topic Re: Debugging inside a BLOCK construct in Intel® Fortran Compiler

Debugging inside a BLOCK construct

Andrew_Smith — Wed, 08 Oct 2025 12:39:16 GMT

With IFORT (Intel® Fortran Compiler Classic 2021.12.0 [Intel(R) 64]) the variables declared local to the block cannot be seen by the debugger.

So I switched to IFX (Intel® Fortran Compiler 2025.2.1 [Intel(R) 64]) and now they can be seen. But the local variables declared outside the block are now invisible!

Attached files replicate the problem with IFX. The use of an include file triggers the issue.

Re: Debugging inside a BLOCK construct

jimdempseyatthecove — Wed, 08 Oct 2025 13:45:17 GMT

You win some - you lose some.

As a hack

block associate(VariableOfInterest_=>VariableOfInterest) real :: blockLocalVariable ... end associate end block

Jim Dempsey

Re: Debugging inside a BLOCK construct

JohnNichols — Wed, 08 Oct 2025 18:28:05 GMT

Jim:

Why would you use a block?

It is like starting a fire with a match when someone has a torch.

JMN

Re: Debugging inside a BLOCK construct

andrew_4619 — Wed, 08 Oct 2025 19:21:45 GMT

You would use one because they can be quite useful and they exist!

Re: Debugging inside a BLOCK construct

jimdempseyatthecove — Thu, 09 Oct 2025 14:03:37 GMT

Think of a BLOCK as an inline contained subroutine complete with its own local variable declarations yet within the scope of the procedure containing the block.

Jim Dempsey

Re: Debugging inside a BLOCK construct

Andrew_Smith — Thu, 09 Oct 2025 17:40:09 GMT

The c language and most others allow variables to be declared when needed anywhere within your statements.

Fortran restricts us to declare all variables before any statements. I am not sure why Fortran has always had this restriction. Has there been proposals to change the standard in this area? I cant imagine it would affect performance but would greatly improve the coding experience.

A Fortran BLOCK allows you to create new variables almost anywhere. Its not as nice or as flexible as c.

Re: Debugging inside a BLOCK construct

JohnNichols — Mon, 13 Oct 2025 23:40:15 GMT

Accepting all that is said, I like a nice subroutine name such as

INPUT

LOAD

ANALYSIS

EIGENVALUES

OUTPUT