Re: [Pnet-developers] Re: back to Emit problems

[Michal Moskal]

On Fri, Apr 30, 2004 at 09:18:07AM +0200, Jonas Printzen wrote:
> 
> >I believe there are no guarantees about validity of code generated by
> >S.R.E. It is even possible to insert some junky opcodes by accident
> >by using wrong overload of Emit method (like Emit(Ldarg_s,
> >0xdeadf00ddeadf00dULL)). Neither Mono nor MS .NET catches this kind of
> >error during generation. And verification is far more complex then just
> >valid opcodes.
> 
> Hmm... pretty interresting. That would imply that with
> current Emit implementations one major argument for managed code
> is down the drain? If there is a way to actually execute junk... 

Managed code != S.R.E. Under Windows one need to have reflection
permission to actually generate and run code. And that ,,major
argument'' is that one can verify code before running it -- you can
generate assembly and the verify it.

S.R.E. is low level API. Plain and simple.

> But on the other hand, the argument probably gets truncated by the
> next Emit() or EndScope(). 

Nope. When using Emit(Ldarg_S, ((int)1)) you will easly see that

  ldarg.s 1
  nop
  nop
  nop

is generated. Opcode for nop is 0. Figure out the rest.

> After all, isn't the ILGenerator and
> others hiding the 'program pointer'? 

Emit() call simply puts bytes into method body.

-- 
: Michal Moskal :: http://www.kernel.pl/~malekith :: GCS !tv h e>+++ b++
: When in doubt, use brute force. -- Ken Thompson :: UL++++$ C++ E--- a?