[Ur] json_variant/destrR in meta leaving behind anonymous functions

[Adam Chlipala]

On 09/16/2012 10:23 PM, Edward Z. Yang wrote:
> Trace is below. Perhaps the inline thresholds have to be bumped up for
> this case?
>
> ezyang at javelin:~/Dev/logitext/meta/tests$ urweb testJson
> /home/ezyang/Dev/logitext/meta/variant.ur:96:24: (to 96:33) Anonymous function remains at code generation
>    

Fixed, by making a test for legality of inlining less conservative.

The latest Logitext from GitHub compiles (up until C compilation, at 
least) for me, with these two compiler changes.