Re: Util/helpfiles and locale

["Jun T." <takimoto-j@kba.biglobe.ne.jp>]

On 2013/11/13, at 15:48, Bart Schaefer <schaefer@brasslantern.com> wrote:

> On my Ubuntu box it chooses en_AG.utf8 even though the default locale is
> set to en_US.utf8.  On my somewhat older CentOS host it chooses en_AU.utf8.
> 
> Does it really not matter which of these is used?

I tested (on Fedora 17) several locales for building help docs by, e.g.,

$ lc_ctype='-fja_JP.utf8' make runhelp

and found that the only difference is the character used by colcrt for
hyphenation (split a word at the end of a line):

  '‐' (Unicode: U+2010, UTF-8: E2 80 90) if any of UTF-8 locales is used,
  '-' (ASCII minus, 0x2D) if C locale is used (lc_ctype='-fC'), and
  a two-byte code 0xA1 0xBE if ja_JP.eucjp is used.

        ('‐' and '-' may look almost identical with some fonts)

Obviously EUC encoding must be avoided, but I feel using UTF-8 encoding may
not be the best way, if we must consider the possibility that the machine
on which a user views the docs is different from the host on which the docs
are built, and the user's OS (or terminal) doesn't support UTF-8.

If the zsh document sources (the .yo files) are pure ASCII (I guess they are,
and I guess yodl doesn't support UTF-8), then always using the C locale
when building the run-help docs would be the best and simplest way.


On 2013/11/10, at 17:57, Martin Vaeth <vaeth@mathematik.uni-wuerzburg.de> wrote:

> 2. It does not work out-of-the box: With current man/groff versions,
> one must export (an empty) GROFF_NO_SGR, MANWIDTH=80, possibly
> unset MANPL, MANROFFSEQ, set utf8-aware locales etc.

Are you sure utf8-aware locale is necessary? Did you try LC_CTYPE=C on your
Gentoo box?