gsed fails to handle non-ASCII characters (bytes with top-bit set) in C locale

[vinc17@…]

For instance:

$ echo "abécd" | LC_ALL=C gsed -e 's/.*//'
écd

With the sed from Mac OS X and GNU sed under GNU/Linux, one gets a blank line, thus I suppose that this is what the user expects even though é isn't part of the US-ASCII character set specified by the C locale (and even though the result could depend on the encoding with some expressions).

The consequence is that building ocaml fails if gsed is installed with the with_default_names variant (see bug #20275).

[jabronson@…]

Cc Me!

[nox@…]

I don't get an empty line with Mac OS X sed:

Bellcross:~ nox$ which sed gsed
/usr/bin/sed
/opt/local/bin/gsed
Bellcross:~ nox$ echo "abécd" | LC_ALL=C sed -e 's/.*//'
écd
Bellcross:~ nox$ echo "abécd" | LC_ALL=C gsed -e 's/.*//'
écd

[vinc17@…]

That's strange. I have Mac OS X 10.4.11. If you have Leopard, perhaps Apple introduced a bug.

[vinc17@…]

BTW, does bug #20275 occur on your machine?

[nox@…]

I don't use +with_default_names.

[vinc17@…]

Yes, but even without +with_default_names (or without gsed installed), you should probably reproduce the bug because your Mac OS X sed is buggy too.

[rbubley]

I encounterd this problem too. For me (on Tiger)

machine:~/bin user$ which sed gsed
/usr/bin/sed
/opt/local/bin/gsed
machine:~/bin user$ echo "ab\303\251cd" | LC_ALL=C sed -e 's/.*//'

machine:~/bin user$  echo "ab\303\251cd" | LC_ALL=C gsed -e 's/.*//'
écd

[jmroot (Joshua Root)]

Has this been reported upstream?

[vinc17@…]

Replying to jmr@…:

Has this been reported upstream?

Yes: ​sed fails to handle bytes with top-bit set in C locale under Mac OS X

[jmroot (Joshua Root)]

-> nomaintainer

[jabronson@…]

Cc Me!

[Schamschula (Marius Schamschula)]

This ticket can be closed. With the current version of gsed, @4.2.2, I get the expected empty line (Tested on OS X 10.5.8 and 10.9.4).