32-bit rar archiver with libc6-i386 cannot convert filenames to utf8

["Max Alekseyev" <maxale@gmail.com>]

rar archiver is distributed as 32-bit binary. Under debian-amd64 it is
possible to run it with libc6-i386 libraries. Unfortunately, there is
a problem described below.

Usually rar stores filenames in an archive "as is" (in the current
locale encoding) as well as in utf8 encoding. Filenames in utf8
encoding are particularly useful when the archive is supposed to be
processed under other OS (e.g., MS Windows) with different locale.

If for some reason rar cannot convert filenames into utf8, it creates
an archive only with filenames in the current locale encoding. And
that exactly what happens with rar using libc6-i386 libraries. Under
my debian-amd64 it creates archives only with filenames in koi8-r
encoding (as my locale is ru_RU.KOI8-R) that are unreadable under MS
Windows. What can possible prevent rar from converting filenames from
koi8-r to utf8 (locale ru_RU.UTF-8 is also in my system)?
Under (similarly configured) debian-i386 system rar creates filenames
in utf8 as expected.

Thanks,
Max