repository breakage

[Ville Skyttä]

On Sat, 2005-11-26 at 09:56 -0500, seth vidal wrote:

> So if yum updates the other packages and completes its operation, but
> doesn't error out, how does the average user know that anything went
> wrong?
> 
> And if they don't know that something went wrong, then how will they
> know if they missed an update?
> 
> The reality is that most users see if the operation succeeded or failed.
> If it succeeded (no matter the messages it spewed out) they will assume
> all is well and move on. If we do that every time then a user could miss
> lots of valuable updates and never know.

So how many users in this target group are actually able or willing to
do something about it if the whole update fails, instead of just waiting
for a day or two or a week and maybe trying again if it succeeds then?
(I assume this is not how you intend things to happen.)

They'd need to first figure out what are the packages that are causing
their transaction to fail or from which repository are they from, then
issue the update again with proper --exclude and/or --disablerepo
switches.  Don't you think this is a major hurdle in this target group
of "most" or "extremely average" end users?

I would find it surprising if the actual effect in this target group
would really be a matter of "could miss lots of valuable updates" (and I
think the "lots" is an exaggeration in practice), but rather missing
_all_ updates until the issue with one or few at worst is fixed.

>  It's only the more advanced
> users such as yourself who looks for the nuanced warning messages.

Probably, but that depends on how the messages are presented.  Anyway,
IMO this is a far less dangerous thing and cosmetics when compared to
the one of automatically missing only the updates that wouldn't work
anyway at that moment versus missing those and the good ones too.

Orthgonal and partially OT: personally, I as a more advanced user would
be happier with the current behaviour of yum if the failed update run
would take something like 5 seconds from when I hit enter to the moment
I'm told that it didn't work out.  Sadly, that's nowhere near the case
today :/

> if the user has enabled a nightly cron update that most
> users never see the messages b/c they sit in /var/mail/root w/o ever
> being read on many, many workstations and laptops.

Wouldn't that be that place where the messages about missing all updates
would sit unnoticed too?