Print format strings in FormatString.F90 error messages

[lizziel]

I am testing GCHP using MAPL3 with pFlogger v1.17.0 and several times have run into this error:

FormatString.F90         158
format_vector() - not enough position arguments for format string.

To find the location in MAPL where the logger call fails I had to add prints to pFlogger to print the string being parsed. It would be great if the error message included that format so that users could easily find and fix the issue. The same is probably true for the rest of the error messages in FormatString.F90.

[tclune]

@lizziel I agree with the desire for a better error message.

Can you also report the use case that triggers this? Mostly I'm curious, as I have not encountered it.

[lizziel]

It was getting tripped up on these lines for the case where phase_name (optional argument) is not passed to run_child_by_name: https://github.com/GEOS-ESM/MAPL/blob/002885e47562a26ae542ea7c67923c0f8f6dc5bc/generic3g/OuterMetaComponent/run_child_by_name.F90#L37-L39

I think this is a bug in MAPL. When I change phase_name to phase_idx in the prints I can get past the error and successfully get through the run phase for the model.

Note this is all using Intel 2024. When I use GNU 12.2 the model hangs and I do not get a FormatString.F90 error message.

[tclune]

The lines in your permalink look right to me. I will need to investigate further. Certainly 3 strings should not be maxing out the pflogger format vector.

[lizziel]

I think the issue is that the third string (phase_name) is not passed for all of the gridcomp runs: https://github.com/GEOS-ESM/MAPL/blob/002885e47562a26ae542ea7c67923c0f8f6dc5bc/gridcomps/cap3g/CapGridComp.F90#L86-L92

When it is not equal to anything it seems to be not included in the vector.

[tclune]

Ah. Thank you. Yeah - that makes sense. There are some tricky things that I can do with optional arguments there, but pflogger users should not be aware of the fact that under the hood there is a list of optional arguments arg1, arg2, .... Or I can put in a conditional.

Is this in your way at the moment, in which case I can have something by Monday morning.

[lizziel]

This is not a problem since I added the workaround for run_phase, and now have your version of the update too. Plus, now that I'm using Intel2024 I at least get an error message. :)