Re: advance=no problem

On 2008-05-12 10:22:39 -0300, rudra <bnrj.rudra@[EMAIL PROTECTED]
> said:

> dear friends,
>  i am using ifort compiler in linux.
> when i am trying to write something like
> write(*,'(1x,'a20',advance='no')"STARTING CALCULATION"
> .
> !do the calculation.
> .
> write(*,'a18')"calculation done'

Unbalance quotes so it is not what you actually did.

> i am expecting the program to write "STARTING CALCULATION" at the
> beginning of the calc; do the calc. and write "calculation done' in
> the same line when its over.
> but its not doing the same thing..rather its waiting to finish the
> calc. and writing the whole line in one go...anu idea?


When you use the advance="no" it makes sense to think of it as "to be 
continued".
Here you just took a long time to continue you thoughts about the output.

Remove the advance="no" if you want to see the output sooner.

You may then run into the problem that the system only wants to
deliver batchs of ouput and will not actually yield any output until
a full batch is available or the program finishs. The technical name is
buffering. The issue of when the buffers get flushed and whether a
partial buffer line (a prompt is the common case) will be flushed is
an awkward issue for the Fortran record oriented I/O model. Different
vendors have made differing detailed choices. Some expect you attempt
to work and others do not.

You would seem to have two choices. 1. Join the technical debate
on whether this was intended to work and whether it should. 2. Avoid
the fuss and just write full nonprompting lines with perhaps the aid
of politely asking the run time to flush the buffers is whatever way
they want you to ask. "call flush ( )" is common. RTFM. (Read The Fine
Manual - ifort is pretty good.)