[Mageia-discuss] A possible risk ?
Maarten Vanraes
alien at rmail.be
Thu Feb 9 01:07:51 CET 2012
Op donderdag 09 februari 2012 00:39:50 schreef David W. Hodgins:
> On Wed, 08 Feb 2012 07:35:01 -0500, Michael Scherer <misc at zarb.org> wrote:
> > The risk is the lack of QA, and I have been repeating this since a long
>
> On Mandriva, I agree, which is why I haven't allowed users I support on
> Mandriva systems, to install updates themselves. Too many times I've
> seen unsigned packages, or file conflicts, etc.
>
> With Mageia, being on the qa team myself, these type of problems are
> being caught before the update gets moved from Testing to Updates.
>
> There is still the problem of mirrors only getting part of a large
> update, which can break a system, that needs to be fixed.
>
> Regards, Dave Hodgins
maybe possible ideas:
updates are being done by moving them. are we moving them and after move
regenerating hdlists? or using hardlinks? if the time is closer together and
hdlist is generated at a single time, i would guess it's less chance of
mirrors getting a bad time.
maybe the best way to handle this case is to detect incomplete hdlists
(somehow) and skipping this update and falling back to the previous one, and
still having the older uploads available.
I don't know if that means we would need stricter versioned requires for
certain packages?
on cauldron or updates_testing, it could be interesting to have an upload
queue that will hold the package in it, until all dependencies are resolved
(mostly packages which requires these versions), and then uploading them all
at the same time with only once regenerating hdlists.
More information about the Mageia-discuss
mailing list