Hello, I’ve tried to find someone else using OpenBSD in various places for a while now, but with no success, so I’m hoping someone will read this.
I’m wondering what your output is from file(1) on a file you know has text encoded as UTF-8.
On my system (7.3-stable) the output is “Non-ISO extended-ASCII text”, and I’m trying to figure out if this is how it should be, or if I did something wrong setting up the system.
So, if you have a computer with OpenBSD and a minute to spare, could you try running file(1) on a UTF-8 file and see if it identifies it as UTF-8 or “Non-ISO extended-ASCII text”?
Thanks in advance
I don’t have OpenBSD, but to check if a file is UTF-8, try this:
file -i filename.txt
The command should tell you the charset information, and if it’s UTF-8, it should say something like “charset=utf-8.”
The file command might still label UTF-8 files as ASCII text due to its classification rules (UTF-8 is an extension of the ASCII character set).
That result doesn’t necessarily mean there’s something wrong with your system setup.
If I run file(1) on a file containing only characters in the ASCII set, the output is “ASCII text”. So far so good. If I add an “å”, the output of file(1) is “ISO-8859 text”. This is not correct, since if I look closer at what’s there, the “å” is encoded as \xc3\xa5, and this same file is reported to be UTF-8 in Debian and other OSs. If I add more unicode like “· ß ð ŋ” to the file, then file(1) says it is “Non-ISO extended-ASCII text” on OpenBSD. file -i testfile gives “text/plain”. Something is not right here.
edit: the file does not contain a BOM, but that is discouraged in UTF-8 files anyway. I have tried manually adding the correct BOM and it didn’t help.
Make sure your test file contains a decent amount of UTF-8 text, not just a few characters. The file command uses statistical analysis, so having more text might help it make a more accurate determination.
What does the locale command return?? … to set your locale you can use the export command (eg. export LC_CTYPE=“en_US.UTF-8” using whatever code is relevant)
I have all of this page: https://www.w3.org/2001/06/utf-8-test/UTF-8-demo.html as a test file. It renders fine and displays all the languages and special characters in vim.
LC_CTYPE is “en_US.UTF-8” , I export it in .xsession (and in .profile).
XTERM_LOCALE is also “en_US.UTF-8”
Yep I have the same result so most likely you didn’t do anything wrong. My VPS on openbsd.amsterdam shows this and my laptop does too.
Aha, I understand, thank you! file(1) might not be utf8 aware yet then.
I explored the source of file(1) and the part to determine file types of text file seems to be in text.c: https://cvsweb.openbsd.org/cgi-bin/cvsweb/~checkout~/src/usr.bin/file/text.c?rev=1.3&content-type=text/plain
And especially this part:
static int text_try_test(const void *base, size_t size, int (*f)(u_char)) { const u_char *data = base; size_t offset; for (offset = 0; offset < size; offset++) { if (!f(data[offset])) return (0); } return (1); } const char * text_get_type(const void *base, size_t size) { if (text_try_test(base, size, text_is_ascii)) return ("ASCII"); if (text_try_test(base, size, text_is_latin1)) return ("ISO-8859"); if (text_try_test(base, size, text_is_extended)) return ("Non-ISO extended-ASCII"); return (NULL); }So file(1) is not capable of saying if a file is UTF-8 right now. There is some other file (
/etc/magic) which can help to determine if a text file is UTF-7 or UTF-8-EBCDIC because those need a BOM but as you said UTF-8 does not need a BOM. So it looks like we are stuck here :)Which is ironic, given that OpenBSD only supports the UTF-8 encoding :)
Yes it looks like utf8 is a first-class citizen but really it is ASCII which is 100% supported. From the FAQ:
The OpenBSD base system fully supports the ASCII character set and encoding, and partially supports the UTF-8 encoding of the Unicode character set.
Thank you. At least I know now that it’s the expected output of utf-8 files, that’s good to know. Thank you again.
