https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110914
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |jakub at gcc dot gnu.org
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 55813
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55813&action=edit
gcc14-pr110914.patch
Untested fix.
The bogus change in the above mentioned commit was the:
if (olddsi != NULL
- && tree_fits_uhwi_p (len)
&& !integer_zerop (len))
- adjust_last_stmt (olddsi, stmt, false);
+ {
+ maybe_warn_overflow (stmt, len, rvals, olddsi, false, true);
+ adjust_last_stmt (olddsi, stmt, false);
+ }
part. I haven't analyzed what exactly maybe_warn_overflow does, it is some
warning stuff and perhaps can be called when len is not constant, but the
previous guarding of adjust_last_stmt was completely intentional.
As adjust_last_stmt function comment says:
If the last .MEM setter statement before STMT is
memcpy (x, y, strlen (y) + 1), the only .MEM use of it is STMT
and STMT is known to overwrite x[strlen (x)], adjust the last memcpy to
just memcpy (x, y, strlen (y)). SI must be the zero length
strinfo.
so obviously the fact that memcpy (the second one) doesn't have last argument
constant 0 doesn't mean that it is non-zero length memcpy, we only know it
either if it is constant non-zero length, or variable where say value-range
could prove it is not zero. We have an adjust_last_stmt call later in the
function which handles length of strlen (x) + 1 though. So, this patch just
reverts the guard of that function back to what it was before.
