I am having the same problem, although the --postcleanup option doesn't
appear to workaround it.  Luckily the --ignore-small-errors is working.

I mirror from a lower level source mirror, and I don't think I've _ever_
seen a mirror here (which is run once a week) complete without complaining
about at least one missing file.  Consequently, my Packages files were
several _months_ out of date, and though I appeared to have all the
"current" packages when I searched for them in the filesystem, I would
always get that the packages were up to date, even though I could see that
they weren't.

Seeing that it appeared to be a problem with the files under the dist/ tree,
I completely removed my dists tree in the hopes of forcing them to be
refreshed.  But after running debmirror again, it created the dists
directories, but left them completely empty.  And that was defintely NOT the
expected behavior, as it completely breaks, and intentionally leaves
broken, any computer trying to use that repository.

>From my point of view, several things need to be fixed:

1. If the dists files do not exist at all, the current version needs to be
placed there no matter the outcome of the rest of the mirror processing.  A
"possibly wrong" Packages file is better than none at all.

2. If the Packages files in dists/ are older than a certain timeframe (say a
few days), they should also be updated "no matter what".  The possibility of
an upstream mirror problem shouldn't affect whether I can even tell if there
are new packages available, and without updated Packages files, you can't.

3. There needs to be a warning printed out if the Packages files were not
updated so users can figure out what is going on.  A hint to add
--ignore-small-errors would also be nice.

Mirroring from the original source distribution no doubt normally completes
without errors.  But when trying to be nice and mirroring from a lower level
leaf mirror, errors from missing files are much more common, to the point of
being a normal occurance, and this case needs to be handled better.

Evan



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

Reply via email to