1451694 - Snap: Ship nightly on the edge channel

Status: RESOLVED WORKSFORME

(Release Engineering Graveyard :: Release Automation: Snap, enhancement)

Description

[Kestrel]

Until a Developer Edition Snap is available (Bug 1447064) it would be nice if the "edge" channel honored xpinstall.signatures.required so that unsigned extensions could be tested longer in Snap across restarts rather than just being loaded temporarily.

Comment 1

[Johan Lorenzo [:jlorenzo]]

Thank you for filing this issue, Kestrel! Changing the edge channel requires us to make a new build with a different configuration[1]. Although, that's perfectly doable, we'd prefer to stick with one configuration.

`xpinstall.signatures.required` is exposed on Nightly too. We eventually want to have Firefox Nightly on the edge channel. Let me rename this bug to reflect this.

Please let me know if you have any comment or question. 

[1] https://github.com/mozilla-partners/canonical/blob/02e615a41cc1cf9a9fede2020e37517dc3b0707e/desktop/ubuntu/distribution/distribution.ini

Comment 2

[Johan Lorenzo [:jlorenzo]]

In order to prevent further confusion about what's on edge and what's on beta, I just closed the edge track. Since bug 1447263, we now ship beta directly on the eponym channel.

Comment 3

[Kestrel]

Honestly a Nightly Snap doesn't seem that useful at the moment. It adds another variable to testing which won't be replicated in mozregression and since Snap is currently limited to one channel at a time, most will use it for their main browser which is more likely to be Release or Beta. The fact that nobody has filed a bug requesting a Nightly channel is indicative of this.

Nightly is also not a good platform for testing extensions since it is too unstable which is why Developer Edition exists as a sweet spot between new features, permissive settings and stability.

Comment 4

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Is it possible that a Nightly Snap could help get more people to test Nightly in more configurations?  The Linux Nightly userbase is not very large, and there have been some problems with regressions not being found until late in the release cycle (or sometimes *after* release).

It would also help with changes that interact in unexpected ways with the Snap packaging itself (see bug 1450740), but those might turn out to be infrequent.

Comment 5

[Kestrel]

Delta updates (Bug 1453026) should be a prerequisite to avoid full 200MB updates twice a day.

Comment 6

[Jordan Lund (:jlund)]

mconnor, pdol: hi, would either of you know the current priority of this bug? I'm triaging H2 items and trying to determine:

• How important this is? Business impact.
• Any hard or soft deadlines around this bug? Soft being "when we would like to have it done by".
• What happens if we don't do it?

Comment 7

[Vicky Chin [:vchin]]

Hi Romain, can you assess?

Comment 8

[Romain Testard [:RT]]

Here is my summary of the bug discussions:
Opportunity:

• Improve Nightly quality (Get more people to test Nightly through the snap distribution)
• Improve snap package quality (Get more people test the snap packaging)

Cost:

• Per comment 3 it adds another variable to testing which won't be replicated in mozregressionmore testing
• Per comment 1 this would require releng work (a new build with a different configuration)

Comments:

• Nightly is biased towards Linux with over 7.2% of nightly population on Linux VS 4.6% of release population on Linux (https://sql.telemetry.mozilla.org/queries/68974/source#174114) - if representativity is the goal on nightly then it does not seem needed to have more Linux users compared to other OS
• It's hard for me to assess value of having more people testing snap builds on Nightly without data on snap - do we have data on downloads, DAU or share of Linux users through the snap package? We're considering adding telemetry (distribution_id) for the flatpak to help understand volumes of users.

Summary:
The main value seems to be around improving the snap package quality although a comment referred to snap packaging issues not being frequent and we also don't seem to have data on snap usage to assess the user base that could be impacted. Without more data on snap Firefox user base I'd recommend not prioritizing this work that does not seem to provide enough value to balance the QA and releng work necessary.

Comment 9

[Aki Sasaki (not active)]

Olivier, is this something Canonical can/wants to take on?

Comment 10

[Julien Cristau [:jcristau]]

I believe this was done a few months back, afaict 99.0a1 is on the edge channel currently:
https://git.launchpad.net/~mozilla-snaps/+git/firefox-snap/log/?h=nightly

Comment 11

[Olivier Tilloy]

Confirmed. The caveat is that nightly builds are only done for amd64 currently. But that should be good enough for the vast majority of linux desktop users who want to test-drive nightly as a snap.