https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83704
Bug ID: 83704
Summary: pr31243 revisited
Product: gcc
Version: 8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: dominiq at lps dot ens.fr
CC: jb at gcc dot gnu.org
Target Milestone: ---
Revision r256284 makes the request in pr31243 obsolete for 64-bit mode.
On the positive side, -fcheck=bounds now works for the test in comment 9.
On the negative side, the following test compiles now but its behavior at run
time is not the one I expect:
subroutine foo(i)
integer(kind=8), intent(in) :: i
character(len=i) :: x
print *, i
print *, len(x, kind=8)
end subroutine foo
program main
character (len=2_8**32_8+4_8) :: ch
character (len=4) :: a1, a2
ch = '123456789'
a1 = ch(1:4)
ch(2_8**32_8:2_8**32_8+3_8) = 'abcd'
a2 = ch(2_8**32_8:2_8**32_8+4_8)
print *, "'",a1,"', '",a2,"'"
print *, "'", ch(1:2_8**32_8+3_8), "'"
print *, len(ch, kind=8)
print *, len(ch(1:2_8**32_8+3_8), kind=8)
call foo(2_8**32_8+4_8)
end program main
The output is
'1234', 'abcd'
'123'
4294967300
4294967299
Illegal instruction
i.e., print *, "'", ch(1:2_8**32_8+3_8), "'" prints only the first three
characters of the 'ch' substring and not the entire substring.
In addition, if I comment the line
print *, len(x, kind=8)
the Illegal instruction is no longer present in the output
'1234', 'abcd'
'123'
4294967300
4294967299
4294967300