[Bug c++/95638] New: Legit-looking code doesn't work with -O2

[Bug c++/95638] New: Legit-looking code doesn't work with -O2

[officesamurai at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

            Bug ID: 95638
           Summary: Legit-looking code doesn't work with -O2
           Product: gcc
           Version: 10.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: officesamurai at gmail dot com
  Target Milestone: ---

Created attachment 48718
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48718&action=edit
The failing code

The attached code produces wrong result when build with -O2 and without
-fno-strict-aliasing, but I don't believe it does anything illegal. The only
fishy thing it does is creating a pointer to T before an object of that type is
constructed at that location (in Storage::Storage and Storage::push_back), but
the pointer is just passed to the placement new, which should be fine according
to [basic.life] (because "...using the pointer as if the pointer were of type
void*, is well-defined").

------------------------
$ g++-10.1.0 -v
Using built-in specs.
COLLECT_GCC=g++-10.1.0
COLLECT_LTO_WRAPPER=/home/brd/soft/gcc-10.1.0/libexec/gcc/x86_64-pc-linux-gnu/10.1.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ./configure --prefix=/home/brd/soft/gcc-10.1.0
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 10.1.0 (GCC) 

$ g++-10.1.0 -O2 gcc10_aliasing_issue.cpp -o test -std=c++14 && ./test
Null!!!

$ g++-10.1.0 -O2 -fno-strict-aliasing gcc10_aliasing_issue.cpp -o test
-std=c++14 && ./test
OK
------------------------

P.S. The same code works fine with GCC 9 and earlier versions.

[Bug c++/95638] Legit-looking code doesn't work with -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amker at gcc dot gnu.org,
                   |                            |jakub at gcc dot gnu.org

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
All I can say is that bisection shows (at least when preprocessed with g++
8.3.1 first) that this changed behavior in
r10-7184-ge4e9a59105a81cdd6c1328b0a5ed9fe4cc82840e
No time to analyze if it is a bug in the code or on the GCC side.
CCing patch author.

[Bug c++/95638] Legit-looking code doesn't work with -O2

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
The code works with -flifetime-dse=1, so I bet there's some object that goes
out of life:
https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
https://gcc.gnu.org/gcc-6/porting_to.html#flifetime-dse

[Bug c++/95638] Legit-looking code doesn't work with -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #2)
> The code works with -flifetime-dse=1, so I bet there's some object that goes
> out of life:
> https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
> https://gcc.gnu.org/gcc-6/porting_to.html#flifetime-dse

The difference between -flifetime-dse=1 and the default is not in uses after
end of scope, but in relying on content being defined across start of the
constructor.

[Bug c++/95638] Legit-looking code doesn't work with -O2

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
I don't see anything obviously wrong with the code. Nothing seems to write to
the storage before the constructor, let alone rely on those writes being
preserved.

[Bug c++/95638] Legit-looking code doesn't work with -O2

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #5 from bin cheng <amker at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #1)
> All I can say is that bisection shows (at least when preprocessed with g++
> 8.3.1 first) that this changed behavior in
> r10-7184-ge4e9a59105a81cdd6c1328b0a5ed9fe4cc82840e
> No time to analyze if it is a bug in the code or on the GCC side.
> CCing patch author.

Thanks for ccing, I will look into it this WE.

[Bug c++/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |9.3.0
            Summary|Legit-looking code doesn't  |[10/11 Regression]
                   |work with -O2               |Legit-looking code doesn't
                   |                            |work with -O2
           Keywords|                            |wrong-code
   Target Milestone|---                         |10.2

[Bug c++/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #6 from bin cheng <amker at gcc dot gnu.org> ---
We call graphds_scc twice to break alias dependence, with alias dependence
edges skipped in the second call.  The code (both before and after
r10-7184-ge4e9a59105a81cdd6c1328b0a5ed9fe4cc82840e) tries to rectify post order
information after the second call, however it never gets it right.  Actually I
don't think it can be easily rectified (if possible).

Will test another patch which records/restores post order information for the
second call.

[Bug tree-optimization/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

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

https://gcc.gnu.org/g:2c0069fafb53ccb7a45a6815025dfcbd2882a36e

commit r11-1565-g2c0069fafb53ccb7a45a6815025dfcbd2882a36e
Author: Bin Cheng <bin.cheng@linux.alibaba.com>
Date:   Sat Jun 20 15:42:12 2020 +0800

    Record and restore postorder information in breaking alias sccs.

    gcc/
            PR tree-optimization/95638
            * tree-loop-distribution.c (pg_edge_callback_data): New field.
            (loop_distribution::break_alias_scc_partitions): Record and restore
            postorder information.  Fix memory leak.

    gcc/testsuite/
            PR tree-optimization/95638
            * g++.dg/tree-ssa/pr95638.C: New test.

[Bug tree-optimization/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So fixed on the trunk, waiting for 10 backport?

[Bug tree-optimization/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #9 from bin cheng <amker at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #8)
> So fixed on the trunk, waiting for 10 backport?

Sorry, https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95804 is also in this part
which I believe is related to this fix.  Will backport the full patch after
fixing 95804.

Thanks

[Bug tree-optimization/95638] [10/11 Regression] Legit-looking code doesn't work with -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638
Bug 95638 depends on bug 95804, which changed state.

Bug 95804 Summary: [11 Regression] ICE in generate_code_for_partition, at tree-loop-distribution.c:1323 since r11-1565-g2c0069fafb53ccb7
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95804

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

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |amker at gcc dot gnu.org
   Last reconfirmed|                            |2020-07-09
      Known to fail|                            |10.1.0
            Summary|[10/11 Regression]          |[10 Regression]
                   |Legit-looking code doesn't  |Legit-looking code doesn't
                   |work with -O2               |work with -O2
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |ASSIGNED
      Known to work|                            |11.0

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed on trunk sofar.

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #11 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Bin Cheng <amker@gcc.gnu.org>:

https://gcc.gnu.org/g:f737ca45bee4ea61571393e04495955aeb7d67ab

commit r10-8453-gf737ca45bee4ea61571393e04495955aeb7d67ab
Author: Bin Cheng <bin.cheng@linux.alibaba.com>
Date:   Sat Jun 20 15:42:12 2020 +0800

    Record and restore postorder information in breaking alias sccs.

    gcc/
            PR tree-optimization/95638
            * tree-loop-distribution.c (pg_edge_callback_data): New field.
            (loop_distribution::break_alias_scc_partitions): Record and restore
            postorder information.  Fix memory leak.

    gcc/testsuite/
            PR tree-optimization/95638
            * g++.dg/tree-ssa/pr95638.C: New test.

    (cherry picked from commit 2c0069fafb53ccb7a45a6815025dfcbd2882a36e)

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #12 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Bin Cheng <amker@gcc.gnu.org>:

https://gcc.gnu.org/g:dc7a8afce35eb8948b481b5bcb8d26124a267f55

commit r9-8730-gdc7a8afce35eb8948b481b5bcb8d26124a267f55
Author: Bin Cheng <bin.cheng@linux.alibaba.com>
Date:   Sat Jun 20 15:42:12 2020 +0800

    Record and restore postorder information in breaking alias sccs.

    gcc/
            PR tree-optimization/95638
            * tree-loop-distribution.c (pg_edge_callback_data): New field.
            (loop_distribution::break_alias_scc_partitions): Record and restore
            postorder information.  Fix memory leak.

    gcc/testsuite/
            PR tree-optimization/95638
            * g++.dg/tree-ssa/pr95638.C: New test.

    (cherry picked from commit 2c0069fafb53ccb7a45a6815025dfcbd2882a36e)

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.2                        |10.3

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10.2 is released, adjusting target milestone.

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

--- Comment #14 from bin cheng <amker at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #13)
> GCC 10.2 is released, adjusting target milestone.

Hmm, this should be fixed on GCC10/GCC9.  I backported PR95638/PR95804
separately using cherry-pick, so the backport information for latter PR is not
reflected here.

[Bug tree-optimization/95638] [10 Regression] Legit-looking code doesn't work with -O2

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95638

bin cheng <amker at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
      Known to work|                            |10.2.0
         Resolution|---                         |FIXED

--- Comment #15 from bin cheng <amker at gcc dot gnu.org> ---
Confirmed fixed for 10.2.0 also.  Closing.