j].name;
  strncpy (&dest, _1, 32);

But as the IL changes from the original representation above to the &MEM thing
to make it easier for optimizers to work with, it loses that important detail
and causes the false positive.

The "fix" is to either try to undo the IL change and reconstruct the original
(which is never going to be perfect) or to run the warnings on the IL before it
loses the important detail.  The latter will mean that the warning will not
benefit from some of the optimizing transformations that depend on the MEM
representation.  All solutions involve trade-offs between false positives and
false negatives.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88781
[Bug 88781] [meta-bug] bogus/missing -Wstringop-truncation warnings