------- Comment #4 from paul dot richard dot thomas at cea dot fr 2006-01-04 14:26 ------- (In reply to comment #2) > Actually assumed sized is not the full issue, see PR 18003.
You are right about this. Turn the declaration for a into an explicit full-array, rather than assumed size. I have seen, somewhere, a check for the rank of inquiry intrinsics, that uses the rank calculated from the arguments, if rank 0 is found for the function. I believe that the use of the scalarizer for intrinsic bounds is screwing itself up when tricked into treating the expression as scalarizable. This is confirmed by the work-around of taking the lbound out of the actual array list for bar(). I have taken "assumed size" out of the summary line. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25669