Peter Schwartz

Thanks for following up @ndkeen . Something to try is to remove `-fpe0` that way the bounds and pointer check can work better and provide the source of error (in...

So what I mean by it hindering debugging is that it raises this error but gives a pretty useless backtrace. Removing the fpe flag will give you the exact line...

@ndkeen Sorry I just realized I missed your direct question, I can make a branch this afternoon with the typo fixed to aid you in either confirming or refuting my...

@rljacob I worded that incorrectly: it should be "evaluation of a logical expression involving NaNs" So a floating point exception will get triggered on if statements like this: `if (carr(c)...

@rljacob It would require some additional changes to filter out by spval or `huge` in some loops, but that may be the best option. I did that for the initial...

@ndkeen The branch [peterdschwartz/lnd/cnpbudget-typo-fix](https://github.com/E3SM-Project/E3SM/tree/peterdschwartz/lnd/cnpbudget-typo-fix) fixes the typo that you can use. I guess I didn't explicitly mention it in the original post, but every test I listed there is one...

I don't know. They also overload the assignment operation for entire modules for data types with some special NaN assignment elemental function. `use shr_infnan_mod , only : isnan => shr_infnan_isnan,nan...

> I also tried on cori-knl with Intel `SMS_D.f09_g16.I1850ELMCN` and it fails as well in same way. However, I know that at some point, DEBUG cases were working with Intel...

Ha, hopefully we can get some closure on this 2.5 year issue. Also looks like your last test in November shows the memory corruption/out-of-bounds that I just fixed ``` At...

Backtrace looks very similar to those described Issue #4820 and that this is related.