issue with 2018 ISO_Fortran_binding.h

[Gilles Gouaillardet <gilles at rist dot or dot jp>]

Folks,


I tried gcc-9 snapshot from 20190331 and found quite severe issues with 
ISO_Fortran_binding.h.


I attached a tarball with a reproducer for both GNU and intel (2019) 
compilers, simply run 'make' in both gnu and intel directories.

I get the result I expect with Intel compiler regarding

- rank

- CFI_is_contiguous()

- base_addr


but with current GNU-9 snapshot, only rank is valid, base_addr is always 
incorrect and CFI_is_contigous() is not always correct.


Could you please review this test program and confirm there is indeed an 
issue ?



Best regards,


Gilles


result with intel 2019.3 compiler

x->rank = 0, CFI_is_contigous(x) = 1, x->base_addr = 0x7ffd1f984890, 
expected 0x7ffd1f984890
x->rank = 1, CFI_is_contigous(x) = 1, x->base_addr = 0x6b7430, expected 
0x6b7430
x->rank = 2, CFI_is_contigous(x) = 1, x->base_addr = 0x6b72a0, expected 
0x6b72a0
x->rank = 3, CFI_is_contigous(x) = 1, x->base_addr = 0x6b60a0, expected 
0x6b60a0


result with gnu-9 compiler

x->rank = 0, CFI_is_contigous(x) = 0, x->base_addr = 0x7fffa8733bec, 
expected 0x7fffa8734d4c
x->rank = 1, CFI_is_contigous(x) = 0, x->base_addr = 0x91d650, expected 
0x7fffa8734d20
x->rank = 2, CFI_is_contigous(x) = 0, x->base_addr = 0x91d680, expected 
0x7fffa8734b90
x->rank = 3, CFI_is_contigous(x) = 1, x->base_addr = 0x91d650, expected 
0x7fffa8733bf0

Attachment: f08ts.tar.gz
Description: GNU Zip compressed data