bug with internal file I/O?

[Tobias Burnus]

Hi Magnus,

Magnus Hagdorn wrote:
> the attached program does not run as expected.
(You should add
   intvalues = 0
otherwise not all values are initialized.)

> I tried this code with the NAG compiler where it works as expected. Both gfortran-4.1 and gfortran-4.2 reach the
> end of the do loop.
>   
There is a difference between NAG f95 and sunf95 on one hand and g95,
gfortran, ifort on the other hand. NAG f95 and sunf95 give:
  6  (len)
  3  (i -1)
  2  5  69

gfortran, g95 and ifort have:
  6  (len)
  100  (i -1)
  2  5  59 0 0 0 ...

Now someone needs to study the standard for io-implied-do (cf. F2003,
"9.5.2 Data transfer input/output list").
It either might be that gfortran/g95/ifort have a bug or that it is
undefined by the standard. (Or that nagf95/sunf95 are wrong ;-)

Reading the standard (e.g. "For an io-implied-do, the loop
initialization and execution is the same as for a DO construct
(8.1.6.4)."), I would not be surprised if sunf95 and NAG f95 are correct
and that gfortran and g95/ifort have a bug.

Maybe one should ask at c.l.fortran and see what Richard Maine et al.
will answer.

Tobias