GNUcash crashes on save - losing all changes

[Derek Atkins]

Bill Wohler <wohler at newt.com> writes:

> Let me ask a question: if you have an unclean shutdown, saved or unsaved
> changes notwithstanding, what does the orphaned lock file buy you?

It informs the user that there was an unclean shutdown and they may
have lost data -- or that they may need to replay the last logfile.

> Maybe I'm missing an additional reason for the lock file, but my
> understanding was that it prevented you from having two instances of
> gnucash on the same file. The lock file should thus go away if the
> gnucash process goes away, no matter how.

That's part of the reason, yes.

> Unless, as I mentioned before, the lock file is doing something more.
> 
> In answer to your other concerns about unclean shutdowns, there should
> already be a feature request for replaying the log the next time gnucash
> starts (after first querying the user) if it is more recent than the
> data file.

The log-replay code is already in CVS and will be in 1.8.5.  I'll note
that this doesn't help with changes to the business features -- only
transactional changes are written to the log.

There has not been a request to check the log timestamp and compare it
to the data file time stamp...

-derek

PS: IMHO this whole direction is a goose chase -- the endgame is
moving to an embedded SQL backend and most of this issue will go away.
So, frankly, I really don't want to spend precious developer-hours
working on major changes to what I consider "dead-end backend code".

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord at MIT.EDU                        PGP key available