On Fri, Nov 04, 2016 at 12:43:37PM +0100, Paul Richard Thomas wrote: > > The associate construct does not readily permit the identification of > the associate name as an array during parsing. However, this can be > done whilst matching rvalues within the associate block. This patch > extends this identification in gfc_match_varspec, whilst excluding > character types from the present test. This latter change prevents > scalar substrings from being pegged as array references, which was the > origin of this bug. > > Bootstraps and regtests on FC21/x86_64. > > OK for trunk? >
Looks good to me. -- Steve
