[Bug rtl-optimization/94189] New: [9/10 Regression] -fcompare-debug failure Wstrict-overflow-22.c

[Bug rtl-optimization/94189] New: [9/10 Regression] -fcompare-debug failure Wstrict-overflow-22.c

[zhroma at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

            Bug ID: 94189
           Summary: [9/10 Regression] -fcompare-debug failure
                    Wstrict-overflow-22.c
           Product: gcc
           Version: 9.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhroma at gcc dot gnu.org
                CC: jakub at gcc dot gnu.org
  Target Milestone: ---

This command:
gcc -c -O2 -fcompare-debug gcc/testsuite/gcc.dg/Wstrict-overflow-22.c

fails on master, 9.2 and 9.3, but not in 8.4 or earlier versions.

--
Great thanks to Jakub for looking at PR94166 and PR94167!  Unfortunately, those
patches actually fixed only the tests I have reported and thus doesn't decrease
much the amount of broken tests in -fcompare-debug regtest.

Here is the next one, an it seems to be a real ".text" difference in this case,
not only final-dump.  Brief investigation shows difference since ".expand"
dumps, while tree dumps look fine.

[Bug rtl-optimization/94189] [9/10 Regression] -fcompare-debug failure Wstrict-overflow-22.c

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
On which target?  Can't reproduce on x86_64-linux nor powerpc64le-linux.

[Bug rtl-optimization/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c

[zhroma at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

Roman Zhuykov <zhroma at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[9/10 Regression]           |[9/10 Regression]
                   |-fcompare-debug failure     |-fcompare-debug failure on
                   |Wstrict-overflow-22.c       |Wstringop-overflow-22.c

--- Comment #2 from Roman Zhuykov <zhroma at gcc dot gnu.org> ---
I've screwed up too, sorry :) It should be
gcc/testsuite/gcc.dg/Wstringop-overflow-22.c, file is rather new, available
only in master.  Just rechecked it fails on master on powerpc64le and using
9.2-9.3 on x86_64.

[Bug rtl-optimization/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
const char a[] = { 'a', 'b', 'c', 'd' };

int
foo (void)
{
  return __builtin_strnlen (a, 5);
}

and this one doesn't generate different code between -O2 -g and -O2, but FAILs
with -fcompare-debug because that one does add -w to the second invocation in
order not to emit all the warnings twice, and apparently the code generation in
this case is dependent on whether the warning is emitted or not.

[Bug middle-end/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2020-03-16
           Priority|P3                          |P2
     Ever confirmed|0                           |1
            Summary|[9/10 Regression]           |[9/10 Regression]
                   |-fcompare-debug failure on  |-fcompare-debug failure on
                   |Wstringop-overflow-22.c     |Wstringop-overflow-22.c
                   |                            |since r9-3242
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
   Target Milestone|---                         |9.4

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r9-3242-g6c4aa5f6bd8aacf51ddceb78239a7a2da0a1b959.

[Bug middle-end/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 48038
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48038&action=edit
gcc10-pr94189-1.patch

One possible fix, optimize even if we emit a warning.

[Bug middle-end/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 48039
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48039&action=edit
gcc10-pr94189-2.patch

Another possible fix.  Don't optimize if we would emit a warning and don't emit
it just because the warning isn't enabled, -w or some other warning has been
emitted for the call already.  I think this is my preference.

[Bug middle-end/94189] [9/10 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:7afa3b82918a75a486aad7818f11df9ea7504368

commit r10-7206-g7afa3b82918a75a486aad7818f11df9ea7504368
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Mar 17 10:42:35 2020 +0100

    expand: Don't depend on warning flags in code generation of strnlen
[PR94189]

    The following testcase FAILs with -O2 -fcompare-debug, but the reason isn't
    that we'd emit different code based on -g or non-debug, but rather that
    we emit different code depending on whether -w is used or not (or e.g.
    -Wno-stringop-overflow or whether some other pass emitted some other
warning
    already on the call).

    Code generation shouldn't depend on whether we emit a warning or not if at
    all possible.

    The following patch punts (i.e. doesn't optimize the strnlen call to a
    constant value) if we would emit the warning if it was enabled.
    In the PR there is an alternate patch which does optimize the strnlen call
    no matter if we emit the warning or not, though I think I prefer the
version
    below, e.g. the strnlen call might be crossing field boundaries, which is
in
    strict reading undefined, but I'd be afraid people do that in the real
    world programs.

    2020-03-17  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/94189
            * builtins.c (expand_builtin_strnlen): Do return NULL_RTX if we
would
            emit a warning if it was enabled and don't depend on
TREE_NO_WARNING
            for code-generation.

            * gcc.dg/pr94189.c: New test.

[Bug middle-end/94189] [9 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[9/10 Regression]           |[9 Regression]
                   |-fcompare-debug failure on  |-fcompare-debug failure on
                   |Wstringop-overflow-22.c     |Wstringop-overflow-22.c
                   |since r9-3242               |since r9-3242

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk so far.

[Bug middle-end/94189] [9 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:65de83595faeccd83bc0fefbfb79768f8a3bb2b6

commit r9-8398-g65de83595faeccd83bc0fefbfb79768f8a3bb2b6
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Mar 17 10:42:35 2020 +0100

    expand: Don't depend on warning flags in code generation of strnlen
[PR94189]

    The following testcase FAILs with -O2 -fcompare-debug, but the reason isn't
    that we'd emit different code based on -g or non-debug, but rather that
    we emit different code depending on whether -w is used or not (or e.g.
    -Wno-stringop-overflow or whether some other pass emitted some other
warning
    already on the call).

    Code generation shouldn't depend on whether we emit a warning or not if at
    all possible.

    The following patch punts (i.e. doesn't optimize the strnlen call to a
    constant value) if we would emit the warning if it was enabled.
    In the PR there is an alternate patch which does optimize the strnlen call
    no matter if we emit the warning or not, though I think I prefer the
version
    below, e.g. the strnlen call might be crossing field boundaries, which is
in
    strict reading undefined, but I'd be afraid people do that in the real
    world programs.

    2020-03-17  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/94189
            * builtins.c (expand_builtin_strnlen): Do return NULL_RTX if we
would
            emit a warning if it was enabled and don't depend on
TREE_NO_WARNING
            for code-generation.

            * gcc.dg/pr94189.c: New test.

[Bug middle-end/94189] [9 Regression] -fcompare-debug failure on Wstringop-overflow-22.c since r9-3242

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94189

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed for 9.4+ too.