Re: [gentoo-dev] net-dns/dnssec-root: Blind stable on arm, critical bug 667774

[Jeroen Roovers <jer@gentoo.org>]

On Thu, 11 Oct 2018 19:14:00 +0200
Thomas Deutschmann <whissi@gentoo.org> wrote:

> > 1) Someone blind-stabled something on arm and it broke (doesn't
> > build?) 2) The arm team failed to mark a package stable before a
> > hard deadline (DNSSEC key rotation)

"Blind-stabled"...

> But that's not the point here. The point was to get some attention
> that again we have a lacking architecture (net-dns/dnssec-root is not
> the only package where ARM arch team is lacking behind) which affects
> anyone "trusting" somehow in STABLE keywords.

The trustworthiness of stable keywords has been eroding for years.

It started when ago@gentoo.org found ways to automate "compile-testing"
on many architectures, taking work away from people who actually cared
about those architectures, reducing arch team efforts to trying to
catch up with ago's work. While it was a valiant effort to reduce
architecture teams' backlogs, I couldn't stress enough at the time how
taking decisions on behalf of all users of an architecture isn't
something you can automate, for instance putting effort into
stabilisations for (sets of) packages that may have ceased being useful
on respective platforms, so that users would switch to cherry-picking
their own stable targets instead of relying on stable keywords to still
be meaningful.

Where "compile-testing" failed as runtimes do not necessarily reflect
that what is being compiled does actually work, architecture teams had
to pick up those pieces of now incorrectly stable-keyworded packages
that got strewn around in automation's wake.

Even more recently a new trend arose where just about anybody who
maintains a package takes stabilisation decisions, usually citing some
"all arches" policy, and in this case "blind-stabled", on behalf of
architecture teams. This new direction is likely based on the same
backlog pressure[0], a sense of emergency because of security issues,
and the desire to clean up obsolete ebuilds.

Having mostly stepped away from concerted stabilisation efforts myself
for those reasons among others, I can only speak for myself in stating
that my trust in stable keywords is at its lowest ever ebb.


Kind regards,
     jer


[0] Wait, didn't we get rid of that? Ah no, the automation effort
reduced architecture team involvement to the point of being
non-existent.