[lfs-dev] sed test failures, and locales was Re: tzdata
zarniwhoop at ntlworld.com
Sat Aug 18 17:53:26 PDT 2012
On Sat, Aug 18, 2012 at 04:57:03PM +0100, Ken Moffat wrote:
> On Sat, Aug 18, 2012 at 07:34:04AM +0200, g.esp at free.fr wrote:
> > I am with glibc-2.11 with chroot /etc/localtime as a copy of real /etc/localtime.
> > I don't see test failures in sed.
> > The 4 utf8 test report error 1 and XFAIL : utf8-x, so my understanding is that is intended they fail
> > All 65 tests behaved as expected (4 expected failures)
> I've had this since forever, the earliest log I can find is from
> 2010-06-16. But, it works ok in practice and I normally delete all
> directories after a package is built, so I've never had evidence for
> what is failing. Now that I'm having to break my scripts to examine
> other failures, I thought I'd try to look at all the failures.
> These utf-8 tests are expected to fail on a couple of OS's (hp and
> win), and I think they might get treated as XFAIL if you don't have
> an ru_RU.UTF-8 locale installed.
> ken at jtm1 ~ $grep -e '^FAIL' -e 'tests\ failed' \
> FAIL: utf8-1
> FAIL: utf8-2
> FAIL: utf8-3
> FAIL: utf8-4
> 4 of 65 tests failed
> Please report to bug-gnu-utils at gnu.org
Since I had been able to connect to fedora gitweb, I took a look
around [ possibly, it's easier since they stopped putting everything
on one page ]. sed is on about page 243, and they have a patch:
which changes various LANG= to LC_ALL= (specifically, it causes the
test to override LC_ALL instead of LANG). They've either been doing
that since 2009-10-14 (according to the patch) or since 2004 (fedora
bug 129014). Either way, it fixes it for me in the finished chroot:
root at jtm1 ~$ grep -e 'utf8' -e '=====' -A 2
All 65 tests passed
make: Leaving directory `/building/sed-4.2.1/testsuite'
Bruce - if your build was already ok on sed, do you have the
ru_RU.UTF-8 locale installed ?
Anybody else here install ru_RU.UTF-8 on recent glibc ? Does sed
report 4 tests failed (utf8-*) without the patch ?
I also think that ru_RU.UTF-8 locale should be added to
"the minimum set of locales necessary for the optimal coverage of
tests" in section 6.9 (depending on responses to those questions,
das eine Mal als Tragödie, das andere Mal als Farce
More information about the lfs-dev