Bug 9993 - [3.2/3.3/3.4 regression] destructor not called for local object created within and returned from infinite loop
Summary: [3.2/3.3/3.4 regression] destructor not called for local object created withi...
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: c++ (show other bugs)
Version: 3.1
: P3 normal
Target Milestone: 3.1.x/3.2.x
Assignee: Jason Merrill
URL:
Keywords: wrong-code
Depends on:
Blocks:
 
Reported: 2003-03-07 20:36 UTC by jon_udell
Modified: 2004-07-13 17:53 UTC (History)
1 user (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed:


Attachments
temp.ii (80.25 KB, text/x-c++)
2003-05-21 15:17 UTC, jon_udell
Details

Note You need to log in before you can comment on or make changes to this bug.
Description jon_udell 2003-03-07 20:36:01 UTC
An object created in a do-while(1) loop for possible use as the return object of the function does not get destructed each time through the loop.

The erroneous output of the supplied program is:

Constructed with index 0
Constructed with index 1
Constructed with index 2
Destructed with index 2


The apparently correct output is:

Constructed with index 0
Destructed with index 0
Constructed with index 1
Destructed with index 1
Constructed with index 2
Copy constructed reusing index 2
Destructed with index 2
Destructed with index 2


One workaround is to explicitly return a dummy object after the loop even though the code will never be reached.  I'm guessing that with only the one return point inside the loop the compiler tries to create the object on the stack and return it without destructing it, forgetting to do the destruction when the return does not occur, and that putting in a second return point eliminates this optimization.

This problem does not exist on some previous versions of gcc that I am using, including:
gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-85) (i686)
gcc version 2.95.2 19991024 (release) (SunOS 5.8)

Release:
3.1

Environment:
i686-pc-linux-gnu

How-To-Repeat:
g++ temp.ii
a.out
Comment 1 Wolfgang Bangerth 2003-03-15 05:34:26 UTC
State-Changed-From-To: open->analyzed
State-Changed-Why: Indeed:
    ---------------------------
    extern "C" void printf (const char *, ...);
    #define PRINT printf ("%s\n", __PRETTY_FUNCTION__)
    
    struct Object {
        Object()                      { PRINT; }
        Object(const Object&)         { PRINT; }
        void operator=(const Object&) { PRINT; }
        ~Object()                     { PRINT; }
    };
    
    Object function() {
      int i = 0;
      do {
        Object b;
        if (i++ == 2)
          return b;
      } while (1);
    }
    
    int main() {
      function();
    }
    -----------------------------
    This is the output with 3.2/3.3/3.4:
    g/x> /home/bangerth/bin/gcc-3.4-pre/bin/c++ x.cc
    g/x> ./a.out
    Object::Object()
    Object::Object()
    Object::Object()
    Object::~Object()
    
    Ups. Not right. 2.95 generates
    g/x> ./a.out
    Object::Object()
    Object::~Object()
    Object::Object()
    Object::~Object()
    Object::Object()
    Object::Object(const Object &)
    Object::~Object()
    Object::~Object()
    
    So this is a regression.
    
    W.
Comment 2 Jason Merrill 2003-03-16 23:50:22 UTC
Responsible-Changed-From-To: unassigned->jason
Responsible-Changed-Why: nrvo
Comment 3 Jason Merrill 2003-03-17 15:43:47 UTC
From: jason@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org
Cc:  
Subject: c++/9993
Date: 17 Mar 2003 15:43:47 -0000

 CVSROOT:	/cvs/gcc
 Module name:	gcc
 Changes by:	jason@gcc.gnu.org	2003-03-17 15:43:47
 
 Added files:
 	gcc/testsuite/g++.dg/opt: nrv6.C 
 
 Log message:
 	PR c++/9993
 	* decl.c (finish_function): Only allow the NRVO to use variables
 	declared at function scope.
 
 Patches:
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/nrv6.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
 

Comment 4 Jason Merrill 2003-03-17 15:45:30 UTC
From: jason@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org
Cc:  
Subject: c++/9993
Date: 17 Mar 2003 15:45:30 -0000

 CVSROOT:	/cvs/gcc
 Module name:	gcc
 Changes by:	jason@gcc.gnu.org	2003-03-17 15:45:30
 
 Modified files:
 	gcc/cp         : ChangeLog decl.c 
 
 Log message:
 	PR c++/9993
 	* decl.c (finish_function): Only allow the NRVO to use variables
 	declared at function scope.
 
 Patches:
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3280&r2=1.3281
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&r1=1.1022&r2=1.1023
 

Comment 5 Jason Merrill 2003-03-17 15:46:38 UTC
From: jason@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org
Cc:  
Subject: c++/9993
Date: 17 Mar 2003 15:46:38 -0000

 CVSROOT:	/cvs/gcc
 Module name:	gcc
 Branch: 	gcc-3_2-branch
 Changes by:	jason@gcc.gnu.org	2003-03-17 15:46:38
 
 Modified files:
 	gcc/cp         : ChangeLog decl.c 
 
 Log message:
 	PR c++/9993
 	* decl.c (finish_function): Only allow the NRVO to use variables
 	declared at function scope.
 
 Patches:
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_2-branch&r1=1.2685.2.114.2.71&r2=1.2685.2.114.2.72
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&only_with_tag=gcc-3_2-branch&r1=1.866.2.36.2.17&r2=1.866.2.36.2.18
 

Comment 6 Jason Merrill 2003-03-17 15:47:39 UTC
From: jason@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org
Cc:  
Subject: c++/9993
Date: 17 Mar 2003 15:47:39 -0000

 CVSROOT:	/cvs/gcc
 Module name:	gcc
 Branch: 	gcc-3_3-branch
 Changes by:	jason@gcc.gnu.org	2003-03-17 15:47:39
 
 Modified files:
 	gcc/cp         : ChangeLog decl.c 
 
 Log message:
 	PR c++/9993
 	* decl.c (finish_function): Only allow the NRVO to use variables
 	declared at function scope.
 
 Patches:
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.3076.2.86&r2=1.3076.2.87
 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.965.2.31&r2=1.965.2.32
 
Comment 7 Jason Merrill 2003-03-17 18:08:26 UTC
State-Changed-From-To: analyzed->closed
State-Changed-Why: fixed