Declaration and include confusion

[h-faber]

Hi @all,

latest confusion I have concerns declaration and include files. Consider the following code:

PROGRAM ABC

REAL SD(16)

! several further declarations

include 'SOMEFTN.INC'

! REAL SD(16)

! Some file 10 gets opened

READ(10,' (1X,''SD(1) '',28X,F10.3)',END=20)SD(1)

END

The problem now is:

Error: Constants and expressions are invalid in read-only I/O lists. [SD]

Background: In the someftn.inc this variable SD is used and modified.

When I declare SD again after the include line (outcommented above), the compilation works. But I fear that all the manipulation done within the include file concerning SD gets lost. Am I right or wrong? And how can this be solved better?

(Intel Fortran Compiler 8.0 used)

TIA

Harald

PS: If I leave out the complete include line and the second declaration, I get

"Fatal: There has been an internal compiler error (C0000005). (code 1)"

Whatever that means...

Message Edited by h-faber on 04-04-2005 08:19 AM

[TimP]

If the compiler were accepting a declaration of SD() from the include file, a subsequent re-declaration should be flagged. So I would agree with your apparent suspicion that include processing is in trouble. Likewise, the error you quote appears to imply that no declaration was seen. Pre-processing was broken in the 8.0 Windows Fortrans. I suggest you try a current version.

[Steven_L_Intel1]

I don't see that preprocessing is in use here and I would want to see all the sources to reproduce the problem. My guess is that there is something important in the files not shown and that the compiler error is correct.

[durisinm]

Isn't the compiler complaining because the format string for the READ statement contains the string "SD(1)" as part of the I/O list? Shouldn't the I/O list of a READ statement contain only variables?

Mike D.

[Steven_L_Intel1]

An array element is a variable, at least in this context.

[h-faber]

Well, actually I still do not understand why the compiler speaks of constants or expressions. But the I/O can be because in the included file there is some file input/output operation. I changed it so that this is no longer an included file but a subroutine.

[Steven_L_Intel1]

I would still like to see a complete example that shows the original problem. Would you please send it (as a ZIP) to me at steve dot lionel at intel dot com? Thanks.

[h-faber]

Hi Steve,
it is customer's code so I have to ask for permission or refactor some variables and names.

[durisinm]

I'm talking about the first occurrence of SD(1) in the READ statement, not the second.

READ(10,' (1X,''SD(1) '',28X,F10.3)',END=20)SD(1)

Isn't it illegal to read a character constant, and isn't that what the error message at the beginning of this thread is saying?

Mike D.

Message Edited by durisinm on 04-06-2005 02:04 PM

[Steven_L_Intel1]

Oh, I see. Hmm. Yes, that's true - character constants in a format used for a READ are not allowed. I was looking elsewhere and missed that. Good catch.

[h-faber]

So this is a change? On VMS it worked that way so far, using Fortran 77 code/compiler. Anyway thanks for pointing this out!

[h-faber]

---

durisinm wrote:

I'm talking about the first occurrence of SD(1) in the READ statement, not the second.

READ(10,' (1X,''SD(1) '',28X,F10.3)',END=20)SD(1)

Isn't it illegal to read a character constant, and isn't that what the error message at the beginning of this thread is saying?

Mike D.

Message Edited by durisinm on 04-06-2005 02:04 PM

---

This sounds senseful of course - except that the code compiles when I use

CALL someftn()

instead of

INCLUDE someftn.inc

(After changing the included file to a subroutine inits first line)

The question remains:

Why does it compile without errors when I use CALL, but does not compile when I try to INCLUDE? The READ(...) shouldn't play a role I'd say. Am I wrong?