From: Duncan <1i5t5.duncan@cox.net>
To: gentoo-amd64@lists.gentoo.org
Subject: [gentoo-amd64] Re: a different revdep-rebuild problem and solution
Date: Fri, 29 Feb 2008 12:34:25 +0000 (UTC) [thread overview]
Message-ID: <pan.2008.02.29.12.34.24@cox.net> (raw)
In-Reply-To: Pine.LNX.4.64.0802282342530.26122@thing.com
Steve Herber <herber@thing.com> posted
Pine.LNX.4.64.0802282342530.26122@thing.com, excerpted below, on Fri, 29
Feb 2008 00:05:46 -0800:
> I had a slightly different revdep-rebuild problem[.]
> I could not find the weird -MERGING-mythtv packages. I did remember
> having an emerge problem with MythTV when I first started to try it.
> [I finally found it in] /var/db/pkg/media-tv.
>
> In the directory media-tv were subdirectories with the -MERGING name. I
> removed all the directories in media-tv and reran revdep-rebuild and all
> was well.
As you likely figured out, /var/db/pkg/ is portage's merged package
database.
As a bit of background, portage tries to work as safely with the system
as possible, leaving as little time for a crash to affect system
stability as possible. Thus, it does all its compiling and a "fake
install" to a temp location (PORTAGE_TMPDIR in make.conf, /var/tmp/ by
default, with it adding the portage subdir to wherever the setting
points, so normally /var/tmp/portage/), before doing the final qmerge
step that merges everything from the fake location to the live filesystem.
Of course, as portage does that final qmerge it also has to add the
package information to the system database in /var/db/pkg/. Again, in
ordered to minimize damage risk as much as possible, and to make it easy
to correct the damage if something /does/ go wrong, portage initially
creates and works in a tempdir in its package database, naming it
MERGING- and the package name and version. It then does the copy from
the fake-install into the live filesystem as fast as possible, before
finally renaming the temporary package database dir, removing the
MERGING- prefix.
Thus, any MERGING-* entries in the database indicate partially merged
packages. Portage or the entire system crashed during the critical
qmerge phase, and the merge wasn't able to complete. It's wise to pay
particular attention to such entries, as not only do they take up space
in the database, but they indicate partially merged packages, some files
of which are likely stuck in limbo on the live filesystem, orphaned in
the incomplete qmerge step, not working and what's worse, very possibly
triggering other system issues. These other issues could range from the
revdep-rebuild issues you noted, to conflicting configurations (tho
that's not likely due to portage's config-protect features), to build
complications for other packages due to them picking up the wrong
versions of files, to security issues if someone can manage to run stale
binaries or link in stale libraries with known vulnerabilities, because
neither the package manager nor the sysadmin really realizes those
potentially vulnerable files are still there!
Thus, it's important to properly clean up such partially merged packages
when you find them. Fortunately, if you go back and try merging the same
package over again, portage will cleanup after itself. Since that's what
probably happens in the majority of cases, those MERGING entries are
normally relatively temporary and cause no real problems. However, if
the sysadmin decides to try a different version or simply gives up on
whatever he was trying to merge, THEN there's the potential complications
mentioned above, and a good Gentoo sysadmin will take the necessary steps
to clean things up manually.
Properly cleaning up is *NOT*, however, as simple as just deleting the
MERGING-* entries in the package database. Rather, if the package can't
be merged even if it'll be immediately unmerged (this is often simplest
when possible, as the cleanup is then automatic), one must manually check
to see if any of the files listed in the MERGING-* dir actually exist on
the main filesystem. If they do, one should then run a query (equery
belongs, or similar using your tool of choice) against the portage
database to ensure no other merged packages (as might be the case if a
different version was successfully merged) own the file in question, and
if not delete it (or if one is extra careful, move it elsewhere for
awhile to see if anything complains, just in case... as I said, sometimes
other packages will have detected and built against the wrong version if
it is still around due to a failed merge, they'll need rebuilt, but it
can be useful to be able to move the file back if necessary until you can
get the rebuild done).
So, you removed the MERGING-* temporary database entry, but did you
ensure any files it listed weren't still around causing trouble on your
live filesystem? That's supposed to be done first, so you know exactly
which files to check. However, if you missed doing that, you can usually
do the check anyway, tho at the cost of more hassle and risk of missing
something. In this case, again assuming the stale version is no longer
in portage or otherwise can't be merged and unmerged, thus automating the
process, the idea would be to check another version of the package,
seeing what files it has, and that you don't have other orphaned versions
of the same files still around. Again, be sure to equery belongs
anything that you find, before just removing it, just to be sure.
FWIW... the above is from experience... I had unstable hardware at one
point, and unfortunately ended up with a bit of experience cleaning up
such messes. I also ended up at one point with a portage database that
didn't match what I had actually installed, due to an emergency rescue of
a failing disk -- the backups I had of one part of the system didn't
match the backups of a different part, didn't match the parts of the
system I was able to successfully recover. =8^( Now THAT was a MESS to
clean up! (Since then, I've made it a point to keep all the areas
portage merges files to on the same partition and same backup as the
package database. Thus, if I lose my working system, when I recover, the
entire installed system and package database will be in sync, all from
the same date when I did the backup, so if I get lazy and let the backup
get old and stale, at least it'll all be the SAME staleness!)
Hope that helps! =8^)
--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman
--
gentoo-amd64@lists.gentoo.org mailing list
next prev parent reply other threads:[~2008-02-29 12:34 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-02-28 9:02 [gentoo-amd64] revdep-rebuild keep on detecting a broken link referred to libqt-mt.so.3 manuel
2008-02-28 19:02 ` [gentoo-amd64] " Duncan
2008-02-29 14:43 ` manuel
2008-02-28 20:15 ` [gentoo-amd64] " Beso
2008-02-29 8:05 ` [gentoo-amd64] a different revdep-rebuild problem and solution Steve Herber
2008-02-29 12:34 ` Duncan [this message]
2008-02-29 16:46 ` [gentoo-amd64] revdep-rebuild keep on detecting a broken link referred to libqt-mt.so.3 Chris Brennan
2008-02-29 17:19 ` [gentoo-amd64] " Duncan
2008-02-29 17:24 ` Chris Brennan
2008-03-01 1:52 ` Duncan
2008-03-01 4:11 ` Chris Brennan
2008-03-01 9:08 ` Duncan
2008-03-01 10:10 ` Beso
2008-02-29 16:42 ` [gentoo-amd64] " David Fellows
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=pan.2008.02.29.12.34.24@cox.net \
--to=1i5t5.duncan@cox.net \
--cc=gentoo-amd64@lists.gentoo.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox