Date: Fri, 25 Oct 2013 20:18:01 -0400 From: Glen Barber <gjb@FreeBSD.org> To: freebsd-stable@FreeBSD.org, freebsd-current@FreeBSD.org Subject: 10.0-BETA2, why it is late... Message-ID: <20131026001801.GB1740@glenbarber.us>
next in thread | raw e-mail | index | archive | help
--96YOpH+ONegL0A3E
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
To keep everyone informed on what is taking so long with 10.0-BETA2,
here is where we stand at the time of this writing:
- A problem was found after freebsd-update(8) builds were finished for
10.0-BETA1 which, because of a file within contrib/openpam containing
a tilde ('~'), would cause freebsd-update(8) to error when updating
a system. This is a problem that was discovered during the
9.2-RELEASE cycle, but worked around for that release cycle. A more
permanent fix has been committed to head/, and merged to stable/10.
- During testing of the fix for the openpam filename problem, it was
discovered that order in which freebsd-update(8) would install files
to the new directory '/usr/lib/private' collided with the creation of
the '/usr/lib/private' directory. This caused some shared libraries
('.so.N') files to attempt installation prior to the existence of the
directory, which would cause an error.
- Since between 9.x and 10.x, libc.so is a "regular" file (not
a symlink to a shared library), freebsd-update(8) thought
'/usr/lib/libc.so' was a shared library that should be
removed/replaced as part of the upgrade. This caused the removal of
'/usr/lib/libc.so', and subsequently, unfortunate side-effects.
- CTF was found to be leaking build-host information into the resulting
release build. Specifically, lines 130-133 define 'VERSION' as
'uname -srp', if otherwise unset. The negative effect of this is, in
the event the release build machine and the freebsd-update build
machine are out of sync (userland and kernel), the freebsd-update(8)
run to upgrade a system would unnecessarily update everything within
/boot/kernel/.
It has been pointed out that r257136 to stable/10 has an unfortunate
side effect of warning output from make(1) during the 'make delete-old'
and 'make delete-old-libs'. Unfortunately, the fix for the CTF
pollution is unclear at this point, but this will be correctly fixed.
For 10.0-BETA2, we will need to ignore the messy output for now. I do
not like it either, to be honest.
Please keep in mind: This is *not*, by any evaluation, fault of
freebsd-update(8). This is the natural progression of software, and
these issues are result of infrastructural changes on multiple levels.
Please bear with us; we'll have 10.0-BETA2 started soon.
Glen
--96YOpH+ONegL0A3E
Content-Type: application/pgp-signature
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (FreeBSD)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=OEK1
-----END PGP SIGNATURE-----
--96YOpH+ONegL0A3E--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20131026001801.GB1740>