[llvm-commits] updated patch for PR889

[Chris Lattner]

On Nov 29, 2007, at 1:57 AM, pawel kunio wrote:

> As to the tabs and 80 char limit, already applied.

Thanks!

> Regarding the
> patch, I'd say I'd rather stick
> to the assert(0 && "Unknown 'subtype' instruction in Value  
> destructor");
> The reason is two-fold, 1. Some of the Value inherited class could
> have escaped my searches
> and 2. We will be able to enforce more strictly the destroyThis method
> pattern in newly added
> Value inherited classes.
> What do You think?

I'm ok with either one.  Note that they will both enforce the  
property for new classes.  If you have:

   if (..)
   ...
   else if (...)
    ..
   else
     assert(0 && "bad");

This will assert if none of the if's above match.  Likewise, if you  
have:

   if (..)
     ...
   else
     cast<sometype>(this)

The cast in the 'else' case will abort if 'this' doesn't match sometype.

-Chris