Does core dump file really tell anything useful?

[thrasher@pmc_ucsc_edu]

Still trying to run down segmentation fault source. I find this in my core_lite files:

__diagnostics_NMOD_diag_physvar_ic : 758 # in file
physpkg@OL@1 : 139 # in file
# At location 0x090000000399e164 but procedure information unavailable.
# At location 0x090000000399a958 but procedure information unavailable.
physpkg : 135 # in file
stepon : 228 # in file
cam : 241 # in file


Here's what those lines are:
1. cwat_var => pbuf(ifld)%fld_ptr(1,1:pcols,1:pver,lchnk,itim)
2. call diag_physvar_ic ( c, pbuf )
3. do c=begchunk, endchunk
4. call physpkg(phys_state, gw, ztodt, phys_tend, pbuf)
5. call stepon ()


So it looks like it's having trouble writing or reading QCWAT somewhere? Or is this another useless message that has nothing to do with the actual problem?

 

[rneale]

I've seen this problem before. Which machine are you running on
which code base are you using?

 

[thrasher@pmc_ucsc_edu]

I'm running CAM 3.1.p2 on bluevista....

 

[rneale]

Yep, apparently this is a problem with the current compiler version on bluevista.
I think if you move to blueice, it should work.

 

[thrasher@pmc_ucsc_edu]

Yes! It actually worked! I can't believe how much time I've spent trying to get it running, and all I had to do was switch machines.

Thanks so much for your help, Richard. Seems I'll be able to keep all my hair for now. ;) Maybe this is something that can be noted on the CAM download page? The version 3.1.p2 specifically says it's supposed to be for running on bluevista...

 

[rneale]

Yep, it did run on bluevista until they recently upgraded the xlf compiler. Hopefully they will fix the problem before they do the same upgrade on blueice!