[Bug c++/95226] New: Faulty aggregate initialization of vector with struct with float

[Bug c++/95226] New: Faulty aggregate initialization of vector with struct with float

[fboranek at atlas dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

            Bug ID: 95226
           Summary: Faulty aggregate initialization of vector with struct
                    with float
           Product: gcc
           Version: 8.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: fboranek at atlas dot cz
  Target Milestone: ---

# Steps to reproduce

//prog.cc

#include <vector>
#include <iostream>

struct T {
    uint32_t a;
    float b {.8};
};

int main()
{
    T t = {1};
    std::vector<T> tt = {{1}, {2}};

    std::cout << t.a << ", " << t.b << std::endl;
    std::cout << tt[0].a << ", " << tt[0].b << std::endl;
    std::cout << tt[1].a << ", " << tt[1].b << std::endl;
}

// g++ prog.cc -Wall -Wextra -std=c++17

For gcc 9.3.0, 9.2.0, 9.1.0, 8.1.0 the behavior is correct and output of the
program is:

1, 0.8
1, 0.8
1, 0.8

For gcc 8.3.0, 8.2.0 the output of the program is:

1, 0.8
1, -1.58819e-23
1072273817, 2.8026e-45

The second object in the vector is wrongly initialized.

The code can be fixed by changing of initialization of float T::b to {.8f}, {0}
or {}. It seems that only initialization by double is affected.
I used wandbox.org to test diffrent version of gcc and prevous and newer
version are OK. It sound like regression in 8.3.0 and 8.2.0, but 8.3.0 is in a
current stable Debian, which worried me.


# Used gcc from Debina Buster (current stable)
gcc (Debian 8.3.0-6) 8.3.0

[Bug c++/95226] Faulty aggregate initialization of vector with struct with float

[fboranek at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

--- Comment #1 from Frantisek Boranek <fboranek at gmail dot com> ---
Luckily, the behaviour is the same on old stable Stretch as on Buster. So this
version is also affected, but it is probably not serious as was my first
impression.

gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |8.1.0, 9.1.0, 9.2.0, 9.3.0
            Summary|Faulty aggregate            |[8 Regression] Faulty
                   |initialization of vector    |aggregate initialization of
                   |with struct with float      |vector with struct with
                   |                            |float
   Target Milestone|---                         |8.5

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
                 CC|                            |jakub at gcc dot gnu.org
   Target Milestone|8.5                         |9.0
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The GCC 8 branch is being closed, fixed in GCC 9.1.

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu.org,
                   |                            |mpolacek at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Seems on the 8 branch this regressed with
r8-8138-gbdc2c1ea35c16d3bbd3711430d8035dd54cfcf20
change aka PR85873 fix.
Slightly simplified testcase:

#include <vector>

struct T {
  unsigned a;
  float b {8.};
};

int main()
{
  T t = {1};
  std::vector<T> tt = {{1}, {2}};
  if (t.a != 1 || t.b != 8.0f || tt[0].a != 1 || tt[0].b != 8.0f || tt[1].a !=
2 || tt[1].b != 8.0f)
    __builtin_abort ();
}

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

--- Comment #4 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:4206171605de65df9674a14dd9db75bf4f4ed037

commit r12-801-g4206171605de65df9674a14dd9db75bf4f4ed037
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Fri May 14 16:29:49 2021 +0200

    testsuite: Add testcase for already fixed PR [PR95226]

    2021-05-14  Jakub Jelinek  <jakub@redhat.com>

            PR c++/95226
            * g++.dg/cpp1y/pr95226.C: New test.

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Jason Merrill <jason@gcc.gnu.org>:

https://gcc.gnu.org/g:e7cc4d703bceb9095316c106eba0d1939c6c8044

commit r14-1498-ge7cc4d703bceb9095316c106eba0d1939c6c8044
Author: Jason Merrill <jason@redhat.com>
Date:   Thu Jun 1 14:41:07 2023 -0400

    varasm: check float size

    In PR95226, the testcase was failing because we tried to output_constant a
    NOP_EXPR to float from a double REAL_CST, and so we output a double where
    the caller wanted a float.  That doesn't happen anymore, but with the
    output_constant hunk we will ICE in that situation rather than emit the
    wrong number of bytes.

    Part of the problem was that initializer_constant_valid_p_1 returned true
    for that NOP_EXPR, because it compared the sizes of integer types but not
    floating-point types.  So the C++ front end assumed it didn't need to fold
    the initializer.

            PR c++/95226

    gcc/ChangeLog:

            * varasm.cc (output_constant) [REAL_TYPE]: Check that sizes match.
            (initializer_constant_valid_p_1): Compare float precision.

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

--- Comment #6 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:470f501f31a4bdb9fa04c691ca7db2915ac3ae5b

commit r12-10136-g470f501f31a4bdb9fa04c691ca7db2915ac3ae5b
Author: Jason Merrill <jason@redhat.com>
Date:   Thu Jun 1 14:41:07 2023 -0400

    varasm: check float size [PR109359]

    In PR95226, the testcase was failing because we tried to output_constant a
    NOP_EXPR to float from a double REAL_CST, and so we output a double where
    the caller wanted a float.  That doesn't happen anymore, but with the
    output_constant hunk we will ICE in that situation rather than emit the
    wrong number of bytes.

    Part of the problem was that initializer_constant_valid_p_1 returned true
    for that NOP_EXPR, because it compared the sizes of integer types but not
    floating-point types.  So the C++ front end assumed it didn't need to fold
    the initializer.

    This also fixed the test for PR109359.

            PR c++/95226
            PR c++/109359

    gcc/ChangeLog:

            * varasm.cc (output_constant) [REAL_TYPE]: Check that sizes match.
            (initializer_constant_valid_p_1): Compare float precision.

    gcc/testsuite/ChangeLog:

            * g++.dg/ext/frounding-math1.C: New test.

    (cherry picked from commit e7cc4d703bceb9095316c106eba0d1939c6c8044)

[Bug c++/95226] [8 Regression] Faulty aggregate initialization of vector with struct with float

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95226

--- Comment #7 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:9b8e82ab45d1ad976a824cfd7c9bd2640c8bc8e3

commit r13-8282-g9b8e82ab45d1ad976a824cfd7c9bd2640c8bc8e3
Author: Jason Merrill <jason@redhat.com>
Date:   Thu Jun 1 14:41:07 2023 -0400

    varasm: check float size [PR109359]

    In PR95226, the testcase was failing because we tried to output_constant a
    NOP_EXPR to float from a double REAL_CST, and so we output a double where
    the caller wanted a float.  That doesn't happen anymore, but with the
    output_constant hunk we will ICE in that situation rather than emit the
    wrong number of bytes.

    Part of the problem was that initializer_constant_valid_p_1 returned true
    for that NOP_EXPR, because it compared the sizes of integer types but not
    floating-point types.  So the C++ front end assumed it didn't need to fold
    the initializer.

    This also fixed the test for PR109359.

            PR c++/95226
            PR c++/109359

    gcc/ChangeLog:

            * varasm.cc (output_constant) [REAL_TYPE]: Check that sizes match.
            (initializer_constant_valid_p_1): Compare float precision.

    gcc/testsuite/ChangeLog:

            * g++.dg/ext/frounding-math1.C: New test.

    (cherry picked from commit e7cc4d703bceb9095316c106eba0d1939c6c8044)