Bugzilla Bug 44201

Several man pages -- the only one I wrote down is bash -- are truncated at
a line ending with colon.  I.e.

  $ man bash
   ...
  COMMAND EXECUTION ENVIRONMENT
       The shell has an execution environment, which consists of  the  follow-
       ing:
  $ // ^%@#!!

The problem goes away if you change -Tlatin1 to -Tascii in /etc/man.conf.


Reproducible: Always
Steps to Reproduce:
1. see above
2.
3.

Actual Results:  
See above


Expected Results:  
"man bash" should produce 4512 lines ending with

GNU Bash-2.05b			 2002 July 15			       BASH(1)


Happens with two different x86 installations starting with
stage1-x86-20030910.tar.bz2.  One athlon-xp aggressive opts, one
pentium2 conservative opts.

I've had the same problem:  With the default `man.conf', some man pages are
truncated when piped through `col -b`.  My problem seems to be that `col -b`
truncates its output at any charcter with its high bit set, and the default
`man.conf' allows the output of such characters.  (I use `vim -R -' as my
manpager, and the input to vim has to be stripped of backspaces.)

With the default `man.conf', look at the output of `/usr/bin/man -P less bash`.
In the section `COMMAND EXECUTION ENVIRONMENT', there is a bulleted list.  The
bullets are displayed as `<B7>' in reverse video.  This is the character
`\0xB7', and less displays characters this way when the high bit is set.  If
you pipe the `man` output through `col -b`, the output is truncated at this
point.  Try `/usr/bin/man -P "col -b | less" bash`, you'll see the truncated
output.

Other characters and charater sequences found in the man files are also
translated into characters with the high bit set.  For example in
`/usr/share/man/man1/sendmail.1.gz' supplied by the `postfix' ebuild, in the
description of the option `-i', the word "don't" is found in the file as
"don\'t".  This two-byte sequence, "\'", is translated to `0xB4'.  `col -b`
truncates the manpage at this point.

As suggested by the comments in `man.conf', I removed the option `-Tlatin1'
from the definition of NROFF.  Now all is well.  Instead of getting `0xB4' for
the bullets in the bash manpage, I get the letter `o'.  The apostrophe appears
correctly in the sendmail manpage, and `col -b' doesn't truncate its output.

i too am seeing the <B7> junk ...

but when i remove the '-Tlatin1' from my man.conf, it still shows up :(

root@vapier 0 ~ # grep ^NROFF /etc/man.conf
NROFF           /usr/bin/nroff -c -mandoc

ok, man-1.5o_p2 fixes the <B7> crap ... turns out i couldnt get it to work on
my box because my man was reading the cached version :)

thanks guys for the heavy research !

All comments in here seem to prove that just removing the "-Tlatin1" solves the
original problem.
Why on earth then it was replaced with -Tascii as that breaks viewing of
man-pages on utf8-systems. For example it criples german umlaut (

All comments in here seem to prove that just removing the "-Tlatin1" solves the
original problem.
Why on earth then it was replaced with -Tascii as that breaks viewing of
man-pages on utf8-systems. For example it criples german umlaut (äöü and
ß).

I suggest removing -T... completely and let /usr/bin/nroff use its
autodetection.

col -p works w/out -Tascii ?

considering neither groff nor man supports UTF8 properly, this will be handled
properly in due time

both upstream projects are aware of the issues and working on them