Call for Testers: HBCI in GnuCash

[Christian Stimming]

Andreas Bogk wrote:

> Good work!


Thanks.


> It took me a few days to get everything up and running.  The gnucash
> autogen.sh is an unpleasant experience, but I could make it work.  I
> disliked that the default install prefix for libopenhbci is
> /usr/local/openhbci 


I absolutely agree that openhbci's default prefix used to be a problem, 
but it has been fixed on October 24 and now /usr/local is the default 
prefix. This change is also already incorporated in the 0.9.3 package 
since it was released on October 28. The pre-compiled rpm's always 
installed in standard locations, though. Anyway, this is fixed.


> Dependencies seem to be wrong, the HBCI code uses symbols from
> src/gnome, but HBCI is built before src/gnome, so linking fails. After
> manually invoking make in src/gnome, the application actually
> compiled.


Yes, this is a known problem, unfortunately. Proposals were made on how 
to fix it, but so far nobody has done it. Some code in gnome/ depends on 
stuff in import-export/qif-import/, so there is a circular dependency at 
the moment. The fix might not be too difficult but it's not trivial 
either, unfortunately.


> I'm currently using the commercial software Moneyplex for HBCI (at
> least it's a native Linux program!), and so I already have an
> established key with the bank.  Unfortunately, libopenhbci (and thus
> both gnucash and aqmoney) was unable to open my key file, with error:
> 
> MediumKeyfile::mountMedium: This seems not to be a medium file (bad type). (0) at MediumKeyFile::_reallyReadFile()


Right. The HBCI standard doesn't specify how keys should be stored in a 
file, so every application chooses its own file format. Since Moneyplex 
is closed source, we from openhbci cannot know its file format, and 
therefore we cannot use the keys generated in Moneyplex.


> Now that doesn't work for me. I keep getting: 
> 
> GnomeUI-WARNING **: gnome_app_insert_menus_custom: couldn't find insertion point for menus!


In very current CVS this problem occurres (introduced approx. two days 
ago by changes of jsled for gnc-split-reg widget et al) and IT IS NOT 
YET FIXED. It means that I cannot use HBCI functions, too, which I'm 
quite unhappy about, too. I suggest you should stick to the 1.7.2 
tarball for HBCI functions until I can confirm that this has been fixed. 
There haven't been any HBCI-related changes since the 1.7.2 tarball.


>>* Currently each HBCI action can only be executed while you are
>>online; support for off-line preparation and queueing is not yet
>>implemented.
> 
> Shouldn't that be easy to implement wby just using the account for
> queueing, and tagging the transaction (or split)?


Sure, it is *relatively* easy to implement. Nevertheless it does require 
some coding and GUI work, and as I said currently *I* have not the 
resources to do it but instead I focus on fixing what already has been 
done. If you want to give it a try, I will happily provide any 
assistance you need.

Thanks again for your feedback.

Christian