www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - [Issue 9386] New: struct destructor called erroneously

reply d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386

           Summary: struct destructor called erroneously
           Product: D
           Version: D2
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody puremagic.com
        ReportedBy: pisetta.gianni alice.it


--- Comment #0 from Gianni Pisetta <pisetta.gianni alice.it> 2013-01-24
07:26:39 PST ---
I had a nasty bug with the std.stdio.File type with the files i had opened and
immediately after closed.
I striped it down to a bug with the management of dynamic arrays of structs.
The code below allocate a dynamic array of Test struct, that output a line for
the constructor, postblit constructor and destructor, like the File struct.
The output shows that the destructors for each struct are called after
constructing the entire array. It isn't the behavior expected, since the struct
inside the dynamic array aren't garbage collected yet. I think that the more
appropriate behavior is inplace construction without destructor.

Here is the code:
--------------------------
import std.stdio;

struct Test {
    public string name;

    public this(string name) {
        this.name = name;
        writeln( "Created ", name, "..." );
    }

    public this(this) {
        writeln( "Copied ", this.name, "..." );
    }

    ~this() {
        writeln( "Deleted ", this.name );
    }
}

void main(string[] args)
{
    Test[] tests = [ Test( "one" ),
             Test( "two" ),
             Test( "three" ),
             Test( "four" ) ];

    foreach( Test test; tests ) {
        writeln( "Foreach ", test.name );
    }
}
--------------------------
And the output:
--------------------------
Created one...
Created two...
Created three...
Created four...
Deleted four
Deleted three
Deleted two
Deleted one
Copied one...
Foreach one
Deleted one
Copied two...
Foreach two
Deleted two
Copied three...
Foreach three
Deleted three
Copied four...
Foreach four
Deleted four
--------------------------

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Jan 24 2013
next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386



--- Comment #1 from Gianni Pisetta <pisetta.gianni alice.it> 2013-01-24
07:29:07 PST ---
I forgot to mention that I'm using dmd 2.061.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Jan 24 2013
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386


Maxim Fomin <maxim maxim-fomin.ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |maxim maxim-fomin.ru


--- Comment #2 from Maxim Fomin <maxim maxim-fomin.ru> 2013-01-24 08:21:27 PST
---
The problem is that dmd lowers the code to like below:

void main()
{
   Test _temp1;
   Test _temp2;
   Test _temp3;
   Test _temp4;
   Test[] tests = [ Test(), Test(), Test(), Test() ]; //these ctors are not
called
   _temp1.ctor("one"); 
   tests[0] = _temp1;
   _temp2.ctor("two");
   tests[1] = _temp2;
   _temp3.ctor("three");
   tests[2] = _temp3;
   _temp4.ctor("four");
   tests[3] = _temp4;
   _temp1.dtor();
   _temp2.dtor();
   _temp3.dtor();
   _temp4.dtor();
   ...
}

So, the first 8 lines of output comes from 4 stack temporaries, not array.
Perhaps you need to look at issue 9335 and issue 9334

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Jan 24 2013
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386



--- Comment #3 from Gianni Pisetta <pisetta.gianni alice.it> 2013-01-24
13:03:09 PST ---
Maybe I didn't explained well the problem in my first message.
In issue 9335 the problem is that dtors aren't called for structs in dynamic
arrays, which is fine for me because is the gc at collection that call the
dtors. In issue 9334 the problem is that postblit and dtors are called at
allocation of a dynamic array only if isn't called the default constructor, and
i don't know if it is good or bad.
But if you see the output, there aren't no lines that say "Copied ...." in the
output, so here we have the problem that the dtor is called without the
postblit, which invalidates any struct that performs the RAII idiom. If you
substitute my Test struct of the example in the first message with the File
struct of std.stdio module, you will have a Segmentation Fault in linux or a
exception FileNotOpened in windows, and it isn't clear what is going on because
the structs aren't collected by the gc, no one have called detach() on the
files and nobody suspects that the files in the dynamic array are actually been
closed at allocation time.
I think the best solution here is to have the struct initialized directly in
the heap, without the need to first allocate the struct in the stack, as
suggested from Maxim Fomin in issue 9334. Then no need to call postblit and
dtors, no overhead.
But if it isn't possible, then at least it must call the postblit after the
structs are copied in the heap and before the call to the dtors on the stack
structs, so that any resource remains retained by the struct.
The lowered code must be like this:

void main()
{
   Test[] tests = [ Test(), Test(), Test(), Test() ]; //these ctors are not
called
   tests[0].ctor("one"); 
   tests[1].ctor("two");
   tests[2].ctor("three");
   tests[3].ctor("four");
   ...
}

or like this:

void main()
{
   Test _temp1;
   Test _temp2;
   Test _temp3;
   Test _temp4;
   Test[] tests = [ Test(), Test(), Test(), Test() ]; //these ctors are not
called
   _temp1.ctor("one"); 
   tests[0] = _temp1;
   tests[0].postblit(); // call to postblit constructor
   _temp2.ctor("two");
   tests[1] = _temp2;
   tests[1].postblit(); // call to postblit constructor
   _temp3.ctor("three");
   tests[2] = _temp3;
   tests[2].postblit(); // call to postblit constructor
   _temp4.ctor("four");
   tests[3] = _temp4;
   tests[3].postblit(); // call to postblit constructor
   _temp1.dtor();
   _temp2.dtor();
   _temp3.dtor();
   _temp4.dtor();
   ...
}

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Jan 24 2013
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386


Kenji Hara <k.hara.pg gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |pull, wrong-code


--- Comment #4 from Kenji Hara <k.hara.pg gmail.com> 2013-04-08 22:28:42 PDT ---
https://github.com/D-Programming-Language/dmd/pull/1875

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Apr 08 2013
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386



--- Comment #5 from github-bugzilla puremagic.com 2013-04-09 16:43:39 PDT ---
Commits pushed to master at https://github.com/D-Programming-Language/dmd

https://github.com/D-Programming-Language/dmd/commit/49be239434d2f1d61d645498773d9525b539427c
fix Issue 9386 - struct destructor called erroneously

https://github.com/D-Programming-Language/dmd/commit/cfab27e4656a5d5f6fac578976a968f83df27cf0
Merge pull request #1875 from 9rnsr/fix9386

Issue 9386 - struct destructor called erroneously

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Apr 09 2013
prev sibling next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386


Walter Bright <bugzilla digitalmars.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |bugzilla digitalmars.com
         Resolution|                            |FIXED


-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Apr 09 2013
prev sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=9386


Kenji Hara <k.hara.pg gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dmitry.olsh gmail.com


--- Comment #6 from Kenji Hara <k.hara.pg gmail.com> 2013-05-09 02:12:44 PDT ---
*** Issue 8740 has been marked as a duplicate of this issue. ***

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
May 09 2013