On Sat, Mar 02, 2002 at 07:15:09PM +0000, Nick Ing-Simmons wrote:
There are some other warnings running compile without -Q
e.g. the attached.
Yes, that's a known issue. GNU liviconv's test/ lists CP950, EUC-TW,
ISO-IR-165, BIG5-HKSCS as four encodings containing many-to-one
(irreversible) mappings.
It seems that some of these encoding are not round-trip safe.
One reason for prefering .ucm is that by declaring one of multiple
map chars a fallback one can get the "right" thing for e.g. <U00F3>
is that 2B2E or 282E ?
Understood. But if '.enc' will consistently default to the larger one,
it seems good enough to me, as there is no well-defined "right" behaviour
as far as I know.
(in this particular case, as the 2B plane is the larger half-width pinyin
character, it's also arguably the right thing.)
Test case?
Against the table itself, I suppose? Ok, I'll write one.
/Autrijus/
pgp72rHtqjQXl.pgp
Description: PGP signature
