[PATCH] Fix C/43248, ICE with ?: and invalid compound literal

[Joseph S. Myers]

On Wed, 3 Mar 2010, Andrew Pinski wrote:

> Hi,
>   The problem here is that when we build the compound literal after an
> error has occurred for the initializer we still produce a compound
> literal.  This confuses the gimplifier later on when it tries to build
> a temporary for the ?: operator result.  The fix is just not to build
> the compound literal but just return error_mark_node as we do if the
> type is an error_mark_node.
> 
> OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com