Request for delaying stable release: Multi-currency code still not mature enough

Nigel Titley nigel@titley.com
29 Dec 2002 19:56:28 +0000


On Sun, 2002-12-29 at 18:46, Derek Atkins wrote:

> > I would agree with Christian, for what my opinion is worth. I *am* using
> > it for day to day real data, including multi-currency, and I have at
> > least three bugs open with bugzilla each of which is pretty much of a
> > show-stopper for me, two of them cause crashes, and one freezes Gnucash
> > solid.
> 
> I know at least one of these is a SX bug.  What are the other two?
> Besides, aren't you using some home-grown Linux-based machine that
> you built yourself?  Can you supply gdb traces and other debugging
> tools (or perhaps give a developer an account on your system) in
> order to aid debugging?

I'm not sure where you got the idea that I'm running on a home-grown
Linux machine. As a matter of fact, although I *do* build systems, the
main one I use to test gnucash is a commercially build model running an
out-of-the-box copy of Redhat 7.3, fully patched using Redhat's RHN
system. I'm not sure how more conventional you can get than that. I also
have a number of years experience in unix systems management. My home
systems are run professionally, as I would run a commercial setup. That
means no funny kernels, standard supported hardware, full daily backups
to tape, and all the rest. 

One of the bugs is indeed in the SX system. Another is a reproducible
crash after importing OFX data, for which you have full guile
tracebacks. One is a very peculiar one which may be kernel thread
related, associated with OFX imports, which freezes gnucash completely.
I've given you strace logs for this one, and also sent a copy of the
offending data to Benoit in response to a request from Christian, and
asked what further information I can supply.

I've supplied full guile/gdb tracebacks, strace logs and all the
information I can, even down to a complete copy of my financial records.
Josh, thank heavens, has finally managed to reproduce one of my bugs,
which he has marked urgent and major. I've repeatedly asked what
additional information I can supply, and where I've had requests I've
supplied it. I've spent many, many hours beating on the CVS code to try
and exercise it and shake out the bugs for you. I've also spent hours on
the en_GB translation. 

I know how difficult it is sometimes to accept a bug report from someone
you don't know, and don't trust, especially when you can't reproduce the
bug yourself. I have worked on support desks and I know how annoying the
lusers can be. 

> >  Now Josh is (hopefully) on top of one of these, but the other two
> > haven't even been succesfully reproduced by anyone but me (although I
> > can reproduce them without problem). Sure as fate, if you make a full
> > release, once more people are hammering on the code, you will find these
> > bugs are reproducible. I really don't think it is sufficiently stable to
> > release at the moment.
> 
> I admit that crashing is clearly a Bad Thing...  However not being
> able to reproduce a problem makes it very hard to bug-track.  Considering
> you can reproduce it, perhaps you could add a bunch of printf()'s in
> the code and help determine where it is dying, when, and how?

I will probably end up doing this if I have to, but I know from
experience that the author of the code is far better placed to do this
than an outsider and there will probably be a couple of weeks learning
curve before I can do anything sensible.

I may sound upset, this is *not* the case. I may sound as though I am
complaining, this is also *not* the case. I just want gnucash to be the
best financial package out there, and in its current state, at least on
my extremely vanilla, RH7.3 system it just isn't there.

If you want to release then go ahead. Its your privilege. I'll carry on
testing 'cos I believe in the product. But don't be surprised when the
great unwashed masses complain shortly afterwards. 

All the best

Nigel