Hello,
LFS-related patches that are used in the utf8 branch of the Live CD are
available at:
http://www.linxfromscratch.org/~alexander/patches/
Please review them for inclusion into the book (preferrably utf8-only
patches should not go into the trunk, but into a separate utf8 branch).
Details are below (what LSB testsuite tests is in fact mandated by POSIX).
1) [NON-UTF] coreutils-5.2.1-dateseconds-2.patch: already discussed,
accepted upstream but rejected for LFS as "the issue is too minor to
fix". The issue is that date --date "some-date-specification" without
the patch will use the number of seconds from the current date, not
zero, if some-date-specification does not have seconds.
2) coreutils-5.2.1-i18n_fixes-1.patch: this is a combination of several
RedHar patches. coreutils-i18n.patch fixes LSB testsuite failures such
as "When -t option is not specified, verify blank characters according
to the current locale are used as input field separators.".
3) diffutils-2.8.1-i18n-1.patch: fixes the LSB testsuite failure that
says: "* When -b option is specified, verify this utility interprets
whitespace character according to the current locale."
4) grep-2.5.1a-redhat_fixes-1.patch: this is a combination of many
RedHat patches. Testcases for some of them are at the top of the patch.
The -i18n RedHat patch is required for LSB testsuite to pass (one of
broken cases is "Can't handle a pattern \"ＡｂｃＤ\" with -i.").
coreutils-jday.patch and coreutils-zh_CN.patch make "ls" print Japanese
and Chinese dates correctly. coreutils-overflow.patch fixes the fact
that translations of "IDLE" and other strings from "who -Hu" header
overflow the buffer (even in some non-UTF-8 locales). This may be
relevant in gcc4 branch due to libmudflap.
5) [NON-UTF] psmisc-21.6-acsc-1b.patch: Testcase: boot with the non_UTF
CD, select the ru_RU.KOI8-R locale and the ru-ms keymap. Activate the
virtual console. Look at nicely formatted date in the prompt. Run
pstree. Look at garbage in the prompt where Russian characters are
supposed to be.
6) [NON-UTF] readline-5.0-redhat_fixes-1.patch: this fixes the crash
that occurs in "read -e foo" when $PS1 is longer than the screen width.
The bug doesn't exist in the pristine Readline, it's introduced by
readline-5.0-fixes-1.patch
7) texinfo-4.8-multibyte-1.patch: testcase is on the attached
screenshots. screenshots 1 and 3 were taken in the ru_RU.KOI8-R locale,
screenshots 2 and 4 are from ru_RU.UTF-8. You can see that lines are
wrapped incorrectly on screenshot 2, and info even broke one multibyte
character (that resulted in the empty swuare after "to determine if an
XML document is well"). Also the status line and the prompt are
truncated prematurely. See also the wrong text position at the search
prompt on screenshot 4. The patch replaces all problematic screenshots
with their English equivalents (which is, of course, not a real fix).
Also the following patch already in the archive is applied:
http://www.linuxfromscratch.org/patches/downloads/sysklogd/sysklogd-1.4.1-8bit-1.patch
Gnome likes to log messages in the current locale. Without this patch,
they become garbage in UTF-8 locales.
Non-patch changes (such as replacing GNU groff 1.19.1 with Debian Groff
1.18.1.1-8 and replacing man-1.6 with man-db-2.3.4) and BLFS patches
will be discussed separately.
--
Alexander E. Patrakov
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
