[Bug c++/105937] New: [12] Regression with maybe-uninitialized

[Bug c++/105937] New: [12] Regression with maybe-uninitialized

[janisozaur+gcc at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105937

            Bug ID: 105937
           Summary: [12] Regression with maybe-uninitialized
           Product: gcc
           Version: 12.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: janisozaur+gcc at gmail dot com
  Target Milestone: ---

Created attachment 53125
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53125&action=edit
Test case extracted from original code

Commit https://gcc.gnu.org/g:6feb628a706e86eb3f303aff388c74bdb29e7381 "Improve
warning suppression for inlined functions [PR98512]. " introduced a regression
when using std::optional.

Ever since that commit, when building optimized code compiler complains:

/home/mijn/workspace/openrct2/src/openrct2/drawing/Drawing.String.cpp: In
function ‘void ttf_process_string_literal(OpenRCT2::IContext*,
std::string_view)’:
/home/mijn/workspace/openrct2/src/openrct2/drawing/Drawing.String.cpp:215:62:
error: ‘*(long unsigned int*)((char*)&ttfRunIndex + offsetof(std::optional<long
unsigned int>,std::optional<long unsigned
int>::<unnamed>.std::_Optional_base<long unsigned int, true,
true>::<unnamed>))’ may be used uninitialized in this function
[-Werror=maybe-uninitialized]
  215 |                 auto len = it.GetIndex() - ttfRunIndex.value();
      |                                                              ^


gcc -v:

Using built-in specs.
COLLECT_GCC=/home/mijn/workspace/gcc-install-master/bin/c++
COLLECT_LTO_WRAPPER=/home/mijn/workspace/gcc-install-master/libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc/configure
--prefix=/home/mijn/workspace/gcc-install-master --enable-languages=c++
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20210702 (experimental) (GCC)



Provided an extracted test case for the issue mentioned.
Original code:
https://github.com/OpenRCT2/OpenRCT2/blob/b0ffa9d28e62b862f68a37788931265cb4528835/src/openrct2/drawing/Drawing.String.cpp#L839
Original bug report for OpenRCT2:
https://github.com/OpenRCT2/OpenRCT2/issues/17371
Test case on godbolt: https://godbolt.org/z/6Y3cn1Ks1

c++ -O2 -c -o foo test.cpp -Werror=maybe-uninitialized

[Bug tree-optimization/105937] [12] Regression with maybe-uninitialized

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105937

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note your example no longer compiles on the trunk of GCC as it requires to add:

#include <stdint.h>

[Bug tree-optimization/105937] [12/13 Regression] maybe-uninitialized with std::optional

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105937

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |needs-reduction
   Target Milestone|---                         |12.2
            Summary|[12] Regression with        |[12/13 Regression]
                   |maybe-uninitialized         |maybe-uninitialized with
                   |                            |std::optional

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---

  # ttfRunIndex_64 = PHI <ttfRunIndex_6(52), [/app/example.cpp:207:27] 0(30)>
  # ttfRunIndex$8_65 = PHI <ttfRunIndex$8_8(52), [/app/example.cpp:207:27]
0(30)>
...


<L17>:
  [/app/example.cpp:213:13] if (ttfRunIndex$8_65 != 0)
    goto <bb 22>; [33.00%]
  else
    goto <bb 48>; [67.00%]

...

  <bb 22> [local count: 157594832]:
  [/app/example.cpp:216:62] len_15 = _20 - ttfRunIndex_64;

...
  # ttfRunIndex_6 = PHI <ttfRunIndex_14(D)(26), ttfRunIndex_64(48), _20(21),
ttfRunIndex_64(49)>
  # ttfRunIndex$8_8 = PHI <0(26), 0(48), [/app/example.cpp:227:29] 1(21),
1(49)>

So if anything this is both a missed optimization and the maybe uninitialized
predicates is not doing its job correctly.

[Bug tree-optimization/105937] [12 Regression] maybe-uninitialized with std::optional

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105937

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

https://gcc.gnu.org/g:627132c9bf9439bf0ac83bb092182055c1e0f3ab

commit r12-8818-g627132c9bf9439bf0ac83bb092182055c1e0f3ab
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Aug 19 14:12:52 2022 +0200

    tree-optimization/105937 - avoid uninit diagnostics crossing iterations

    The following avoids adding PHIs to the worklist for uninit processing
    if we reach them following backedges.  That confuses predicate analysis
    because it assumes the use is happening in the same iteration as the the
    definition.  For the testcase in the PR the situation is like

    void foo (int val)
    {
      int uninit;
      # val = PHI <..> (B)
      for (..)
        {
          if (..)
            {
              .. = val; (C)
              val = uninit;
            }
          # val = PHI <..> (A)
        }
    }

    and starting from (A) with 'uninit' as argument we arrive at (B)
    and from there at (C).  Predicate analysis then tries to prove
    the predicate of (B) (not the backedge) can prove that the
    path from (B) to (C) is unreachable which isn't really what it
    necessary - that's what we'd need to do when the preheader
    edge of the loop were the edge with the uninitialized def.

    So the following makes those cases intentionally false negatives.

            PR tree-optimization/105937
            * tree-ssa-uninit.cc (find_uninit_use): Do not queue PHIs
            on backedges.
            (execute_late_warn_uninitialized): Mark backedges.

            * g++.dg/uninit-pr105937.C: New testcase.

    (cherry picked from commit c77fae1ca796d6ea06d5cd437909905c3d3d771c)

[Bug tree-optimization/105937] [12 Regression] maybe-uninitialized with std::optional

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105937

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
      Known to fail|                            |12.2.0
         Resolution|---                         |FIXED
      Known to work|                            |12.2.1

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