As testing by Alessandro revealed, vector subscripts weren't properly
handled.
This patch fixes the compiler side (or at least those issues I found).
In particular, for expressions ("get") it wrongly passed a NULL
pointer, additionally, I used the wrong "ar". For it and for
assignments/push ("send", "sendget"), I also used the wrong rank value
as one also passes DIMEN_ELEMENT as DIMEN_RANGE.
Build and regtested on x86-64-gnu-linux.
OK for the trunk?
* * *
I still have to add vector subscript support to libcaf_single. I
didn't include an -fdump-tree-original test case, but I can add one if
there regarded as useful.
Attached is â besides the patch for trans-intrinsic.c â a debuging
patch for libcaf_single. I tested it with:
integer :: A(2,3)[*]
A(2,:) = A(1,[1,3,2])[1]
end
integer :: A(2,3)[*]
A(1,[1,3,2])[1] = A(2,:)
end
integer :: A(2,3)[*]
integer :: B(2,3)[*]
A(1,[1,3,2])[1] = B(1,[1,3,2])[1]
end
The output looks like (for the first one):
DEBUG: CAF_GET: 0x7fffb72f71d0
DEBUG: have vector for rank 2 [1]
DEBUG: dim=0: nvec = 0
DEBUG: (1:1:1)
DEBUG: dim=1: nvec = 3
DEBUG: 0: 1
DEBUG: 1: 3
DEBUG: 2: 2
Tobias