GnuCash 2.2.0 under Windows (problems)

[Mike or Penny Novack]

As promised, I resumed testing today and have solved the mystery.

To repeat the description of the problem: (with a little more detail)

   In Windows XP GnuCash will not save preference changes. Or to be more 
precise, whether GnuCash will or will not properly go through the 
initial configuration process during which .gconf is created depends 
upon the NAME of Windows user account. I do not know ALL the "special 
characters" which are allowed in Windows user's names that will cause 
this problem to manifest. But for example:

If the name of the Windows user were "TestOnly" GnuCash will work properly.
If the name of the Windows user were "Test & Retest" it will not.
(if curious how I figured out to try this out see *)


This is obviously a bug -- there should be no dependency on something 
like user name! Until the problem is fixed, we can suggest work arounds 
such as the user creating a special user account in which affected user 
can run GnuCash (they may not be able to change their account name 
without a great deal of work**). It is likely that the percentage of 
users who would experience this problem is small so the bug is more of a 
nuisance in spite of the fact that something has to be seriously 
miscoded to make the processing dependent on the name. If a developer 
would get back to me with the names of the routines I should examine 
(those executed when GnuCash comes up the first time and creates the 
.gconf directory and its contents) I could take a look before filing the 
bug report. I did after all offer to analyze the code in cases where a 
bug was affecting me personally.

Michael

* I knew from earlier tests that:
a) whether with or without "administrator" rights made no difference.
b) GnuCash worked fine in a test account SET UP EXACTLY THE SAME as our 
regular account.
   At this point, looks impossible/crazy --- WHAT could be the 
difference? Here decades of experience come into play, because whenever 
you are faced with a situation where there appear to be no difference 
you need to stop, step back, and ask what are ALL the possible 
differences no matter how silly they might seem. Well our account name 
is "Mike & Penny" which contains both <space> and <ampersand> and the 
name of the test account was simple. Hmmmmm -- it's a difference, and 
yes, the name of the account would appear in the path name strings and a 
programmer might have made a string processing error that only mattered 
if there were "special characters" and something like this might well 
have escaped both alpha and beta testing since account names with 
special characters are not common.

 ** Example -- all their other application data may have path data. In 
our case that would be mainly T'Bird, F'Fox, and SeaMonkey data (the 
rest of our data, documents, etc. would be portable, not internally 
containing path information). The GnuCash configuration data contains 
path information (isn't portable) while the books themselves do not (and 
so are portable).