digitalmars.D.bugs - [Issue 9099] New: segfault while returning an atomicLoad value
- d-bugmail puremagic.com (39/39) Nov 30 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (34/34) Nov 30 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (16/16) Dec 26 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (7/7) Dec 26 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (19/28) Dec 27 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (10/10) Dec 28 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (9/9) Dec 29 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (8/8) Dec 29 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (9/9) Dec 29 2012 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (8/8) Jan 02 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9099
- d-bugmail puremagic.com (8/8) Feb 12 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9099
http://d.puremagic.com/issues/show_bug.cgi?id=9099
Summary: segfault while returning an atomicLoad value
Product: D
Version: D2
Platform: All
OS/Version: All
Status: NEW
Severity: regression
Priority: P2
Component: DMD
AssignedTo: nobody puremagic.com
ReportedBy: puneet coverify.org
---
Works with dmd-2.059. Does not work with the current github snapshot and with
dmd-2.060.
Strangely, if the "this" function is not defined for the Foo struct. The code
works fine -- no segfault. Alternatively if I store the value returned by
alomicLoad in a local variable before returning it from the function foo, the
code still runs fine.
// minimal code that gives segfault
import core.atomic;
struct Foo {
private ulong _s = 0;
public this (ulong s) { // no segfault if this function is commented out
this._s = s;
}
}
auto foo () {
shared static Foo bar;
// shared frop = atomicLoad(bar);
// return frop; // no segfault
return atomicLoad(bar); // segfaults here
}
void main() {
foo();
}
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Nov 30 2012
d-bugmail puremagic.com
http://d.puremagic.com/issues/show_bug.cgi?id=9099
---
This is related. The result printed out by the code on my Ubuntu amd64 machine
is
42
4294967296
Upto version dmd-2.059, it correctly prints 42 and 42.
When the "this" constructor for Foo is commented out again the results are
correctly printed out as 42 and 42.
//
import core.atomic;
struct Foo {
public ulong _s = 0;
// atomicLoad gives out wrong value when this (constructor) is defined
public this (ulong s) {
this._s = s;
}
}
auto foo () {
import std.stdio;
shared static Foo bar;
bar._s = 42;
writeln(bar._s);
atomicStore(bar, bar);
shared Foo frop = atomicLoad(bar);
writeln(frop._s);
return frop;
}
void main() {
foo();
}
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Nov 30 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099
Walter Bright <bugzilla digitalmars.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |bugzilla digitalmars.com
Component|DMD |druntime
02:52:19 PST ---
The problem is that atomicLoad() only works with POD (Plain Old Data) types.
Structs with constructors are not POD. Non-POD structs are returned with a
different calling convention than POD, hence the crash.
The fix is for atomicLoad() to reject attempts to instantiate it with non-POD
types.
The test case here should fail to compile.
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 26 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099 02:58:38 PST --- Unfortunately, there is no convenient isPOD trait detection, so I'm going to defer this for now. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Dec 26 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099
Dmitry Olshansky <dmitry.olsh gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |dmitry.olsh gmail.com
04:39:38 PST ---
The problem is that atomicLoad() only works with POD (Plain Old Data) types.
Structs with constructors are not POD.
It used to work.
Non-POD structs are returned with a
different calling convention than POD, hence the crash.
Then the definition of POD changed?
And besides what's the difference for the atomic load? AFAICT in D all structs
are PODs unless there is a postblit. There is no virtual table nor inheritance.
Even C++11 extends POD definition to include member functions. Why should we go
the opposite direction?
The fix is for atomicLoad() to reject attempts to instantiate it with non-POD
types.
Even if there was a clear definition of it in D...
The test case here should fail to compile.
... I expect atomicLoad to work with anything suitably sized w/o postblits and
opAssign.
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 27 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099 16:10:42 PST --- atomicLoad()'s implementation assumes that 8 byte structs are returned in EDX:EAX. This is not correct for non-POD structs, which are always returned via a hidden pointer to the return result on the caller's stack. There is no way for atomicLoad() to detect this. See Issue 9237. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Dec 28 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099 Commit pushed to master at https://github.com/D-Programming-Language/druntime https://github.com/D-Programming-Language/druntime/commit/a15d228ed0c1a3bfb5bd05ec1d491802bb29aac0 fix Issue 9099 - core.atomic.atomicLoad() cannot handle non-POD structs -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Dec 29 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099 Commit pushed to staging at https://github.com/D-Programming-Language/druntime https://github.com/D-Programming-Language/druntime/commit/878a183a51825ea5dce25f4289a8ef001f6eeea1 fix Issue 9099 - core.atomic.atomicLoad() cannot handle non-POD structs -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Dec 29 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099
Walter Bright <bugzilla digitalmars.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution| |FIXED
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 29 2012
http://d.puremagic.com/issues/show_bug.cgi?id=9099 Commit pushed to staging at https://github.com/D-Programming-Language/druntime https://github.com/D-Programming-Language/druntime/commit/a15d228ed0c1a3bfb5bd05ec1d491802bb29aac0 fix -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Jan 02 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9099 Commit pushed to master at https://github.com/D-Programming-Language/druntime https://github.com/D-Programming-Language/druntime/commit/878a183a51825ea5dce25f4289a8ef001f6eeea1 fix Issue 9099 - core.atomic.atomicLoad() cannot handle non-POD structs -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 12 2013