remote z39.50 search returns no results for terms with diacritics

I did this search because I knew it would return results from the NOBLE catalog.

Note how the author's name appears in the results list. This suggests some kind of encoding problem as the data passes through Evergreen, yaz, and back.

OK, what I said about LOC above is not totally true.

The searches with diacritics still fail, but if I search by the English title, I get results, similar to that from NOBLE.

Not that in this screen shot, the characters of the author's name are not "mangled." They appear as they should in UTF-8.

Additionally, I have discoverd that certain UTF characters will cause a record not to display when retrieved. Having done some experimentation with yaz-client, it looks like these characters are getting encoded/decoded improperly at that level.

At this time, I believe the fix for these problems may be the same, i.e. tell yaz to use utf-8. However, if the solution proves to be different for these two problems, I'll create a new bug.

Came upon the same issue while investigating why yaz-client returned garbled characters for diacritics in records from Evergreen. In this case, the records all had e acute in their 65X fields.

Using Evergreen client on a separate system to retrieve the same records with diacritics via z39.50 from the 1st system returned no results. Other records returned as expected. SRU retrieved records but diacritics still mangled.

Tried to force yaz-client to treat as UTF-8 using 'marccharset UTF8', showed field values with diacritics as blanks. The MARC Edit Z39.50 client also returned improperly encoded/decoded diacritics, whether forcing UTF-8 or not. Tried MARC Edit Z39.50 to create file of records to import into originating system to see if Evergreen would translate the characters back to the proper diacritics, but it did not.

Confirmed that the record leaders do contain 'a' in position 09 to indicate Unicode.

Editing the configuration files for Simple2ZOOM to use UTF-8 instead of MARC-8 (see http://evergreen-ils.org/dokuwiki/doku.php?id=evergreen-admin:sru_and_z39.50 for the MARC-8 set up) and restarting Simple2ZOOM yielded same results.

I wonder whether the problem of retrieving record is not related rather to remote server than to Evergreen. We are using diacritics for retrieving record via Z39.50 from the Czech remote servers on daily basis and always get the existing record. I have attached snapshot of Slavoj Žižek search using The Union Catalog of the Czech Republic
My setting for the Union Catalog:
Source: SKC
Host: aleph.nkp.cz
Port: 9991
DB: SKC-UTF
Record format: FI
Transmission format: usmarc
Auth: False
Z39.50 Attributes for SKC are almost identical to OCLC, except for the truncation set to "1"

However the discussion reminds me partially of the problem of Z39.50 server querry encoding, seehttp://list.georgialibraries.org/pipermail/open-ils-general/2016-February/012781.html