We used the 26.04 charming survey to shape the 26.10 cycle. To close the loop, here’s a summary of requests from the survey that made it into our plan, broken down by what’s shipped, what’s in progress, and what has slipped into the next cycle.
This isn’t just the pulse reports concatenated. It covers the things where the survey is a part of the reason that the work happened, not everything we did this cycle. The cycle also isn’t quite over yet, so some of the “in flight” items might land before it closes.
What landed
Jubilant’s logs
The most consistent complaint about Jubilant in 26.04 was the logging: too verbose by default, and hard to pick the interesting line out of. That resulted in jubilant#351, which moves the noisy diff lines down to DEBUG and logs application and unit status changes at INFO. pytest-jubilant got its own logger and stopped writing lines on failure, and there’s now a how-to for configuring the logs.
Pinning a Juju version in Concierge
Concierge can now install a specific Juju snap revision, to use when a Juju release breaks and you have to hold still for a while. Note that we recommend that your CI tests against edge, and that you’re tolerant of failures there: investigate them, report regressions to the Juju team, and don’t treat this as a blocking issue for your charm. Hopefully, that avoids these kinds of problems making their way to stable and needing to use this feature.
Concierge documentation
“Better docs please” and “document the concierge.yaml fields thoroughly” both came up. Concierge now has a Diataxis documentation site, rather than only the README.
charmcraft test versus Spread
One person asked why the two behave differently and whether they could be unified. There’s not much Charm Tech can do on this directly, but the integration testing guidance now recommends Spread directly (rather than charmcraft test), and the integration test how-to was split into its own guide. We had several conversations with the Starcraft team and others about this, and we’re hoping that there will be improvements in 27.04. You should also take a look at charm-ci, if you haven’t already.
Library documentation that isn’t just reference
Several answers said the charm library docs are broad but shallow, that you end up reading the source, and that you have to find a charm that uses a library to work out how. Packages in charmlibs can now carry tutorials, how-tos and explanations, not only generated reference docs. Note that the follow-up work is for each of you that are maintaining libraries to make sure you have these other types of docs! 
Charmlibs metadata and links
We populated the library metadata fields across charmlibs and fixed the project URLs, and added library tags, which the store will use.
A git workflow for outside contributors
We were asked to be clearer in terms of our expectations. Ops now has a section for this in CONTRIBUTING.md, and charmlibs got a reworked CONTRIBUTING.md and an AGENTS.md.
Moving libraries into charmlibs
We helped other teams moving their libraries, and did the tracing interface library ourselves. The snap library also reached 2.0.
Documentation examples and the “bloated pages” problem.
The Ops documentation moved to its new home, best-practice admonitions got stable anchors so they can be linked to individually, and there are new guides on subordinate charms, debugging Kubernetes charms, machine workloads, managing files in containers, and charm naming.
That moves things in the right direction for examples (and we’ll continue with work in that area), and we hope it helps address “bloating”. There’s still work to be done in terms of the constellation of charming docs (Juju, ops, charmcraft, charm libraries, concierge, Jubilant, …).
What’s in flight
- Pebble’s trust store, which is the TLS certainty request (pebble#780), is spec’d and being implemented. OTLP telemetry forwarding is in progress.
- Multiple-charm state-transition testing is spec’d and in the 27.04 draft roadmap as implementation work.
- Modelling the HTTP proxy environment variables in Ops was on the 26.10 roadmap, but it’s looking likely that it will slip to 27.04.
- The same applies to Terraform-first integration testing documentation, unfortunately.