https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107874
--- Comment #21 from john.harper at vuw dot ac.nz ---
I now have a new test case that avoids the possibility of recursive I/O
by tstuff and fstuff doing internal writes to two different character
variables. It still reveals the merge problem. It compiles and runs with
gfortran-12 and ifort, giving different outputs.
! Must merge evaluate all 3 arguments?
program testmerge10
implicit none
character(7):: string(2) = ' '
logical:: x(2) = [.true., .false.], y
integer i
do i = 1,2
y = merge(tstuff(),fstuff(),x(i))
print *,y,string
string = ' '
end do
contains
logical function tstuff()
write(string(1),"(A)") ' tstuff'
tstuff = .true.
end function tstuff
logical function fstuff()
write(string(2),"(A)") ' fstuff'
fstuff = .false.
end function fstuff
end program testmerge10
Good luck with your bughunt!
John H
On Fri, 2 Dec 2022, anlauf at gcc dot gnu.org wrote:
> Date: Fri, 2 Dec 2022 21:05:43 +0000
> From: anlauf at gcc dot gnu.org <gcc-bugzi...@gcc.gnu.org>
> To: John Harper <john.har...@vuw.ac.nz>
> Subject: [Bug fortran/107874] merge not using all its arguments
> Resent-Date: Sat, 3 Dec 2022 10:05:53 +1300 (NZDT)
> Resent-From: <john.har...@vuw.ac.nz>
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107874
>
> --- Comment #19 from anlauf at gcc dot gnu.org ---
> (In reply to john.harper from comment #18)
>> An interesting problem! But I thought my original test case did not have
>> recursive I/O because tstuff and fstuff each print something in the
>> statement
>> y = merge(tstuff(),fstuff(),x(i))
>> but y itself is printed only in the next statement,
>> print *,y
>
> John, your original testcase in comment#0 was fine.
> I tried to extend it to check for constant as well as non-constant mask,
> and as you see I made a mistake by trying to make it smaller. Bad idea.
>
>> Or does evaluating merge allow each of tstuff and fstuff to be evaluated
>> at the same time? I was thinking of tstuff and fstuff being evaluated
>> in succession but could there be systems in which they are evaluated
>> simultaneously?
>
> I don't recall having seen a mentioning in the standard of the order of
> evaluation of different function (or subroutine) arguments. Do you?
>
> (Of course, if side-effects happen during that evaluation, such as I/O,
> unexpected things may happen.)
>
>> If so, whether the program is valid Fortran depends on the
>> kind of system on which it is being executed.
>
> Well, even if the print in tstuff/fstuff were a problem, one could construct
> other testcases with side-effects that might be conforming.
>
> --
> You are receiving this mail because:
> You reported the bug.
>
-- John Harper, School of Mathematics and Statistics
Victoria Univ. of Wellington, PO Box 600, Wellington 6140, New Zealand.
e-mail john.har...@vuw.ac.nz phone +64(0) 4 463 5276
