bites assumed set mid UTF-8

bites assumed set mid UTF-8

[Dan Jacobson]

Bad bad bad. Emacs 21.4.1 shows the same Chinese character ("nuclear")
even though the second bit string is not valid UTF-8. Cc'd Handa.
11100110 10100000 10111000
11100110 00100000 10111000
M-x compile this makefile

B=perl -wle '$$_=unpack "B*", <>; s/.{8}/$$& /g; print'
Q=qp-decode
a:
	for i in =E6=A0=B8 =E6\ =B8; do \
	echo $$i|$Q; echo -n $$i|$Q|$B; done
#echo =E6=A0=B8|qp-decode #Chinese "nuclear"
#echo =E6 =B8  |qp-decode #Not legal Unicode

uxterm etc. correctly displays only the first.

Save what you see into a file and indeed they are both the first.
Reality hits when one first saves it into a file bypassing emacs
display, and then find-files it. Thereupon the second version is not
coerced into being the same as the first, and emacs guesses the file
is latin-1.

Re: bites assumed set mid UTF-8

[Kenichi Handa]

In article <87r75fignl.fsf@jidanni.org>, Dan Jacobson <jidanni@jidanni.org> writes:

> Bad bad bad. Emacs 21.4.1 shows the same Chinese character ("nuclear")
> even though the second bit string is not valid UTF-8. Cc'd Handa.
> 11100110 10100000 10111000
> 11100110 00100000 10111000

Thank you for the report.  It is already fixed in the latest
CVS code.  With it, the second byte sequence (invalid utf-8)
is decoded into "\346 \270" (i.e. 8-bit-char #xe6, SPC,
8-bit-char #xb8).

---
Kenichi Handa
handa@m17n.org