Silly bodhi karma games

[Michael Schwendt]

On Thu, 19 Mar 2015 09:44:27 -0400, Stephen Gallagher wrote:

> I think it can be a judgement call on certain packages. For example, I 
> maintain the Review Board packages which almost never get karma from 
> more than one person (and that usually only for whichiver Fedora or 
> EPEL branch that person is currently deploying to).

Always there is at least one example of an unhappy/impatient packager, who
considers Bodhi "unnecessary bureaucracy" or who finds examples of updates
that "don't need any testing" and could be unleashed much more quickly.
[The repo metadata would change a hundred times a day, if packagers could
push changes to the repo without any delay.]


I only have a problem with those packagers who try to outsmart Bodhi (and
its defaults (e.g. karma threshold "3") and shoot themselves into their feet
with runtime breakage or dependency breakage.

 * "It's just a rebuild of packages because of a SONAME bump in some lib!"
 * "I only cleaned up the spec file!"
 * "It's just a minor version update!"
 * "It's just a rename of package(s)!"

It has happened before. Broken rebuilds because of changes in the tool-chain
(previous build several months old!), in the other build dependencies, because
of regression (or even a brown paperbag bug) in the upgraded lib, or even
because of expired buildroot overrides. Some packages "sleep" in the
distribution for months, survive the development cycle and alpha/beta test
releases of the distribution, but are replaced too quickly with updates/upgrades
_without_ any adequate testing.

> Even at karma 1, 
> most Review Board packages sit in updates-testing until the timeout 
> passes.

Which is a good thing.

Waiting for the timeout is a good habit. At least it gives users a chance
to test the update.

"But nobody gives feedback on my test updates!"

And obviously, for those packagers who don't mind the timeout, Bodhi could
improve and offer an option to push an update to stable automatically
after the timeout. Or after a customizable timeout, so e.g. the maintainer
could request "21 days" instead of 7 days. Meanwhile, a sufficient number
of karma changes could still speed up the release or veto it, too.

> Hopefully, some of the new changes in Bodhi 2 will improve upon this 
> situation. I hear that's coming Real Soon Now.

It's the packagers, who should act more responsibly and accept any aid
offered by Bodhi instead of trying to rush out updates.