FESCo wants to ban direct stable pushes in Bodhi (urgent call for feedback)

[Josh Stone]

On 03/01/2010 11:46 AM, Seth Vidal wrote:
>>> One thing to consider: while from a psychological standpoint, a regression
>>> is indeed perceived as much worse than an unfixed bug, from a technical /
>>> practical standpoint it's actually the smaller issue: you can rollback to
>>> the version of the package before the regression, you can't rollback an
>>> unfixed bug as there's nothing to roll back to!
>>
>> Given that we don't provide an easily accessible user-friendly rollback
>> mechanism, I don't know that that's actually applicable to the general case,
>> though.
> 
> yum history undo works pretty well. Not flawless, to be sure - but it's 
> not bad for the simple-ish cases.

I think the yum-history-undo is great when I just want to try out a new
package -- install something with all dependencies, try it, and then I
can easily remove everything that was just installed.

But for rolling back an update, yum requires that the old package is
still available.  We only keep the very latest version in the updates,
so unless your previous version was from the initial release, you're out
of luck.  My last yum-update hit 19 packages, and only 7 can be
downgraded by yum-history-undo.

Josh