llvm dev - Jan 2015 - [LLVMdev] Heads up, nuking a feature of ValueHandle

ValueHandle has this lovely feature:

  // A subclass may want to store some information along with the value
  // pointer. Allow them to do this by making the value pointer a
pointer-int
  // pair. The 'setValPtrInt' and 'getValPtrInt' methods below
give them
this
  // access.
  PointerIntPair<Value*, 2> VP;

It was originally used by Metadata, but thanks to Duncan's work, it is no
longer needed and has no in-tree usage. IT also adds a lot of complexity to
the whole thing. I'm planning to remove it.

Initially, I thought it might be likely that folks are using this
out-of-tree and was worried about removing it. But Richard and others have
convinced me that I'm probably wrong. I don't really see how it could be
used. Sending this thread so you can shout if this breaks you, and i'll
quickly re-instate it so we can figure out what to do.

-Chandler
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150108/359e786b/attachment.html>

> On 2015-Jan-08, at 16:02, Chandler Carruth <chandlerc at google.com>
wrote:
> 
> ValueHandle has this lovely feature:
> 
>   // A subclass may want to store some information along with the value
>   // pointer. Allow them to do this by making the value pointer a
pointer-int
>   // pair. The 'setValPtrInt' and 'getValPtrInt' methods
below give them this
>   // access.
>   PointerIntPair<Value*, 2> VP;
> 
> It was originally used by Metadata, but thanks to Duncan's work, it is
no longer needed and has no in-tree usage. IT also adds a lot of complexity to
the whole thing. I'm planning to remove it.
> 
> Initially, I thought it might be likely that folks are using this
out-of-tree and was worried about removing it. But Richard and others have
convinced me that I'm probably wrong. I don't really see how it could be
used. Sending this thread so you can shout if this breaks you, and i'll
quickly re-instate it so we can figure out what to do.
> 
> -Chandler
I think this was only around for MDNodeOperands.  Nuke it :).