https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119827
Bug ID: 119827
Summary: Out of bounds check fails on substrings for upper
bound
Product: gcc
Version: 14.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: v2000kiara at gmail dot com
Target Milestone: ---
Hi all,
I encountered a possible failure of -fcheck=all to pick up a substring going
out of bounds for the upper index at runtime. Consider the following code:
program substring_bug
implicit none
character(len=12) :: mystr
integer :: i
mystr='Hello World!'
i = 0
! This does raises a runtime error
! print *, 'MYSTR1= ', mystr(i:i+1)
! This raises a compilation error
! print *, 'MYSTR2= ', mystr(13:13)
! This does not raise a compilation or runtime error
print *, 'MYSTR3= ', mystr(1:1-1)
! This will not raise an error either...
print *, 'MYSTR4= ', mystr(1:i-1) ! i-1=0
! ... nor does this...
i=-1
print *, 'MYSTR5= ', mystr(1:i)
end program substring_bug
Running:
gfortran -O0 -pedantic -Wall -Wconversion -Wcharacter-truncation
-Wdo-subscript -Wsurprising -Wunused -Wextra -fno-strict-aliasing -fwrapv
-fcheck=all
./a.out
gives the following output
MYSTR3=
MYSTR4=
MYSTR5=
The first two cases in the program are correctly picked up at compilation time
or runtime using -fcheck=all. The final three cases, which are out-of-bounds on
the upper index, result in an empty string being returned as seen in the
output.
I would expect a runtime error to be raised saying something along the lines of
"substring out of bounds: upper bound of 'mystr' is larger than <something>"
