Re: Migrating file from the old gnumeric version to the current version

[Morten Welinder <mwelinder gmail com>]

For the record, I figured out what is going on:

* Old (1.0.x) gnumeric files are saved without an encoding.
* Non-ascii characters are embedded as "ê" but the meaning is
locale dependent.
* Gnumeric can read these old files by converting such characters from
locale to UTF-8
  during the load process.

This works great, until a long time passes and someone upgrades their
system and in the
process changes locale, for example to a UTF-8 locale.  See? 
Converting from locale is
no longer doing any good.

So the solution is to load those files while temporarily setting the
locale to whatever it was
when the files were created.  Once re-saved, the files should be fine.

CVS HEAD will now fall back and try ISO-8859-1.  That's a hack, of
course, but it will
cover a lot of people.

Morten