Ulrich Drepper wrote:
Jungshik Shin wrote:
localedef didn't complain, but the collation result was all messed up
with the locale generated from it. So, I gave up using '..' to
represent the range and listed all characters in the patch.
That's no good solution. The locale description gets unmaintainable
in
this form. localedef should handle the case correctly. Provide a
test
case and we can look into fixing it.
Yes, I agree that we have to fix the root cause instead of working
around. Attached is a bzip2'd tar file with the following files:
1. ko_KR.v2 : the locale definition file with the range notation
2. ko_KR.test: 11,172 Hangul syllables and 8,999 Chinese characters
with Korean reading specified in Unihan-3.2.0.txt
are listed.
With the locale generated by the definition file I sent you earlier,
'sort sorted' reproduced 'sorted' file. However, with the locale
generated from ko_KR.v2 (I'm attaching here), 'sort sorted' generates
a totally different output.