Fix 28280 vla regression

[Mike Stump]

On Jul 7, 2006, at 6:06 PM, Andrew Pinski wrote:
> Instead of TREE_TYPE (e), it seems like you should use  
> void_type_node instead.

Nice catch, thanks.

While I'd rather have an appointed maintainer review this, none seems  
willing, and I hate each day that goes by with a regression that I  
caused in the tree, so I'm checking it in as obvious.  If a  
maintainer would care to review it, I'd be happy to further enhance  
it.  I feel the compiler is better off with this fixed in this way,  
than it just sitting there broken.  :-(

Tested on powerpc-apple-darwin8.5.0.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: vla-5.diffs.txt
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20060712/bfa6430b/attachment.txt>
-------------- next part --------------