http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60596
kargl at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |kargl at gcc dot gnu.org
--- Comment #1 from kargl at gcc dot gnu.org ---
Looks like a bug in the program (and more importantly
a bug and/or processor-dependent behavior in ifort).
!! Unformatted I/O is faster in ifort under linux.
close(input_unit)
open(unit=input_unit,access='stream',form='unformatted')
You've closed the input_unit. The standard states:
The CLOSE statement is used to terminate the connection
of a specified unit to an external file.
You then open a new file and associate that new file with
input_unit. You'll find that you have a file named 'fort.5'
and it has, surprise, a size of 0.
program foo
use iso_fortran_env
integer fd,sz
real x
fd=10
open(unit=fd, file='tmp.dat', access='stream')
do i = 1, 100
call random_number(x)
write(fd) x
end do
close(fd)
open(unit=fd, file='tmp.dat', access='stream')
inquire(unit=fd,size=sz)
print *, sz
close(fd)
open(unit=fd, access='stream')
inquire(unit=fd,size=sz)
print *, sz
close(input_unit)
open(unit=input_unit, access='stream')
inquire(unit=input_unit,size=sz)
print *, sz
close(input_unit)
end program
troutmask:sgk[239] rm fort.*
troutmask:sgk[240] gfc4x -o z r.f90 && ./z
512
0
0
troutmask:sgk[241] ls -al fort.*
-rw-r--r-- 1 sgk sgk 0 Mar 19 16:14 fort.10
-rw-r--r-- 1 sgk sgk 0 Mar 19 16:14 fort.5
