Hi! I have been testing this problem and finally got some results, which might be interesting/important to users from countries with non-Latin scripts: We have Koha 3.0. with Marc21 and nozebra. Entries made in Items fields in Georgian, Russian or even Latin characters with diacritics (in UTF8) were corrupted at some moment :-( which was leading to corrupted nozebra indexes and finally problems with searching. As a result many searches on Georgian terms were unsuccessful - which obviously is irritating - OPAC which cannot search records! I have tested this on KOHA demo sites... all the same! We have no experience with Zebra or KOha 3.2 yet, but I am sure the same bug would corrupt records similarly. So what we have found is: when we create biblios with Unicode, everything goes fine, when we add, modify, delete,... items with Unicode, again everything is fine, but all fields in Items are corrupted when we make changes in biblio! I.e. correcting spelling in Title or Author fields do not result in any problems with "biblios" but all fields like "callnumber", Item Note (public or private), etc which contain non-latin characters are corrupted! Then we examined MySQL fields and found following: All entries in biblios and items tables are fine, while in biblioitems table all fields except MARCXML are fine and all corrupted characters are located in MARCXML field, in parts corresponding to 952 MARC subfields!
From this what I guess is (as I cannot find any documentation on this issues):
1. problems come from addbiblio.pl or some functions called from this file. 2. as all fields except MARC 952 are fine, the problems come from that "function" which takes data from MySQL fields (obviously from Items and Biblioitems tables) and converts it to XML record to be placed in MARCXML field. To proceed with this, we would appreciate any help from people who created this module (cataloguing): 1. excellent if you could provide appropriate patch, 2. or if you could guide me where to look for a bug in the code (which files at least). As I say, I believe this is important to all Koha versions and users as even demo versions have this problem. Thanks in advance Irakli