On Sun, Feb 13, 2005 at 08:38:58AM +0100, Goswin von Brederlow wrote:
> Jeroen van Wolffelaar <[EMAIL PROTECTED]> writes:
> 
> > On Sat, Feb 12, 2005 at 07:47:40PM +0100, Jeroen van Wolffelaar wrote:
> >> My mirror was running a bit out-of-diskspace, that I took a look at
> >> where this space went. I turns out, that debmirror doesn't cleanup
> >> properly.
> >> 
> >> For example, I have pool/main/k/kernel-patch-2.4.25-mips and kvim on my
> >> mirror, but that source package was removed from unstable in october
> >> resp. november 2004.
> >
> > Ah, I find the issue here, debmirror uses 'find', but due to space
> > constraints, part of my mirror were on other partitions, and symlinked
> > to there.
> >
> > find doesn't follow symlinks, so there it goes wrong.
> 
> Try adding "-follow" to the find call and run with --dry-run. I fear
> that adding -follow will remove /dists/unstable/* and /dists/testing/*
> though if you have those thinks, so --dry-run it.

You're correct -- that would have removed almost all of dists via
symlinks if I didn't have hadded --ignore=^dists, but otherwise it
works. I'm unsure what the best course of action would be here, find
doesn't have an option of 'follow symlink if pointing outside of
$ftproot' afaik. Otoh, this issue is mostly only relevant for pool, as
dists is not so easily splitteable outside of $ftproot via symlinks.

I can't fink of a clean solution to this problem that would work, unless
dists is handled special. I think it's best left up to you as
author/maintainer to decide whether or not you think this handling would
outweigh the hackiness, or whether you can find a non-hackish solution.
(doing this only for pool, and regular find for dists? Add the symlinks
in the top level of dists to a 'recursive ignore' for removal purposes?)

--Jeroen

-- 
Jeroen van Wolffelaar
[EMAIL PROTECTED] (also for Jabber & MSN; ICQ: 33944357)
http://Jeroen.A-Eskwadraat.nl


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to