[Bug c++/57510] New: initializer_list memory leak

[Bug c++/57510] New: initializer_list memory leak

[matt at ookypooky dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

            Bug ID: 57510
           Summary: initializer_list memory leak
           Product: gcc
           Version: 4.7.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: matt at ookypooky dot com

Created attachment 30247
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30247&action=edit
Code demonstrating the problem

It looks as though initializer_list can suffer from memory leaks - if during
construction of the initializer_list members an exception is thrown then the
previously allocated members are not correctly destroyed. Unfortunately I do
not have access to a newer g++ version than 4.7.2, however I could not see any
mention of this issue in the closed bugs.

Code (also attached):
--
#include <memory>
#include <random>

struct X
{
    X () : m_n (std::unique_ptr<int> (new int)) { if (random () & 1) throw 1; }
    std::unique_ptr<int> m_n;
};

void foo (std::initializer_list<X>)
{
}

int main ()
{
  for (int i = 0; i < 10; ++i)
  {
    try
    {
      foo ({ X{}, X{} });
    }
    catch (...) {}
  }
}
--

$ g++ -v
Using built-in specs.
COLLECT_GCC=/software/thirdparty/gcc/4.7.2-0.el6_64/bin/g++
COLLECT_LTO_WRAPPER=/software/thirdparty/gcc/4.7.2-0.el6_64/libexec/gcc/x86_64-unknown-linux-gnu/4.7.2/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../configure --prefix=/software/thirdparty/gcc/4.7.2-0.el6_64
--with-system-zlib --enable-shared --enable-threads=posix --enable-laguages=all
Thread model: posix
gcc version 4.7.2 (GCC)

$ g++ -std=c++11 memory_leak.cpp

$ valgrind ./a.out
==13775== Memcheck, a memory error detector
==13775== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al.
==13775== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==13775== Command: ./a.out
==13775==
==13775==
==13775== HEAP SUMMARY:
==13775==     in use at exit: 12 bytes in 3 blocks
==13775==   total heap usage: 23 allocs, 20 frees, 1,244 bytes allocated
==13775==
==13775== LEAK SUMMARY:
==13775==    definitely lost: 12 bytes in 3 blocks
==13775==    indirectly lost: 0 bytes in 0 blocks
==13775==      possibly lost: 0 bytes in 0 blocks
==13775==    still reachable: 0 bytes in 0 blocks
==13775==         suppressed: 0 bytes in 0 blocks
==13775== Rerun with --leak-check=full to see details of leaked memory
==13775==
==13775== For counts of detected and suppressed errors, rerun with: -v
==13775== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 6 from 6)

[Bug c++/57510] initializer_list memory leak

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-06-03
     Ever confirmed|0                           |1

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Testcase without <memory and <random> and not requiring valgrind:

#include <initializer_list>

struct counter
{
  static int n;

  counter() { ++n; }
  counter(const counter&) { ++n; }
  ~counter() { --n; }
};

int counter::n = 0;

struct X
{
    X () { if (counter::n > 1) throw 1; }

    counter c;
};

int main ()
{
  try
  {
    auto x = { X{}, X{} };
  }
  catch (...)
  {
    if ( counter::n != 0 )
      throw;
  }
}

[Bug c++/57510] initializer_list memory leak

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
           Priority|P3                          |P2
                 CC|                            |jason at gcc dot gnu.org

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Thus I would say wrong code and P2.

[Bug c++/57510] initializer_list memory leak

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mmehlich at semanticdesigns dot co
                   |                            |m

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
*** Bug 57930 has been marked as a duplicate of this bug. ***

[Bug c++/57510] initializer_list memory leak

[tavianator at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

--- Comment #4 from Tavian Barnes <tavianator at gmail dot com> ---
I have a testing tool that automatically inserts operator new failures, to help
test exception safety and check for leaks.  This bug causes all kinds of
spurious failures that I have to work around, since anything like this

    vector<string> vec = {"some", "strings"};

leaks memory if the second string constructor fails.  Usually it can be worked
around like this

    string arr[] = {"some", "strings"};

but it's still quite annoying.

If someone can point me in the right direction or give an outline of how to fix
this bug, I'm happy to try and write up a patch myself.  Thanks!

[Bug c++/57510] initializer_list memory leak

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jason at gcc dot gnu.org

[Bug c++/57510] initializer_list memory leak

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

--- Comment #5 from Jason Merrill <jason at gcc dot gnu.org> ---
Author: jason
Date: Fri Dec 12 03:48:55 2014
New Revision: 218653

URL: https://gcc.gnu.org/viewcvs?rev=218653&root=gcc&view=rev
Log:
    PR c++/57510
    * typeck2.c (split_nonconstant_init_1): Handle arrays here.
    (store_init_value): Not here.
    (split_nonconstant_init): Look through TARGET_EXPR.  No longer static.
    * cp-tree.h: Declare split_nonconstant_init.
    * call.c (set_up_extended_ref_temp): Use split_nonconstant_init.

Added:
    trunk/gcc/testsuite/g++.dg/cpp0x/initlist90.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/call.c
    trunk/gcc/cp/cp-tree.h
    trunk/gcc/cp/typeck2.c

[Bug c++/57510] initializer_list memory leak

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED
   Target Milestone|---                         |5.0

--- Comment #6 from Jason Merrill <jason at gcc dot gnu.org> ---
Fixed for GCC 5.

[Bug c++/57510] initializer_list memory leak

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

--- Comment #7 from Jason Merrill <jason at gcc dot gnu.org> ---
Author: jason
Date: Fri Jan 23 16:30:00 2015
New Revision: 220047

URL: https://gcc.gnu.org/viewcvs?rev=220047&root=gcc&view=rev
Log:
    PR c++/64314
    PR c++/57510
    * typeck2.c (split_nonconstant_init_1): Remove a sub-CONSTRUCTOR
    that has been completely split out.

Added:
    trunk/gcc/testsuite/g++.dg/init/array38.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/typeck2.c

[Bug c++/57510] subobjects not destroyed when exception thrown during list-initialization

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510

Jason Merrill <jason at gcc dot gnu.org> changed:

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

--- Comment #14 from Jason Merrill <jason at gcc dot gnu.org> ---
Let's close this one and handle the remaining cases under PRs 52320 and 66139.