Jim Gifford wrote:
I've been heavily researching this utf-8 build stuff for Cross-LFS.
Being the skeptical person I'm, the first thing that caught my eye was
the inclusion of Man-db and Berkeley DB into the book. Thinking this
was overkill I made a few inquiries.
1 - Fedora - Man-db is no longer being maintained is not needed.
Fedora uses a patched version of Man, and it doesn't work at all in
non-UTF-8 locales. This is unacceptable.
2 - Debian - Simple to use that's why we do
Also note that the Debian packager and the "upstream" author are the
same person (Colin Watson). So Man-DB is maintained, at least in Debian.
Only the savannah page is abandoned.
3 - Gentoo - Don't need it, don't waste your time.
Indeed, there is no man-db ebuild. They use Man, but ignore all the
problems. Also unacceptable.
4 - SuSE - uses Man-db.
Just an idea, couldn't we implement man, with something like the bash
script from
http://www.eglug.org/node/1759
Yes, this is possible. Just read the man-i18n.txt hint and implement the
manual page search algorithm and all the hacks :) Man-DB was used not
because of the DB, but because it knows these hacks.
But it is not a good idea to depend upon groff-utf8 (as in the example
you cited), since this would require conversion of all translated manual
pages to UTF-8 (impossible amount of job for BLFS). I would change my
opinion if Randy or Bruce agreed to audit all BLFS packages for manual
page installation, and add one "iconv" command for each non-English
manual page installed.
In other words, any implementation of manual browser is acceptable if it:
1) works in general
2) passes the following testsuite
3) doesn't introduce too many "ifs" in the book
4) doesn't put too much load on BLFS WRT recoding manual pages
5) doesn't make the procedure of reverting from a UTF-8 locale to the
old good non-UTF-8 locale more painful. I.e., the requirement to convert
all installed manual pages when reverting is unacceptable
The testsuite is below, it is assumed that all glibc locales are
installed. All images are located at
http://www.linuxfromscratch.org/~alexander/test-results/
1. Build LFS and Xorg-6.8.2 with
http://ftp.lfs-matrix.net/pub/blfs/conglomeration/Xorg/xorg-6.8.2-luit_race-2.patch.
This patch is a part of Xorg 6.9.0, but it looks like there's something
else wrong with that release.
2. LANG=ru_RU.KOI8-R xterm -lc (this is the simplest way of getting a
known-good terminal emulator), and do steps 3-5 in that xterm. Closing
xterm is not allowed between these steps.
3. yes --help, the result should be the same as on the
"yes--help-ru.png" screenshot. This step actually verifies that the
terminal is indeed set up properly.
4. man ump (assuming that the "ump" manual page doesn't exist). This
should result in either "man-ump-ru.png" or "man-db-ump-ru.png"
screenshot. English error message is also OK. "bad-ump-ru.png" is an
incorrect result.
5. man 5 passwd. This should result in "passwd-ru.png".
6. exit that xterm, start a new copy with: LANG=ru_RU.UTF-8 xterm -lc.
Redo steps 3-5, they should result in the same screenshots:
"yes--help-ru.png", "man-ump-ru.png", "man-db-ump-ru.png",
"passwd-ru.png". A sequence of empty squares is always an incorrect result.
7. exit the xterm, start a new one with: LANG=ja_JP.EUC-JP xterm -lc.
Redo steps 3-5, the correct results are "yes--help-ja.png",
"man-ump-ja.png", "man-db-ump-ja.png", "passwd-ja.png".
8. exit the xterm, start a new one with: LANG=ja_JP.UTF-8 xterm -lc.
Redo steps 3-5, the correct results are "yes--help-ja.png",
"man-ump-ja.png", "man-db-ump-ja.png", "passwd-ja.png".
Reconfiguration of Man is allowed (but still considered a bad thing)
only when xterm is closed.
--
Alexander E. Patrakov
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
