[Bug tree-optimization/66738] New: [5/6 Regression] optimizer bug related to exceptions and static symbols

[Bug tree-optimization/66738] New: [5/6 Regression] optimizer bug related to exceptions and static symbols

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

            Bug ID: 66738
           Summary: [5/6 Regression] optimizer bug related to exceptions
                    and static symbols
           Product: gcc
           Version: 5.1.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: doko at gcc dot gnu.org
  Target Milestone: ---

[forwarded from https://bugs.debian.org/788299]

seen with the gcc-5-branch 20150625 and trunk 20150610 (r224324)

$ g++-5 -g -O2 code.cpp && ./a.out 
Segmentation fault (core dumped)
$ g++-5 -g -O0 code.cpp && ./a.out 
terminate called after throwing an instance of 'MyException'
  what():  MyException: No!
Aborted (core dumped)
$ g++-4.9 -g -O2 code.cpp && ./a.out 
terminate called after throwing an instance of 'MyException'
  what():  MyException: No!
Aborted (core dumped)

$ cat code.cpp
#include <exception>
#include <string>

class MyException : public std::exception
{
protected:
    typedef std::exception Parent;

public:
    MyException(const char* cStr): Parent(), m_reason(cStr) {
        updateMessage();
    }

    virtual ~MyException() throw() {}

    virtual const std::string& exceptionName() const {
        return exceptionNameValue;
    }

    virtual const char* what() const throw() {
        return m_exceptionMessage.c_str();
    }

    inline void updateMessage() {
        m_exceptionMessage = exceptionName() + ": " + m_reason;
    }

private:
    std::string m_reason;
    std::string m_exceptionMessage;
    static const std::string exceptionNameValue;
};

const std::string MyException::exceptionNameValue("MyException");

int main() {
    throw MyException("No!");
    return 0;
}

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-07-02
                 CC|                            |hubicka at gcc dot gnu.org,
                   |                            |jamborm at gcc dot gnu.org
          Component|tree-optimization           |ipa
   Target Milestone|---                         |5.2
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed (with the old libstdc++ ABI).  -O1 works.  We segfault in libgcc:

Program received signal SIGSEGV, Segmentation fault.
_Unwind_Resume (exc=exc@entry=0x0)
    at /space/rguenther/tramp3d/trunk/libgcc/unwind.inc:229
229       if (exc->private_1 == 0)
(gdb) bt
#0  _Unwind_Resume (exc=exc@entry=0x0)
    at /space/rguenther/tramp3d/trunk/libgcc/unwind.inc:229
#1  0x0000000000400ad0 in main () at t.C:37

We optimize main to

Eh tree:
   1 cleanup
     3 cleanup land:{2,<L4>}

  <bb 2>:
  _3 = __cxa_allocate_exception (24);
  MEM[(struct MyException *)_3].D.12544._vptr.exception = &MEM[(void
*)&_ZTV11MyException + 16B];
  _7 = &MEM[(struct MyException *)_3].m_reason;
  std::basic_string<char>::basic_string (_7, "No!", &D.13102);
;;    succ:       3 [100.0%]  (FALLTHRU,EXECUTABLE)
;;                4 (EH,EXECUTABLE)

  <bb 3>:
  D.13102 ={v} {CLOBBER};
  MEM[(struct _Alloc_hider *)_3 + 16B]._M_p = &MEM[(void
*)&_S_empty_rep_storage + 24B];
  __builtin_unreachable ();

<L4>:
  D.13102 ={v} {CLOBBER};
  _18 = &MEM[(struct MyException *)_3].D.12544;
  std::exception::~exception (_18);
  __builtin_eh_copy_values (1, 3);
  __cxa_free_exception (_3);
  _9 = __builtin_eh_pointer (1);
  __builtin_unwind_resume (_9);


but the std::string constructor doesn't throw and thus we run into the
__builtin_unreachable ().

t.C.068t.fixup_cfg4:  __builtin_unreachable ();

is where that appears.  It's from devirtualization:

ipa-prop: Discovered a virtual call to a known target (void
MyException::updateMessage()/143 -> void __builtin_unreachable()/264), for stmt
_8 = OBJ_TYPE_REF(_6;this_2(D)->3) (this_2(D));
t.C:25:37: note: converting indirect call in void MyException::updateMessage()
to direct call to void __builtin_unreachable()
Speculative indirect call void MyException::updateMessage()/143 => virtual
const string& MyException::exceptionName() const/265 has turned out to have
contradicting known target __builtin_unreachable

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
-fno-devirtualize fixes it.

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> ---
For the record, this PR can be bisected to r218024:

Author: hubicka
Date:   Mon Nov 24 16:15:46 2014 +0000

        PR ipa/63671
        * ipa-inline-transform.c (can_remove_node_now_p_1): Handle alises
        and -fno-devirtualize more carefully.
        (can_remove_node_now_p): Update.

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|5.2                         |5.3

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 5.2 is being released, adjusting target milestone to 5.3.

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Mine

[Bug ipa/66738] [5/6 Regression] optimizer bug related to exceptions and static symbols

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66738

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
r218024 probably just uncovers symptoms. This is the same negative offset bug
as the other PR.

*** This bug has been marked as a duplicate of bug 67056 ***