[Bug c/110458] New: -Warray-bounds=2 new false positive

[Bug c/110458] New: -Warray-bounds=2 new false positive

[sirl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

            Bug ID: 110458
           Summary: -Warray-bounds=2 new false positive
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sirl at gcc dot gnu.org
  Target Milestone: ---

Created attachment 55412
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55412&action=edit
testcase

Since somewhere between r14-1870 and r14-2097 a new -Warray-bounds=2 false
positive is shown for this little testcase:

typedef struct {
  unsigned arr1[4];
  unsigned arr2[4];
} data;

void f_notok(void *arrayOut) {
  int i;
  unsigned *arr2;
  unsigned *arr1;
  data *dataOut;
  dataOut = (data *)arrayOut;
  arr1 = dataOut[0].arr1;
  arr2 = dataOut[0].arr2;
  i = 0;
  for (; i < 4; i++) {
    arr1[i] = 0;
    arr2[i] = 0;
  }
}

When compiled with trunk@r2097 "gcc -O2 -W -Wall -Warray-bounds=2 -c
bug-Warray-bounds-eq-2.c" the warning is:

bug-Warray-bounds-eq-2.c: In function 'f_notok':
bug-Warray-bounds-eq-2.c:16:13: warning: '__builtin_memset' offset [16, 31]
from the object at 'arrayOut' is out of the bounds of referenced subobject
'arr1' with type 'unsigned int[4]' at offset 0 [-Warray-bounds=]
   16 |     arr1[i] = 0;
      |     ~~~~~~~~^~~
bug-Warray-bounds-eq-2.c:2:12: note: subobject 'arr1' declared here
    2 |   unsigned arr1[4];
      |            ^~~~

gcc-13.1.1 and earlier didn't warn here. The attached full testcase also shows
that slight variations in the code silence the warning.

[Bug tree-optimization/110458] -Warray-bounds=2 new false positive

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |tree-optimization
           Keywords|                            |diagnostic

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The problem is ldist does:

  arr1_8 = &MEM[(struct data *)arrayOut_7(D)].arr1;
  __builtin_memset (arr1_8, 0, 32);

(which is kinda of correct).

I think there are other bug related to this similar thing ...

[Bug tree-optimization/110458] -Warray-bounds=2 new false positive

[sirl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

--- Comment #2 from Franz Sirl <sirl at gcc dot gnu.org> ---
This has been exposed by commit
r14-2013-gfb0447b1f6b7373f57cb3a3d17a46803cfd9909d "Hide IVOPTs strip_offset".

[Bug tree-optimization/110458] [14 Regression] -Warray-bounds=2 new false positive

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|-Warray-bounds=2 new false  |[14 Regression]
                   |positive                    |-Warray-bounds=2 new false
                   |                            |positive
   Target Milestone|---                         |14.0

[Bug tree-optimization/110458] [14 Regression] -Warray-bounds=2 new false positive

[sirl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

--- Comment #3 from Franz Sirl <sirl at gcc dot gnu.org> ---
Actually Comment 2 is only true for the original testcode (which was quite
fragile to reproduce). The reduced testcode started to fail with the backwards
jump threader rewrite in r12-2591-g2e96b5f14e4025691b57d2301d71aa6092ed44bc and
a simple -Warray-bounds. This is proven by the fact that compiling the testcase
with gcc@r12-2591 with an additional --param=threader-mode=legacy makes the
warning go away.

So this bug maybe a duplicate of the other bugs (couldn't find a metabug)
related to r12-2591.

[Bug tree-optimization/110458] [14 Regression] -Warray-bounds=2 new false positive

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
                 CC|                            |law at gcc dot gnu.org

[Bug tree-optimization/110458] [14/15 Regression] -Warray-bounds=2 new false positive

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110458

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|14.0                        |14.2

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 14.1 is being released, retargeting bugs to GCC 14.2.