sft_crnkNic fails on Windows and macOS

[jdhughes-usgs]

The compiled version of mt3d-usgs fails on Windows and MacOS when running on GitHub actions using Intel compiled executables downloaded from https://github.com/MODFLOW-USGS/executables. The same happens locally on MacOS. The error from my local run is:

 TIME STEP NO.    1
 FROM TIME =   41400.     TO    86400.    

 Transport Step:    1   Step Size:   180.0     Total Elapsed Time:   41580.    
forrtl: error (140): floating inexact
Image              PC                Routine            Line        Source             
mt3dusgs           0000000107DF490A  Unknown               Unknown  Unknown
libsystem_platfor  00007FFF6CDE9F5A  Unknown               Unknown  Unknown
mt3dusgs           0000000107D7B078  _sft1fm_                  523  sft1.f
mt3dusgs           0000000107D8369F  _gnt1fm_                   23  gnt1.f
mt3dusgs           0000000107D9716A  _MAIN__                   376  mt3d-usgs.f
mt3dusgs           0000000107BF36AE  Unknown               Unknown  Unknown

Currently overcoming this CI issue on GitHub actions by excluding this test (t012_test.py -- test_mfnwt_CrnkNic()). Travis does not fail because we are only running tests on Linux.

It is possible that modified compiler switches would resolve this but the error suggests there is something smoldering.

[briochh]

Hi @jdhughes-usgs, I am seeing a similar issue (although maybe not identical). Running a mt3d-usgs simulation using .exe from https://github.com/MODFLOW-USGS/executables on Windows. In fact, I invariably get "floating invalid" type errors for all externally compiled version of mt3d (even back to the version 1.0 compile on https://github.com/MODFLOW-USGS/executables). e.g.:

 TIME STEP NO.    1
 FROM TIME =   2953.0     TO    2981.0

 Transport Step:    1   Step Size:   28.00     Total Elapsed Time:   2981.0
forrtl: error (65): floating invalid
Image              PC                Routine            Line        Source
mt3dusgs.exe       00007FF771A65D9A  SFT1FM                    527  sft1.f
mt3dusgs.exe       00007FF771BC9874  GNT1FM                     23  gnt1.f
mt3dusgs.exe       00007FF771A8173C  MAIN__                    376  mt3d-usgs.f
mt3dusgs.exe       00007FF771CA933E  Unknown               Unknown  Unknown
mt3dusgs.exe       00007FF771CA9724  Unknown               Unknown  Unknown
KERNEL32.DLL       00007FFD93BF84D4  Unknown               Unknown  Unknown
ntdll.dll          00007FFD95DCE871  Unknown               Unknown  Unknown

Interestingly (maybe) the point in the simulation where it falls over is not consistent for any given executable version, with repeated executions failing at different points. <- does this point towards compilation a issue? Note, my issues seem to relate more generally to the SFT, floating errors occur regardless of the value of WIMP (although the inconsistency of the failure point makes it hard to judge if this has any impact.)

Feel free to ignore this, if it is unhelpful. Always a strong chance of user error!

[jdhughes-usgs]

@briochh I think @emorway-usgs is going to try to address this in the next MT3DUSGS release.