On 08/08/2012 02:47 PM, Paolo Carlini wrote:
AFAICS, the most uncertain case is the conditional operator test, otherwise we could split the PR.Turned out to be really trivial: a missing check for error_mark_node in build_conditional_expr_1.
I'll send in a separate message a complete regtested patch + testcase covering all the tests unrelated to instantiation dependent.
Paolo.