[DFTB-Plus-User] Bug in src/Makefile.common; Feature Request: Pass Arbitrary Input File

EISUKE KAWASHIMA kawashima+dftbplus at g.ecc.u-tokyo.ac.jp
Wed Dec 28 00:49:53 CET 2016


Dear Dr. Bálint Aradi,

Thank you for your reply.

>> This just copies binary to the $(INSTALLDIR), which causes error when
>> the directory does not exist.
>> make install should create $(INSTALLDIR) before the copy, like
>>          install -d -m 755 $(INSTALLDIR)
>
> You are right, that would be more comfortable, indeed. However,
> 'install' is not necessarily a tool, which is available on all Unix
> (non-Linux) workstations. So probably we would have to come up probably
> with some combinations of 'mkdir' and 'cp' instead.

Yes, mkdir will do.

>> 2. Feature Request: Pass Arbitrary Input File
>> Currently dftb+ seems to accept no arguments and executes only
>> dftb_in.hsd or dftb_in.xml on the current working directory;
>> dftb+ raises error if the directory has both of hsd and xml.
>> It is convenient if dftb+ accepts an input on other directory or one
>> with different name,
>> like dftb+ /path/to/dftb_in2.hsd.
>> I appreciate if you implement it.
>
> This is something I was also considering. However, I had two objections
> so far:
>
> * Handling command line arguments was not standardized in Fortran 95, so
> one would have to come up with a compiler dependent solution. This issue
> is fortunately solved by now, as Fortran 2003 defines the
> get_command_argument() function.
>
> * Is that really helpful for the reproducibility of your scientific
> results, if you can specify an arbitrary input file? As the output files
> are not prefixed by DFTB+, how do you know, when looking up your results
> a few months/years later, which input file was used to create the output
> files. As long as the input file name is fixed and must be in the
> current folder, you will have to use separate folders for each
> calcultion, so things are more or less clear. With the possibility of
> specifying an arbitrary input file, things get little bit harder to
> follow, don't they?
>
> But, I am not too dogmatic on this, so if there is a wide interest on
> that feature, we can easily implement it.

DFTB+ prints "Interpreting input file 'dftb_in.hsd'" and creates
dftb_pin.hsd or dftb_pin.xml on the working directory, so I believe we
can find the input file as long as we manage output files
properly—different directories for different inputs.
Thank you for your consideration.

Best Regards,

--
Eisuke Kawashima


More information about the DFTB-Plus-User mailing list