Intel oneAPI 2025.3 appears to be broken with the latest Visual Studio 2022. 2025.2 as well.

[Nolan_H]

Visual Studio 2022 - 17.14.16, with the same experiences with 17.14.7

Intel oneAPI Base Toolkit - 2025.3

Intel oneAPI HPC Toolkit - 2025.3

Program optimizations are disabled.

List of problems:

1. Watch window cant see module variables through either <VariableName> or <ModuleName::VariableName>
2. Cant see local structures or variables in Locals or Watch windows.  I seem to be able to duplicate this for any array with more than 1 dimension.
   • Local single dimensional arrays/variables sometimes can be found in the watch window with all caps.
3. Conditional breakpoints set before running the program cause an error
4. Debugger as you are stepping through a program bounces around randomly.
   • Nested IF/DO statements bounce back to the top, or enter blocks that it doesn't execute
   • Stopping inside a routine with FORMAT at the end hits the FORMAT line before executing the rest of the routine.  IFORT doesnt seem to do this.

Side question, is there a way to get module variables to show in the watch window without the module syntax when the variable is not explicitly referenced in the current routine?  Like we would ideally like to use <VariableName> instead of <ModuleName::VariableName> in the watch window.

I haven't found a usable IFX + Visual Studio combination that works for debugging since repeatedly trying to move from IFORT in 2024.0.  We use a lot of legacy keywords like COMMON that should still be supported.  Any guidance would be greatly appreciated.

[e87tn95h]

It looks like the 2025.3 package lost an component of Fortran for Visual Studio - Intel® Fortran Expression Evaluator (FEE) extension. I think you could try to download and install this.

If you are using/having Intel oneAPI HPC Toolkit 2025.2 or earlier, this VS extension is included in these installation by default and persists until uninstall 2025.2 or earlier. You can see installed VS extensions at "Extensions" - "Manage Extensions..." on Visual Studio GUI menu Reference Link.

--

NOTE: I'm not Intel's employee nor moderator.

[TomArge]

Thank you! Installing FEE from VS Extensions solved the issue!

best regards

[Nolan_H]

Installing the Intel Fortran Expression Evaluator separately was a major step in the right direction.  It did solve items 1-3 for me.  I can now see all the variables available, which seems to be a major improvement over 2025.2.

I'd accept your solution, but I'd still like to get #4 addressed if possible.  The debugger as you step through enters blocks of code that it does not execute.  This is really confusing to debug/troubleshoot.  See attached video.

[Steve_Lionel]

I installed 2025.3 yesterday and FEE is there, so it doesn't affect everyone.

[jimdempseyatthecove]

I am surprised Ron Green hasn't posted a notice that 2025.3 is ready for download.

Lately I've been having some weird debugging issues with 2025.2.

Some of the source files (Debug build) stop being able to examine some variables, while being able to examine others.

Hopefully 2025.3 corrects for this.

Jim

[jonejames235]

Confirmed — seeing the same debugger issues with Intel oneAPI 2025.2 and 2025.3 on Visual Studio 2022 (v17.14.x). Module and local variables often don’t appear in the Watch/Locals window, and stepping through code behaves unpredictably, especially in nested IF/DO blocks. Debugging is effectively broken for IFX. Requesting confirmation from Intel if this is a known regression and if any compatible VS or compiler patch is planned.

[Nolan_H]

The stepping through the code unpredictably and not executing current steps are still present and make this compiler very difficult to debug still.

However, installing the Intel Fortran Expression Evaluator separately from @e87tn95h's comment took care of all the variable visibility problems in the locals, watch window, and highlighting variables.  That being said, I haven't seen a completely functional version of IFX (2024.0, 2025.0, 2025.1, 2025.2, 2025.3), and most of my coworkers have had to revert back to IFORT (2024.0).

[andrew_4619]

"The stepping through the code unpredictably and not executing current steps are still present and make this compiler very difficult to debug still."  I agree this is a problem but whilst is is "erratic" after some time it is mostly predicable and it doesn't prevent debugging IMO it does however reduce productivity. I am guessing the LLVM code is not structured like the IFORT  was an the debug support needs more work to achieve better alignment.

[Nolan_H]

If I dont use any IF statements, loops, and avoid nesting anything - it's mostly predictable for me.  That doesn't seem reasonable IMO.

It's the entering blocks of code that it doesn't execute that causes loads of confusion for me.  My experiences has been:

• This evaluated an IF statement as FALSE, why is it entering the IF block and walking through those lines?  Now I need to check every variable along the way to make sure there's not corruption
• Now it's starting to walk through the ELSE part of that same IF statement and is executing things
• Now its bouncing around randomly and appearing to not execute things.  Now I need to check every variable again.

Where before, it would just evaluate the IF statement as FALSE and execute the ELSE block where things get set as expected.

I'd like to see the base rudimentary debugging features in IFX get more TLC, as opposed to optimization bugs, multiprocessor issues, etc.  It does seem odd to get forced onto a compiler that at best doesn't work well after waiting a year or two for updates.

[andrew_4619]

"his evaluated an IF statement as FALSE, why is it entering the IF block and walking through those lines?  Now I need to check every variable along the way to make sure there's not corruption"

It moves the carrot to the first line in the block that isn't executed and then hops back to the if and then hops to the else case for example so two steps after the WTF moment you know that 'didn't happen'. You can still follow the logic it is more work.

And yes I agree IFX was not ready it to me looks like a senior management decision  along the lines "we are spending too much time and money, it WILL be ready by this date and we WILL retire IFORT." 

[TiborKibedi1]

I also experiencing the same problems in the integration of Intel Fortran 2025.3 and VS2022. As a desperate measure I try to reinstall Intel® Fortran Compiler & Intel® Fortran Compiler Classic for Windows* Version: 2024.0.2 using the link from the email received from Intel Registration center in late December 2023. It seems the web page of the Intel Registration center is not working. For many years I have been a registered user of Intel Fortran and I believe I should have access to older versions of the compiler. 

[jonejames235]

These are known IFX debugger issues in recent oneAPI + VS2022 builds. Multi-dimensional arrays, COMMON blocks, and module variables often fail to resolve in Watch due to symbol mapping bugs. Try enabling /debug:full and /Od, and disable Intel Optimization Reports. Also, attach VS via ‘Native Only’ debug mode instead of ‘Mixed’. For module visibility, IFX currently requires ModuleName::VariableName — no shorthand yet. Intel is tracking fixes for this in the next compiler update.