Thread overview
[Issue 8832] New: Segfault when accessing range returned by function that has delegate referencing local variables
Oct 26, 2012
Maxim Fomin
Dec 29, 2012
Walter Bright
Jan 29, 2013
Walter Bright
Jan 29, 2013
Walter Bright
Jan 29, 2013
Walter Bright
October 17, 2012
http://d.puremagic.com/issues/show_bug.cgi?id=8832

           Summary: Segfault when accessing range returned by function
                    that has delegate referencing local variables
           Product: D
           Version: D2
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody@puremagic.com
        ReportedBy: hsteoh@quickfur.ath.cx


--- Comment #0 from hsteoh@quickfur.ath.cx 2012-10-16 17:18:50 PDT ---
import std.algorithm;
import std.range;
import std.stdio;

auto boo() {
    auto C = [2];
    return [1,1].map!((a) => C).joiner;
}

void main() {
    writeln(boo().take(12));
}

This code will either segfault or produce nonsensical output. Replacing (a)=>C with (a)=>[2] makes the problem go away; shortening [1,1] to [1] also makes the problem go away, and removing .joiner also makes the problem go away. Removing the .map makes the problem go away too.

The problem is suspected to be the delegate (a)=>C which references the local
variable C, which goes out of scope when boo() returns. For whatever reason,
dmd isn't emitting code to allocate the delegate's context on the heap, causing
a crash when writeln() tries to read the second element off the range. (I'm not
sure why it doesn't crash with the first element. Maybe luck.)

Replacing the delegate with function(a)=>[1] makes the problem go away.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
October 17, 2012
http://d.puremagic.com/issues/show_bug.cgi?id=8832



--- Comment #1 from hsteoh@quickfur.ath.cx 2012-10-16 21:50:50 PDT ---
See also issue 7978, probably the same bug in a different context.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
October 18, 2012
http://d.puremagic.com/issues/show_bug.cgi?id=8832



--- Comment #2 from hsteoh@quickfur.ath.cx 2012-10-18 08:44:25 PDT ---
I think I may have figured out the cause of this bug. If boo() is modified as
follows:

auto boo() {
    auto C = [2];
    auto d = delegate(int) => C;
    return [1,1].map!d.joiner;
}

then everything works. Seems to me that the compiler is failing to pick up the reference to C *when the delegate is defined inside the compile-time parameter* to map. Hopefully this helps narrow it down enough to find the problem in the dmd code.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
October 26, 2012
http://d.puremagic.com/issues/show_bug.cgi?id=8832


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

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


--- Comment #3 from Maxim Fomin <maxim@maxim-fomin.ru> 2012-10-26 11:41:40 PDT ---
(In reply to comment #0)
> import std.algorithm;
> import std.range;
> import std.stdio;
> 
> auto boo() {
>     auto C = [2];
>     return [1,1].map!((a) => C).joiner;
> }
> 
> void main() {
>     writeln(boo().take(12));
> }
> 
> This code will either segfault or produce nonsensical output. Replacing (a)=>C with (a)=>[2] makes the problem go away; shortening [1,1] to [1] also makes the problem go away, and removing .joiner also makes the problem go away. Removing the .map makes the problem go away too.
> 
> The problem is suspected to be the delegate (a)=>C which references the local
> variable C, which goes out of scope when boo() returns. For whatever reason,
> dmd isn't emitting code to allocate the delegate's context on the heap, causing
> a crash when writeln() tries to read the second element off the range. (I'm not
> sure why it doesn't crash with the first element. Maybe luck.)
> 
> Replacing the delegate with function(a)=>[1] makes the problem go away.

The problem is in erroneous treating (a) => C of type void. If this is fixed
to:
- delegate(int a) { return C; }
- (int a) { return C; )
- (int a) => C;
everything works fine.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
December 29, 2012
http://d.puremagic.com/issues/show_bug.cgi?id=8832


Walter Bright <bugzilla@digitalmars.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bugzilla@digitalmars.com


--- Comment #4 from Walter Bright <bugzilla@digitalmars.com> 2012-12-29 00:41:00 PST ---
A clearer test case:

import std.algorithm;
import std.range;
import std.stdio;

auto boo() {
    auto C = [2];
    auto bar(int) { return C; }
    return [1,1].map!(bar).joiner;
}

void main() {
    writeln(boo().take(12));
}

The problem is that joiner!(MapResult!(bar,int[])) is not recognized as a
'local' template instantiation, even though MapResult!(bar,int[]) is marked as
local. Hence, boo() is not marked as needing a closure (case (4) in
FuncDeclaration::needsClosure()).

The fix listed for Issue 8863 correctly fixes this one. But there are other problems with that fix, as listed in 8863.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
January 29, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=8832



--- Comment #5 from Walter Bright <bugzilla@digitalmars.com> 2013-01-28 19:34:47 PST ---
Interestingly, Don's fix:

https://github.com/D-Programming-Language/dmd/pull/1554

fixes my clearer test case, but not the original.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
January 29, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=8832



--- Comment #6 from Walter Bright <bugzilla@digitalmars.com> 2013-01-28 21:09:49 PST ---
https://github.com/D-Programming-Language/dmd/pull/1575

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
January 29, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=8832



--- Comment #7 from github-bugzilla@puremagic.com 2013-01-29 11:54:38 PST ---
Commits pushed to master at https://github.com/D-Programming-Language/dmd

https://github.com/D-Programming-Language/dmd/commit/63bba9d02ad4ff783c10c543c2c9ea4e91248873
fix Issue 8832 - Segfault when accessing range returned by function that has
delegate referencing local variables

https://github.com/D-Programming-Language/dmd/commit/fe7583317aefdc63e5dc233a627296c5df5594e3 Merge pull request #1575 from WalterBright/b43

fix Issue 8832 - Segfault when accessing range returned by function that...

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
January 29, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=8832


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: -------