Ken Moffat wrote:
> On support, Markku Pesonen pointed out that up to glibc-2.15 the
> tzdata was installed in both /usr/share/zoneinfo{,/posix} for data
> without leap seconds, and the data with leapseconds was installed
> into /usr/share/zoneinfo/right.
>
> He also noted that debian still do this.
>
> Taking a look at tzdata in debian, they create the zones with -L
> /dev/null, and then, except on embedded, repeat this for /posix,
> and then use -L leapseconds for /right.
>
> I assume that some apps use the /posix version and others use the
> 'regular' version. The amount of data in posix/ is only 1.9MB, I
> don't think it's worth attempting to be clever (e.g. symlinks from
> /posix/ ) and possibly again breaking things.
>
> What debian also do is run zic with -p America/New_York to
> create the posixrules file - again, looks as if we need that.
>
> Summary - all debian systems has posix values in zoneinfo/ plus a
> posixrules file. Most also have posix in zoneinfo/posix and values
> with leapseconds in zoneinfo/right.
>
> At the moment I'm debugging recent changes to my own buildscripts.
> Once I've managed to boot the new system, I'll go back to chroot,
> change the tzdata instructions and see what happens to the
> testsuites that were reporting errors.
I've still got some experimenting to do, but I determined that the
/etc/localtime file in the glibc instructions affects these tests. For
testing, I set
cp -v --remove-destination /usr/share/zoneinfo/GMT /etc/localtime
When I delete /etc/localtime, the time test in coreutils passes.
-- Bruce
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
