OK, I have found a simpler way to show my problem.

Try this on a UTF8 database:

select
case when 'Übergeek' collate unicode_ci_ai like 'ÜB%' collate unicode_ci_ai
then '=' else '<>' end as test_1,
case when 'Übergeek' collate unicode_ci_ai like 'üb%' collate unicode_ci_ai
then '=' else '<>' end as test_2,
case when 'Fußball' collate unicode_ci_ai like 'fu%' collate unicode_ci_ai
then '=' else '<>' end as test_3,
case when 'Fußball' collate unicode_ci_ai like 'fuß%' collate unicode_ci_ai
then '=' else '<>' end as test_4,
case when upper ('Fußball') like upper ('fuß%')
then '=' else '<>' end as test_5
from rdb$database

TEST_4 will show a mismatch, which is wrong.

Is this a bug? What can I do to get a case and accent insensitive
comparison?

Database: Firebird 2.5.2.26540 SuperServer 64-Bit
OS: Windows 7 Pro 64-Bit


Regards

Stefan