I realize I am doing bad things, anyway...

I'd like to modify the vtbl inside ClassInfo. My goal is to stick a pointer inside each instance of a series of types, and I thought I'd reserve a slot in the vtbl and put the pointer in place of the function, like this:

import std.stdio;

class MethObject
{
  void methSlotTable() { }
}

int[] a_slots = [ 0, 1 ];

void main()
{
  int firstUserVFunc = 5;
  assert(MethObject.classinfo.vtbl[firstUserVFunc] == &MethObject.methSlotTable);
  writeln(MethObject.classinfo.vtbl);
  MethObject.classinfo.vtbl[firstUserVFunc] = cast(void*) a_slots.ptr;
}

Not so surprisingly, I get a segfault:
ldc2 -unittest --run abi.d
[624230, 40CC50, 40C9E0, 40CC70, 40CB70, 40CED0]
Error: abi failed with status: -2
Error: message: Segmentation fault (core dumped)
Error: program received signal 2 (Interrupt)

Is the vtbl mprotected? Is there a chance of making this work?

On 2017-06-27 00:19, Jean-Louis Leroy wrote:
> I realize I am doing bad things, anyway...
>
> I'd like to modify the vtbl inside ClassInfo. My goal is to stick a
> pointer inside each instance of a series of types, and I thought I'd
> reserve a slot in the vtbl and put the pointer in place of the function,
> like this:
>
> import std.stdio;
>
> class MethObject
> {
>    void methSlotTable() { }
> }
>
> int[] a_slots = [ 0, 1 ];
>
> void main()
> {
>    int firstUserVFunc = 5;
>    assert(MethObject.classinfo.vtbl[firstUserVFunc] ==
> &MethObject.methSlotTable);
>    writeln(MethObject.classinfo.vtbl);
>    MethObject.classinfo.vtbl[firstUserVFunc] = cast(void*) a_slots.ptr;
> }
>
> Not so surprisingly, I get a segfault:
> ldc2 -unittest --run abi.d
> [624230, 40CC50, 40C9E0, 40CC70, 40CB70, 40CED0]
> Error: abi failed with status: -2
> Error: message: Segmentation fault (core dumped)
> Error: program received signal 2 (Interrupt)
>
> Is the vtbl mprotected? Is there a chance of making this work?

It works for me on macOS using DMD. Might behave differently on different systems with different compilers.

-- 
/Jacob Carlborg

You could make it work by making a copy of the vtbl[], modifying it as you please, and seeing __vptr to it.

If you're trying to implement signals and slots, it's already in the library:

http://dlang.org/phobos/std_signals.html

On Saturday, 22 July 2017 at 01:18:12 UTC, Walter Bright wrote:
> You could make it work by making a copy of the vtbl[], modifying it as you please, and seeing __vptr to it.
>
> If you're trying to implement signals and slots, it's already in the library:
>
> http://dlang.org/phobos/std_signals.html

Thanks for the suggestion. I am working on open methods, not signals. So far I've hijacked the 'deallocator' field but now I am experimenting with a perfect hash of the vptr. Seems to work quite well...