CVF6.6C Update / Winteracter

[monmouth]

I am not sure if this is the place to discuss this anymore but I thought you maybe interested Steve? We have been using the CVF 6.6C update for a little while now without any problems. Today we discovered that Winteracter libraries no longer work, giving rise to the error "This module file was generated for a different platform or by an incompatible compiler or compiler release. It cannot be read." Reverting to version 6.6A or 6.6B the problem does not exist.This is not a problem for us as we will leave one machine at 6.6B for modifications to this program but thought that the forum may be interested.

Stuart

ps if you are out there "Bear of Little Brain" Happy New Year!!

[Steven_L_Intel1]

Get the new version of 6.6C I posted on the first.

[monmouth]

Thanks Steve. The only excuse I have is that I was a little worse for wear on the first!!

Stu

[Steven_L_Intel1]

Do let me know if this solves the problem for you.

[monmouth]

Steve. Everything is working well now. Thanks for your help.

Stu

[larsm]

Steve!

The new update works for me too, including the program sources that I submitted to vf-support@compaq.com.
Thanks!

But I noticed the following in the release notes:

3993 - Allow SAVE only within a subprogram.

Surely SAVE is allowed in both the main program (although with no effect) and within modules according to the standard?

LarsM

[Steven_L_Intel1]

Hmm - you're right, the standard does allow that. I wonder what we missed here. We'll investigate.

[Steven_L_Intel1]

The actual change is that the compiler issues a standards warning for use of the SAVE attribute in a main program, if you ask for standards warnings. The use of the SAVE statement doesn't get this warning.

Nonetheless, the warning is incorrect - as you point out, the standard says "The SAVE attribute may appear in a main program and has no effect." We'll fix this on the Intel side for the future. Thanks for noticing.

[larsm]

Sorry to be nagging, but just to make it clear: the same warning occurs for the SAVE attribute in modules as well and not only in the main program.

Best wishes

LarsM

[Steven_L_Intel1]

Right.

[Steven_L_Intel1]

I can't reproduce the problem for SAVE in modules. Do you have an example? SAVE in modules is legal and meaningful.

[larsm]

Steve!

In fact, it occurred in the sources that I submitted to you
for reference CVF21217 on December 3rd.

File CVFTEST.ZIP.

For example, in module COMBINE on this line:

INTEGER, SAVE :: ndim, sel_node

Best wishes

LarsM

[Steven_L_Intel1]

Sorry, Lars, I don't see that behavior with your volter.f90 source.

[larsm]

Steve!

I tried the sources exactly as I sent to you and it did occur.

BUT: it only occurred when I had "Fortran Standards Checking" on and set to "Fortran 95".
Does that account for our different behaviors?

Lars

[Steven_L_Intel1]

My error - was using the wrong compiler. I see it now - thanks.