On 3/7/25 09:48, Benny Siegert wrote:
On Fri, Mar 7, 2025 at 1:50 PM Jason Bacon <jtocino%gmx.com@localhost> wrote:It's very unfortunate that it has come to this, but sometimes a fresh start is the only real solution.Reading this, I wonder: Is pkgsrc dead? I see you, Jonathan and Nia moving over. That's a huge loss.
It's far from dead at this moment, but I worry about its future if the community doesn't embrace some strict QA measures. I'd actually go a step further than the dreckly CI, and require testing of direct dependent packages as well. That's resource-intensive, especially for things like cmake, boost, gcc, but it could be managed if updates to such common dependencies are kept to a minimum (another one of Jonathan's suggestions). Dependent checks could be CI, or policy + some trust of our committers. Recursive testing via a partial bulk build is the best case, of course, but I think building updated package + direct dependents would prevent more than 99% of regressions.
With the future of quarterly branches in doubt, there may be no such thing as a dependable pkgsrc tree going forward. The quarterly branches were the only thing that made pkgsrc viable for scientific computing on Mac and Linux. And as Jonathan has tried hard to convey, the "break things for 3 months and fix it all in a week" model isn't sustainable.
The "zero regressions" model employed by dreckly makes much more sense to me. I think this approach saves a lot of man hours in the long run + I've always viewed firefighting as the worst possible use of my time.
-- Life is a game. Play hard. Play fair. Have fun.