[Pw_forum] projwfc.x crashes when natomwfc iz zero

[Jure Varlec]

Dear all,

I have run into a situation where projwfc.x crashes with error

  On entry to ZHEEV parameter number  5 had an illegal value

This issue was reported about a year ago, but the poster never followed
through with more information. I have tracked the error down to the
basis::natomwfc variable, which for some reason is zero. Looking at
data-file.xml from the scf run, NUMBER_OF_ATOMIC_WFC is, in fact, zero.
This happens when the scf is done using sg15_oncv norm-conserving
pseudopotentials, but not when using GBRV ultrasoft pseudopotentials;
apart from ecut{wfc,rho}, everything is the same. How to proceed?

A related question. Not being very familiar with fortran, finding my
way around QE code is tedious. I could not figure out where
basis::natomwfc is actually set. I tried to use gdb watchpoints on the
__basis_MOD_natomwfc symbol, which is what gfortran produces. While the
read watchpoint got tripped many times as this variable is used quite a
lot, the write watchpoint was never tripped, implying that this
variable is not set at all, which I find very strange. However, the
watchpoint was tripped in qexml_module::qexml_read_bands_info when
data-file.xml was produced using GBRV pseudopotentials. To me, this
looks like NC pseudopotentials don't use natomwfc, but surely this
can't be, this variable is used all over the place. Are watchpoints on
global variables in fortran realiable at all?

My ignorance of the inner workings of QE now being obvious :) , I hope
the above information will be of some use for finding the problem.

Best regards,
Jure Varlec