forums.silverfrost.com

LitusSaxonicum · Posted: Sun Sep 30, 2012 6:04 pm Post subject:

Aha! I see it now. If only I hadn't been too lazy to retype the routine name, and hadn't just cut'n'pasted it from FTN95.CHM ...

Eddie

PaulLaidler · Posted: Sun Sep 30, 2012 7:53 pm Post subject:

/DCLVAR requires that you declare all subprograms when they are called.

I have corrected FTN95.chm.

DanRRight · Posted: Mon Oct 01, 2012 12:33 am Post subject:

Is declaring subroutines as external legal portable standard Fortran?

JohnCampbell · Joined: 16 Feb 2006 Posts: 2625 Location: Sydney

/dclvar is like an extension to /implicit_none.
I have tried to adopt this as a coding convention to document all variables and called routines, but without using /dclvar, I have not consistently updated my EXTERNAL declarations.
Adopting /dclvar could enforce this convention.
There is a question as to how using EXTERNAL as a documention of routines used conforms to the standard, but it can assist as a documentation of the code when returning to update.
Are warnings provided for both undeclared routines and declared routines that are not used ?
The use of EXTERNAL for documenting routines being used may conflict with the use of clearwin.ins, where the external declaration is in the include file. (not sure of this)

John

PaulLaidler · Posted: Mon Oct 01, 2012 10:21 am Post subject:

EXTERNAL is included in the Fortran standard.

Warnings are only provided (with /DCLVAR) for undeclared routines otherwise there would be a problem with INCLUDE statements for standard libraries.

LitusSaxonicum · Posted: Mon Oct 01, 2012 11:16 am Post subject:

Strictly speaking, EXTERNAL is required for FUNCTION names passed as a parameter to a subprogram, as otherwise a compiler could flag the name as a variable name.

For example

PaulLaidler · Posted: Mon Oct 01, 2012 2:06 pm Post subject:

Yes I fixed this just a few days ago.

DanRRight · Posted: Tue Oct 02, 2012 6:11 pm Post subject:

By some reason i do not like declarations of subroutines as external, or just lazy to add somewhere between 1000 and 10000 more lines, but may be i'm still not getting something. I understand that compiler might be confused with the functions (taking them as variables, so i declared that properly despite of it probably was done same 10000 times) but it seems to me that compiler can easily find if this is a subroutine or not because it is always CALLed.

JohnCampbell · Joined: 16 Feb 2006 Posts: 2625 Location: Sydney

Dan,

There are times, such as a function as an argument, where EXTERNAL must be used.
I also try to extend the implicit_none approach to use EXTERNAL to document, at the start of a routine, all routines that are being used.
When I do it properly, I find this to be useful to document what can be assumed of this routine for using other functions. It is useful when you come back to maintain or change the code.

This is not really related to the original post, as it has evpolved somewhat.

John

LitusSaxonicum · Posted: Wed Oct 03, 2012 11:46 am Post subject:

John and Dan,

What does (ab)using EXTERNAL give you that a comment line does not? Indeed, in comments, one can list:

Subroutines CALLed by this subprogram
Functions used by this subprogram
This subprogram called from ...

(Ab)using EXTERNAL can at best provide the first two, and even then, gives no pointer to which source code file (and line!) to find it in.

You don't need EXTERNAL at all if you go the extra distance and use MODULEs.

Eddie

DanRRight · Posted: Thu Oct 04, 2012 5:34 am Post subject:

Modules do not need external -- good. But I still do not see strong reason that subroutines MUST be declared as external for using /DCLVAR for all the rest cases. Compiler can exclude them from that itself.

Also, returning to original restriction on control length - isn't it just possible to increase the limit from current 1280x1280 to say 2560x2560 ?