wrong enconding in man pages [NOT solved]

[Crono81]

Man pages in my language (italian) have a bad encoding: words like "può" look like "puA2" and so on. This happens both on text consoles and on terminals (like konsole). How to fix that?

Some basic infos:

Code: Select all

Crono ~ # locale
LANG=it_IT@euro
LC_CTYPE="it_IT@euro"
LC_NUMERIC="it_IT@euro"
LC_TIME="it_IT@euro"
LC_COLLATE="it_IT@euro"
LC_MONETARY="it_IT@euro"
LC_MESSAGES="it_IT@euro"
LC_PAPER="it_IT@euro"
LC_NAME="it_IT@euro"
LC_ADDRESS="it_IT@euro"
LC_TELEPHONE="it_IT@euro"
LC_MEASUREMENT="it_IT@euro"
LC_IDENTIFICATION="it_IT@euro"
LC_ALL=

Crono ~ # locale -a
C
en_US
en_US.utf8
it_IT
it_IT@euro
POSIX

[selig]

My locale is cs_CZ.iso8859-2 and I am suffering the same problems with 3/4, A2 and similar stuff appearing in localised manpages. I am therefore using only the English ones - it has been like this for a very long time. Actually, when you open up the man page in a text editor, the econding is OK. I guess that the "man" program is simply having trouble with that.

[Crono81]

you may try to open /etc/man.conf and edit the line that starts with NROFF (can't remember it now, however there are some UTF8 comments above). Try to cut the -Tascii option and see if it's going to work for you.. not for me

[selig]

Hmm, I tried playing with the nroff/troff settings, but it does nothing.

[SirYes]

I use a system-wide pl_PL.UTF-8 encoding. However, most manual pages that have Polish versions are in the 8-bit ISO-8859-2 encoding. In the end I came up with this setup:

Code: Select all

NROFF           /usr/bin/nroff -Tlatin1 -c -mandoc -p
COL             /usr/bin/iconv -f iso-8859-2 -t utf-8

This effectively tells man not to clobber anything with NROFF, and convert 8-bit encoding to the UTF-8 one, using a non-standard COL program.

An ugly hack, but it works for me. And it is ugly, because it shouldn't be needed - but as of today man and groff don't support UTF-8 locales for their output very well. Additionally, this is also a system-wide setting (bad!) and completely misses the situation where there are many users on the system that use different locales (think: multiuser server).

If you're inclined to try this, be sure to specify the right input locale instead of "iso-8859-2" in the iconv invocation. Good luck!

[Crono81]

It's an ugly hack but using iconv actually works.

I use this for italian man pages:

Code: Select all

NROFF           /usr/bin/nroff -Tlatin1 -c -mandoc -p
COL             /usr/bin/iconv -f utf-8 -t iso-8859-1

[selig]

Thanks for the tips. I tried:
/etc/man.conf:

Code: Select all

NROFF           /usr/bin/nroff -Tlatin2 -c -mandoc -p
COL             /usr/bin/iconv -f utf-8 -t iso-8859-2

But it did not help. The manpages are obviously encoded in ISO-8859-2... but no matter what I tried with COL, it did not make the output correct.

[Crono81]

did you also try the opposite

Code: Select all

COL             /usr/bin/iconv -f iso-8859-2 -t utf-8 

(from 8859 to utf)?

[selig]

Yes, it did absolutely nothing - the display of characters with these option is the same as without them (both the NROFF and COL options). Some NROFF options produced unusable output, which did not even get displayed. COL options seem to have no effect whatsoever. My terminal encoding is ISO-8859-2, the manpages are in ISO-8859-2, almost everything in my system is ISO-8859-2. I really don't see what problems "man" is having.

I will now try

Code: Select all

TROFF           groff -Tps -man
NROFF           nroff --legacy NROFF_OLD_CHARSET -man
EQN             /usr/bin/geqn -Tps
NEQN            /usr/bin/geqn -Tutf8
TBL             /usr/bin/gtbl
# COL           /usr/bin/col
REFER           /usr/bin/grefer
PIC             /usr/bin/gpic
VGRIND
GRAP
PAGER           /usr/bin/less -iRs
CAT             /bin/cat

[Crono81]

well however that was not a solution. Some man pages have strange characters for indented lists, and the page becomes unreadable. That iconv should work only on pages that effectively need to be translated, not ALL...