Un-initialised variables

[ukmo-ccbunney]

This issue will be used to log any un-initialised variables found in the code.
I will update it as/when I find any.

ww3_profsmd_pdlib.f90:

• RD10 and RD20. Used on line 6573,6574. Declared on line 6414.
  

  WW3/model/src/w3profsmd_pdlib.F90

  Lines 6573 to 6574 in 5cc3a01

  	RD1 = RD10 - DTG * REAL(ITER(IK)-IT)/REAL(ITER(IK))
  	RD2 = RD20

• fhndl: Used on line 3070 in a write statement. Unlikely to be cause any B4B reproduciblity errors, but could cause a crash if unit number is way out of range. I'm don't believe this subroutine is actually called (looks like debugging).
  

  WW3/model/src/w3profsmd_pdlib.F90

  Line 3073 in 5cc3a01

  WRITE(fhndl,*) 'IX=', IX, 'eSum=', eSum

ww3_ounf.F90

• FLGFC is unitilialised if not running with namelist input:
  

  WW3/model/src/ww3_ounf.F90

  Lines 245 to 246 in 5cc3a01

  	LOGICAL :: FLG2D(NOGRP,NGRPP), FLG1D(NOGRP), &
  	VECTOR, TOGETHER, FLGNML, FLGFC

  Needs initialising here with other variables not supported via .inp files:
  

  WW3/model/src/ww3_ounf.F90

  Lines 414 to 419 in 5cc3a01

  	! The following are only configurable via the namelist input
  	! and are hardcoded for .inp files:
  	TTYPE = "D"
  	TIMEUNIT = "D"
  	NOVAL = UNDEF
  	VECTOR = .TRUE.

[JessicaMeixner-NOAA]

@ukmo-ccbunney Thank you! Can I ask how you are finding these (and therefore how we can all help you find more?)

[aronroland]

@ukmo-ccbunney thanks, will make a pull request this weekend.

[ukmo-ccbunney]

@ukmo-ccbunney Thank you! Can I ask how you are finding these (and therefore how we can all help you find more?)

I'm compiling selected regtests with as many compiler warning flags enabled as possible.
The only problem is you get a lot of "unused variable" messages due to the nature of our code, so it's worth disabling that warning if possible. I am grabbing the compiler output and doing some judicious grepping to find any worrisome warnings.
The above were from the ww3_tp2.21 regtest compilation.

I am not 100% sure the compiler catches everything, so I am investigating whether I can run with a tool like valgrind to capture uninitialised errors at runtime... (I'm hoping it can do this!)

[JessicaMeixner-NOAA]

@MatthewMasarik-NOAA has run valgrind on WW3 for a few tests, maybe he has some insights on that.

[ukmo-ccbunney]

@MatthewMasarik-NOAA has run valgrind on WW3 for a few tests, maybe he has some insights on that.

I've run it before on serial jobs, but not on an MPI job (I assume that the PDLIB tests need to be MPI?)
Any pointers on running it under MPI would be appreciated!

[MatthewMasarik-NOAA]

@ukmo-ccbunney I have used valgrind on WW3, but also in serial.

[ukmo-ccbunney]

@ukmo-ccbunney I have used valgrind on WW3, but also in serial.

Out of interest, what options do you use when running valgrind?

Currently, I am doing:

valgrind --tool=memcheck --track-origins=yes --leak-check=yes <exex>

[MatthewMasarik-NOAA]

I could clarify that i have run valgrind on WW3 on MPI jobs, but I have not done any steps like using the LD_PRELOAD that they recommend for MPI. So this is not 'MPI-safe', but I think you can still get other useful information. I'd likewise be interested to hear if you end up able to run it using those MPI precautions.

[ukmo-ccbunney]

@ukmo-ccbunney thanks, will make a pull request this weekend.

Hi @aronroland ,
Re: ww3_profsmd_pdlib.F90
Were you still planning on doing this? I am happy to raise a PR for this one myself if that helps.
I just need you to confirm that the initial values of zero for RD10 and RD20 are ok.

[aronroland]

Hi Chris,

this is fine, I will merge it into init_bugs2, and I will also work out the comment on the triads not being initialized correctly. Let me also thank for these findings!

Cheers

Aron