http://d.puremagic.com/issues/show_bug.cgi?id=10451

           Summary: Array of pointers to opaque struct gives forward
                    reference errors.
           Product: D
           Version: D2
          Platform: x86
        OS/Version: Windows
            Status: NEW
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody@puremagic.com
        ReportedBy: aldacron@gmail.com


--- Comment #0 from Mike Parker <aldacron@gmail.com> 2013-06-23 08:42:23 PDT ---
struct foo;

struct bar {
    foo*[] foos;
}

The error only occurs when the array is declared as a struct field. When declared as a class member or in a function, there is no error.

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451



--- Comment #1 from Mike Parker <aldacron@gmail.com> 2013-08-11 07:23:09 PDT ---
I've got some new info on this. The error also occurs outside of structs if the array is initialized:

=========
struct foo;

void main() {
    foo*[] foos = [];
}
=========

Here's the full error spew that I get in both cases:

==================
forref.d(3): Error: struct forrref.foo is forward referenced when looking for
't
oHash'
forref.d(3): Error: struct forrref.foo is forward referenced when looking for
'o
pCmp'
forref.d(3): Error: struct forrref.foo is forward referenced when looking for
't
oString'
forref.d(3): Error: struct forrref.foo unknown size
forref.d(3): Error: struct forrref.foo no size yet for forward reference
forref.d(3): Error: struct forrref.foo unknown size
forref.d(3): Error: struct forrref.foo no size yet for forward reference
===================

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451


Manu <turkeyman@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |turkeyman@gmail.com


--- Comment #2 from Manu <turkeyman@gmail.com> 2013-08-31 17:45:25 PDT ---
http://d.puremagic.com/issues/show_bug.cgi?id=10451

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451


Ali Cehreli <acehreli@yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |acehreli@yahoo.com


--- Comment #3 from Ali Cehreli <acehreli@yahoo.com> 2013-08-31 20:05:39 PDT ---
Manu meant this: :)

  http://d.puremagic.com/issues/show_bug.cgi?id=10766

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451



--- Comment #4 from Manu <turkeyman@gmail.com> 2013-08-31 23:00:46 PDT ---
(In reply to comment #3)
> Manu meant this: :)
> 
>   http://d.puremagic.com/issues/show_bug.cgi?id=10766

Haha, thx ;)

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451


Iain Buclaw <ibuclaw@ubuntu.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ibuclaw@ubuntu.com


--- Comment #5 from Iain Buclaw <ibuclaw@ubuntu.com> 2013-09-18 04:18:33 PDT ---
This happens because, eg: array literals are initialised by the GC by calling _d_arrayliteralTX (typeinfo, dim);

This requires that the struct has a complete runtime type information about a type.  Something that does not get generated for opaque types, which makes this bug not the easiest to fix - and the same is with other array operations which also require typeinfo to be available.

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451



--- Comment #6 from Manu <turkeyman@gmail.com> 2013-09-18 04:39:28 PDT ---
But the compiler doesn't need to know anything about the type to allocate an array of pointers. Surely the compiler can use/share the typeinfo of void* in this case?

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451



--- Comment #7 from Iain Buclaw <ibuclaw@ubuntu.com> 2013-09-18 05:41:37 PDT ---
(In reply to comment #6)
> But the compiler doesn't need to know anything about the type to allocate an array of pointers. Surely the compiler can use/share the typeinfo of void* in this case?

Sure we could use void* if is your expectation for typeid(foo*).toString to
return "void*"  =)

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451



--- Comment #8 from Manu <turkeyman@gmail.com> 2013-09-18 05:46:42 PDT ---
Surely it can't be hard to synthesize a named pointer type... unless it's dereferenced, it shouldn't need an actual definition in the typeinfo. So it would just be a stub or dummy typeinfo that's empty?

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

http://d.puremagic.com/issues/show_bug.cgi?id=10451


grogan.colin@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |grogan.colin@gmail.com


--- Comment #9 from grogan.colin@gmail.com 2013-09-18 05:48:42 PDT ---
A possible workaround to this is to create the struct like:

struct foo {}

Though, in this case it isn't an opaque struct and is merely a struct with no members.

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