ldb: LANG=tr_TR.UTF-8 make test - fails

[Alexander Bokovoy]

On ti, 17 joulu 2019, Alexander Bokovoy via samba-technical wrote:
> On ti, 17 joulu 2019, Isaac Boukris via samba-technical wrote:
> > Hi,
> > 
> > I'm looking into this bug, which I can reproduce on master:
> > https://bugzilla.redhat.com/show_bug.cgi?id=1743531
> > 
> > In short it looks like running the ldb tests with tr_TR.UTF-8 locale causes
> > this failure:
> > 
> > Running Python test with /usr/bin/python3: tests/python/index.py
> > ...F...FF...F......F...FF...F....
> > ======================================================================
> > FAIL: test_delete_index_multi_valued_truncated_keys
> > (__main__.MaxIndexKeyLengthTests)
> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >   File "tests/python/index.py", line 999, in
> > test_delete_index_multi_valued_truncated_keys
> >     b"0123456789abcde1" + b"0123456789abcde1")
> >   File "tests/python/index.py", line 98, in checkGuids
> >     self.assertEqual(len(res), 1)
> > AssertionError: 0 != 1
> > 
> > Does anyone have a clue how to look further or how the locale impacts
> > here ?
> 
> Most likely it is GUID versus Guid in Turkish locale:
> 
> $ for i in tr_TR en_US ru_RU ; do LC_CTYPE=$i.UTF-8 ~/tmp/strncasecmp Guid GUID ; done
> tr_TR.UTF-8: Guid ? GUID = 32
> en_US.UTF-8: Guid ? GUID = 0
> ru_RU.UTF-8: Guid ? GUID = 0

Forgot to add: https://en.wikipedia.org/wiki/Dotted_and_dotless_I

There are at least two locales (tr_TR and az_AZ) in glibc have dotless i
transformation different from Latin scripts and GUID versus Guid
comparison would be different there (attribute name would not match in
the test).


-- 
/ Alexander Bokovoy