[Koha] Another 229 to 30 migration, again - items... a new twist?

[Andrew Séguin]

I'm trying to do the migration from Koha 229 to Koha 30, and haven't
yet experienced full success.

First, upgrading in place didn't work out so well for some reason - I
had copied the virtual machine's disk, so was able to revert quickly.
After looking through the mailing lis, I then tried just running a
Koha 30 system; imported the database of the 229 system and ran the
upgrade scripts... this also didn't work out so well, "lazy me" just
moved on once again. Finally, I took a koha 30 prebuilt virtual
machine (time constraints, surely not because I'm a "lazy" system
admin, right?) and imported the bibliographic entries.

Through the web interface, it worked so-so, rejecting some "slices"
(read: independent files with 1000 exported records from the 229
system) of the exported data... tried the bulkimport.pl script which
rejected certain records as bad marc ... which I took (combined with
some combing of the mail-group archives) as the reason behind the
failure of the web interface. In the end, imported by the web
interface or bulkimport.pl, I would end up with some biblios with the
items all proper, some with the items not showing up properly (without
call number).

The Koha 30 system I'm using is the pre-built virtualbox machine 3.0
release -- version "Koha 3.0001005". Before running bulkimport, I
recreated the same item-types and branches. Both systems are set to
MARC21.

I'm turning to the community here for some advice how to proceed /
what to consider (I'm not afraid of research, scripting, or SQL... but
a bit lost at the moment):

* My first question would be, is there any trick, quick way, of
knowing which biblios were rejected by bulkimport.pl ? I tried with
"-v", "-v 2" without success to get more details.
   (sample/full command: nohup
/home/koha/kohaclone/misc/migration_tools/bulkmarcimport.pl -d -v 2
-file ~/Koha20090811_02.mrc -commit 1 | tee
~/bulkimport_KOP20090811_2.log
    after having set PERL5LIB and KOHA_CONF)

* My second question, is - are there any pointers/ideas concerning the
import of items?

Items that I consider properly imported have the call number
displayed; where not, the call number is missing. In either case at
least the barcode is correctly imported. Maybe somebody has a quick
moment to look at the difference between the two and sees something
obvious I don't?
* Imported correctly:
http://161.53.212.104/cgi-bin/koha/opac-detail.pl?biblionumber=820
* Not imported correctly:
http://161.53.212.104/cgi-bin/koha/opac-detail.pl?biblionumber=9880
(was biblio 11874)

I just (while writing this mail) found a difference looking at the
records from Koha 229 and Koha 30: bibid 820 had under Koha 229 the
"call number prefix" set to the same value as "call number" where 9880
doesn't. Maybe somebody has a comment about this (did I ignore
something basic with the frameworks)? - I'll test the call number /
call number prefix issue (either with marc-edit or directly in 229's
database, for each item, copying the call-number to the
call-number-prefix subfield value ...) within a day or two...

Many thanks for all those in general who take the time work on Koha,
even more to those who patiently read this entire email!

Andrew