digitalmars.D.bugs - [Issue 10789] New: Struct destructor erroneously called
- d-bugmail puremagic.com (24/24) Aug 09 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (7/7) Aug 09 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (40/40) Aug 10 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (13/13) Aug 17 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (11/20) Aug 17 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (10/10) Sep 03 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (11/11) Sep 03 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (10/10) Sep 03 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (20/20) Sep 29 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (10/22) Sep 29 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
- d-bugmail puremagic.com (14/14) Sep 29 2013 http://d.puremagic.com/issues/show_bug.cgi?id=10789
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Summary: Struct destructor erroneously called
Product: D
Version: D2
Platform: All
OS/Version: All
Status: NEW
Severity: normal
Priority: P2
Component: DMD
AssignedTo: nobody puremagic.com
ReportedBy: sludwig outerproduct.org
PDT ---
The attached program simulates a simple reference counted struct. 'fun' is
supposed to return a newly initialized S with a count of 1. Instead it calls
the destructor dropping the count to zero and then returns a copy of the
initialized struct. Leaving out the 'if' statment and returning the fresh 'S'
directly does not exibit this behavior.
This issue is critical as it has a high probability to indroduce hard to
detect/track down bugs when automatic reference counting is used.
Tested on DMD 2.063.2/Win32 and /Win64 and DMD master/Win32.
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 09 2013
d-bugmail puremagic.com http://d.puremagic.com/issues/show_bug.cgi?id=10789 PDT --- Created an attachment (id=1241) Reproduction case -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Aug 09 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Maxim Fomin <maxim maxim-fomin.ru> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |maxim maxim-fomin.ru
---
Reduced:
extern(C) int printf(const char*, ...);
struct S {
static int count;
this(int ignoreme)
{
int oldcount = count;
printf("%X ctor %d=>%d\n", &this, oldcount, ++count);
}
~this()
{
int oldcount = count;
printf("%X dtor %d=>%d\n", &this, oldcount, --count);
}
this(this)
{
int oldcount = count;
printf("%X postblit %d=>%d\n", &this, oldcount, ++count);
}
}
S fun()
{
S s1 = S(42), s2 = S(24);
if (true) return s1;
else return s2;
}
void main()
{
S s = fun();
}
In case of if(true) compiler does not insert postblit call.
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 10 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789 PDT --- I digged a little in the DMD sources and found a commit by Kenji Hara that at least affects this example and has a commented out code block that looks a little suspicious : https://github.com/D-Programming-Language/dmd/commit/b4bc25d72e601436f3999abc5c9c31e464385052#L4R1241 Changing "#if 0//DMDV2" back to "#if DMDV2" inserts a proper postblit call, but then the returned t has its initialized field set to false. This does not happen with the "#if 0" AFAICS. Unfortunately I know far to less about the mechanics at work there to make an informed attempt to fix this. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Aug 17 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Maxim Fomin <maxim maxim-fomin.ru> changed:
What |Removed |Added
----------------------------------------------------------------------------
Severity|normal |regression
---
I digged a little in the DMD sources and found a commit by Kenji Hara that at
least affects this example and has a commented out code block that looks a
little suspicious :
https://github.com/D-Programming-Language/dmd/commit/b4bc25d72e601436f3999abc5c9c31e464385052#L4R1241
Changing "#if 0//DMDV2" back to "#if DMDV2" inserts a proper postblit call, but
then the returned t has its initialized field set to false. This does not
happen with the "#if 0" AFAICS. Unfortunately I know far to less about the
mechanics at work there to make an informed attempt to fix this.
Sounds like a regression.
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Aug 17 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Kenji Hara <k.hara.pg gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |pull
https://github.com/D-Programming-Language/dmd/pull/2523
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Sep 03 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789 Commits pushed to master at https://github.com/D-Programming-Language/dmd https://github.com/D-Programming-Language/dmd/commit/4ca445bb2564997b80d5c00c6dfc1daeff1e30af fix Issue 10789 - Struct destructor erroneously called https://github.com/D-Programming-Language/dmd/commit/cfffc9b02fed9366babb6712cba7d6f26c18df6e [REG2.061] Issue 10789 - Struct destructor erroneously called -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Sep 03 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
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: -------
Sep 03 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Sönke Ludwig <sludwig outerproduct.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|RESOLVED |REOPENED
Resolution|FIXED |
PDT ---
The original test case still fails on DMD HEAD:
---
0018FD74 this() 0
0018FD75 this(this) 1
0018FD74 ~this() 2
0018FD9C ~this() 1
core.exception.AssertError app(47): Assertion failure
---
This is due to the last destructor running on an uninitialized instance
(initialized == false).
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Sep 29 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789The original test case still fails on DMD HEAD: --- 0018FD74 this() 0 0018FD75 this(this) 1 0018FD74 ~this() 2 0018FD9C ~this() 1 core.exception.AssertError app(47): Assertion failure --- This is due to the last destructor running on an uninitialized instance (initialized == false).To me it looks like that the original test case contains a bug. In S.this(this), `initialized` field is incorrectly set to false. It will stop to decrement S.count at the destruction of the copied objects. Therefore the last assertion in main fails because S.count == 1. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Sep 29 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10789
Sönke Ludwig <sludwig outerproduct.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution| |FIXED
PDT ---
Sorry, you are absolutely right. The "initialized = false" was supposed to go
to the destructor instead to test if the destructor is called twice on the same
instance. I'm still seeing a similar issue in my project, but I can't reproduce
it with the fixed this(this).
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Sep 29 2013