7.2.0 fails gradients of bare ecps which <=7.0.2 works

[chemsh]

Dear NWChem Team,

we find a fundamental bug when testing NWChem 7.2.0 for ChemShell. Please see the attached example input and script used for compiling NWChem (both renamed as .txt). This is a test case of gradients on the background point charges which has been stable for many years. With NWChem 7.2.0 it fails with the message, either in serial or parallel:

0:cannot locate region: density matrix [0:-1 ,1:15 ]:Received an Error in Communication
Abort(-999) on node 0 (rank 0 in comm 496): application called MPI_Abort(comm=0x84000002, -999) - process 0

I've traced back to function hnd_elfcon_0 around line 262 in property/hnd_elfcon.F:
call ga_get(g_dens, bf_lo_i, bf_hi_i, bf_lo_j, where bf_lo_i and bf_hi_i became 0 and -1, respectively. I think it's a bug in the code because these lines are new in 7.2.0. @dmejiar

As metal oxide is one of the most common types of simulation in ChemShell, we appreciate if you could take a look at this issue. Thank you very much.

Kind regards,

You

compile_nwchem_on_ubuntu.txt
mgo_nwchem_bug.txt

[edoapra]

A fix now now available both in the master and hotfix/release-7-2-0 branch
5a45dbf

[chemsh]

Dear Edo,
@edoapra

thank you so much for the prompt response and hotfix! I can confirm hotfix/release-7-2-0 works for our tests again. I am closing the current Issue now.

Kind regards,

You