Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 28 Jul 2014 13:09:25 +0200
From:      Baptiste Daroussin <bapt@FreeBSD.org>
To:        Stefan Esser <se@freebsd.org>
Cc:        pkg@freebsd.org
Subject:   Re: [BUG?] pkg: Cannot get a read lock on a database, it is locked by another process
Message-ID:  <20140728110925.GC68875@ivaldir.etoilebsd.net>
In-Reply-To: <53D62D3E.2000004@freebsd.org>
References:  <53D511D6.7040108@freebsd.org> <53D619AB.5080101@freebsd.org> <20140728094609.GO50802@ivaldir.etoilebsd.net> <53D62D3E.2000004@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help

--MnLPg7ZWsaic7Fhd
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Mon, Jul 28, 2014 at 01:00:14PM +0200, Stefan Esser wrote:
> Am 28.07.2014 um 11:46 schrieb Baptiste Daroussin:
> > On Mon, Jul 28, 2014 at 11:36:43AM +0200, Stefan Esser wrote:
> >> Hi Baptiste,
> >>=20
> >> my mail to the ports maillist has not been answered, but I think=20
> >> what I describe is a real issue: Simultaneous "pkg" commands can=20
> >> lock-out each other in such a way, that one of the commands
> >> fails.
> >>=20
> >> This is not expected in existing scripts (e.g. portmaster),
> >> which assume a permanent failure, when in fact the cause was a
> >> temporary access restriction.
> >>=20
> >> I do not like the fact, that read operations (from the point of=20
> >> view of the user, e.g. "pkg info") now fails when other
> >> processes are also only reading the database.
> >>=20
> >> But I think that aborted updates to the package database (e.g.
> >> if a package has been installed, but cannot be registered due to
> >> other pkg commands having a read lock) are a bigger problem,
> >> since they may leave a system in an inconsistent state.
> >>=20
> >> Shall I open a PR on this topic or is it already on your to-do
> >> list?
> >>=20
> >> Regards, STefan
> >>=20
> > This is fixed with pkg 1.3 :)
>=20
> Hi Bapt,
>=20
> thank you for your reply, but I waited for 1.3.0 to be released
> (had tested pkg-devel and noticed this effect, but thought it
> might be a temporary problem) before posting to the ports@ list.
>=20
> I just retried with pkg-1.3.2:
>=20
> # pkg check -Bs
> ^Z
> [1]+  Stopped              pkg check -Bs
> # pkg info
> pkg: Cannot get a read lock on a database, it is locked by another process
>=20
> I never hit this problem with pkg-1.2.x ...

oh fun if I run pkg check -Bs as root, I can properly run pkg info as user =
but I
cannot do it as root, looks like a bug.
>=20
> But it makes it impossible to e.g. run "pkg check -B" (or pkg_libchk)
> in one window and issue "portmaster $PORT_WITH_MISSING_LIB" in
> another. Not that this is required functionality, but in the same
> way other scripted uses of pkg collide with each other ...
>=20
> > btw usually to hit more people on pkg adding pkg@ in CC is a good
> > idea.
>=20
> Thanks, I did not know that list existed. CC is added and I'll
> subscribe to the list ...
>=20
> Regards, STefan

regards,
Bapt

--MnLPg7ZWsaic7Fhd
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iEYEARECAAYFAlPWL2UACgkQ8kTtMUmk6Ey7cgCdEqHQ2kK8PSeZsR9Jc+eBGIAs
qBIAoKuDgWm+/sJx1bWhq5sHPKr+SpOV
=dDc6
-----END PGP SIGNATURE-----

--MnLPg7ZWsaic7Fhd--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20140728110925.GC68875>