Are magnetic host simulations properly parsed?

[JPchico]

@PhilippRue, @broeder-j I was testing bcc Fe to see the SCF when magnetism is considered (NSPIN=2, mag_init: True in the wf_parameters ). And the calculation is failing with the current error Finished [302] KKR parser retuned an error, when I do some checking I find that what is happening is that the parser is giving the error Error parsing output of KKR: orbital moment'.

When I see the output.2.txt I see that no orbital moment is printed, i.e. I'm running without NEWSOSOL, for what I can see in masci_tools it should not even reach this point to get the failure state.

I think that what is happening is that the use_newsosol function in matsci_tools is incorrectly identifying NEWSOSOL to be active (or use_chebychev) even when it is not. (I tried to parse them manually and I get this error) even when the line in question is

 <use_Chebychev_solver>=  F   use the Chebychev solver (former: 'NEWSOSOL')

I think that in the previous inputcard as the presence of the line was all that was necessary the use_newsosol just looks if a line containing NEWSOSOL is larger than 0, case which would not work here. One could perhaps replace it by something like this:

newsosol = True
if line.startswith("<use_Chebychev_solver>="):
     _temp = line.split()
    if _temp[1].lower() == "f":
        newsosol = False

P.S. How does one activate NEWSOSOL in the kkr_params?

[PhilippRue]

yes, I realized this as well and I have just pushed a fix in masci-tools which I had already in my working directory. Since I change the capitalization in some keywords in masci-tools I also had to update aiida-kkr (although this is not properly tested yet) in commit 37329e0.

[PhilippRue]

P.S. How does one activate NEWSOSOL in the kkr_params?

The 'traditional way' to activate this was to set RUNOPT=['NEWSOSOL'] in the kkrparams. Since the change to using the new style runoptions is not tested yet I advise using it like this.
At some point this will be done but there will be backwards compatibility to still support the RUNOPT-way of setting it.

[JPchico]

Thanks @PhilippRue, I guess that then I should for the time being use a local installation of masci_tools instead of the pypi version until a release is done.

[PhilippRue]

Thanks @PhilippRue, I guess that then I should for the time being use a local installation of masci_tools instead of the pypi version until a release is done.

yes that should work for now

[JPchico]

@PhilippRue I think the modification to masci_tools will still fail as

    if itmp>=0 and 'NEWSOSOL' in tmptxt[itmp].upper():
        newsosol = True

will result in True as the line that contains use_Checbychev contains NEWSOSOL

<use_Chebychev_solver>=  F   use the Chebychev solver (former: 'NEWSOSOL')

and the check later

    if itmp>=0 :
        if tmptxt[itmp].split()[1][:1].upper()=='T':
            newsosol = True    

Only checks for a True condition, I think that one would need to add the search for the false condition to be able to capture it, otherwise it will always return True. A simple fix would be amending the function with this

    if itmp>=0 :
        if tmptxt[itmp].split()[1][:1].upper()=='T':
            newsosol = True    
        if tmptxt[itmp].split()[1][:1].upper()=='F':
            newsosol = False    

[PhilippRue]

Oh yes, you are right, feel free to push a quick fix to the develop.

[JPchico]

Sure thing I'll make it as soon as I have a free moment