(new) non-ASCII filenames break unit tests on Linux

Stephan Bergmann stephan.bergmann at allotropia.de
Fri Dec 8 12:04:38 UTC 2023


On 12/8/23 11:30, Michael Stahl wrote:
> ... considering that LO uses UTF-16 strings for everything including 
> file paths, perhaps the best thing would be to add a check for the "C" 
> locale on startup, print an error and abort.

The situation for the ASCII "C" locale is not much different here from 
the situation for any other non-Unicode locale.  If a LO-internal file 
URL encodes a Unicode character that cannot be translated to the target 
character encoding, we have a problem.  This broken-by-design situation 
effectively resolves itself to a non-issue for Unicode locales (modulo 
remaining by-design issues in the opposite direction, around system 
pathname byte sequences that are not UTF-8), but I assume at least some 
non-Unicode CJK locales are still in somewhat popular use, e.g. in Japan.


More information about the LibreOffice mailing list