Closed pawosm-arm closed 3 years ago
Just to keep you posted, @pawosm-arm : removing this line fixes the issue and does not break any of the publicly available tests. I will try to dig deeper and come up with a less intrusive work-around and some explanation of what is going on in that part of apply_use()
function.
Consider following code containing
sub2=> sub
rename list:This causes the following link-time issue:
The removal of the rename list as such:
...results in a successful linking and a correct
sub
symbol resolution:To make this problem even more complicated, just the removal of the
private
clause in themodule b
also solves the link time problem and results in a correctsub
symbol resolution.There was also somewhat symmetric issue observed. The removal of 1) the
private
clause in themodule b
and 2) theuse b
line (while leaving theuse c, sub2=> sub
line intact and still callingsub
instead ofsub2
) results in link-time error with gfortran, yet it results in a correct compilation and linkage (and a correctsub
symbol resolution) with classic flang! This only suggests that something is wrong with processing the rename lists in the calssic flang.