https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84141
--- Comment #16 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
This is strange.
Looking at the small reproducer and grepping for the settings to dtype
with
diff -b -u repro.f90.old repro.f90.new |grep 'dtype.*='
where repro.f90.old is the output of -fdump-tree-original with r257064
and repro.f90.new for r257131, the difference is
- (*(struct vamp_equivalence_t[0:] * restrict)
eq->eq.data)[eq->eq.offset + (integer(kind=8)) *i].permutation.dtype = 265;
+ (*(struct vamp_equivalence_t[0:] * restrict)
eq->eq.data)[eq->eq.offset + (integer(kind=8)) *i].permutation.dtype =
{.elem_len=4, .rank=1, .type=1};
- eq->eq.dtype = 4137;
+ eq->eq.dtype = {.elem_len=72, .rank=1, .type=5};
- eq->permutation.dtype = 265;
+ eq->permutation.dtype = {.elem_len=4, .rank=1, .type=1};
- n_bin.dtype = 266;
+ n_bin.dtype = {.elem_len=4, .rank=2, .type=1};
- desc.53.dtype = 1064;
+ desc.53.dtype = {.elem_len=8, .rank=0, .type=11};
- perm.dtype = 265;
+ perm.dtype = {.elem_len=4, .rank=1, .type=1};
- atmp.61.dtype = 265;
+ atmp.61.dtype = {.elem_len=4, .rank=1, .type=1};
- atmp.70.dtype = 265;
+ atmp.70.dtype = {.elem_len=4, .rank=1, .type=1};
- atmp.73.dtype = 265;
+ atmp.73.dtype = {.elem_len=4, .rank=1, .type=1};
- atmp.75.dtype = 265;
+ atmp.75.dtype = {.elem_len=4, .rank=1, .type=1};
- channel.dtype = 3625;
+ channel.dtype = {.elem_len=64, .rank=1, .type=5};
- (*(struct phs_channel_t[0:] * restrict)
channel.data)[channel.offset + (integer(kind=8)) ch].eq.dtype = 4137;
+ (*(struct phs_channel_t[0:] * restrict)
channel.data)[channel.offset + (integer(kind=8)) ch].eq.dtype = {.elem_len=72,
.rank=1, .type=5};
- eq->perm.dtype = 265;
+ eq->perm.dtype = {.elem_len=4, .rank=1,
.type=1};
all of which looks plausible.
