#๐ v0.11.7 - 11th June 2024
1 messages ยท Page 1 of 1 (latest)
Hallo all ๐ just going to start a release thread, since we've not done one this week.
I'd like to aim to release something next week, though not sure what. It feels like we're getting close to a v0.12.0, with some larger changes in the pipeline (lots of api breaking things we'd like to do, and maybe the telemetry changes that @brazen wave, depends on how breaky that is) is working on.
That said, not sure what's "ready". I think maybe we aim for at least one more release in the v0.11.X series, and then, aim to do a bunch of breaking changes/etc for v0.12.0?
For context:
- v0.11.7 milestone: https://github.com/dagger/dagger/milestone/48
- v0.12.0 milestone: https://github.com/dagger/dagger/milestone/45
cc @brazen wave @bitter seal @winged oak @frail idol @formal dust @left willow @sick dragon
If we're going for v0.11.7 next week, I'd probably say we should aim for Tuesday/Wednesday.
@bitter seal @dreamy linden Done addressing all feedback on https://github.com/dagger/dagger/pull/7452! thanks again ๐
As @dreamy linden mentioned it would be good to have https://github.com/dagger/dagger/pull/7031 in the same release since the OTel changes will make an outdated client hang. (Looked into a fix for backwards compat. and it's a bit complicated, wouldn't want to keep it for long, so kind of just delaying the inevitable.) Taking a look at that now but getting near EOD
Will take a quick look again later today!
I'd really like to get https://github.com/dagger/dagger/pull/7315 in too. Needs to come after OTEL and is another doozy of a PR but will be good to review once OTEL is in main (currently has all the commits from Alex's PR).
It also creates incompatibilities with old CLIs (due to changes to the engine's API it serves, just pure http now rather than insane grpc tunneling), so same story of ideally having 7031
โจ v0.11.7 - 11th June 2024
cool! let's go for a v0.11.7 release, and intentionally break our client-server incompat ๐ i'll revive 7031, and add some tests ๐
i'm also pulling in https://github.com/dagger/dagger/pull/7474 to the milestone, not really any huge changes here, just some progress tidy-ups, and making the cloud url a bit more prominent
๐๏ธ note! please add your issues/prs to the milestone! if they're on the milestone, they'll make it into the release, and the release will block until we can get them fixed/merged ๐
also please add any release notes if applicable (it saves me so much time)
woo ๐ finally rebased https://github.com/dagger/dagger/pull/7031 (and added some automated tests) - from manual testing as well, this works all the way back to v0.10.X, which is kinda neat ๐
(potentially we might also be able to deprecate+remove checkVersionCompatibility for v0.12 as well ๐)
Just FYI that afaict right now all TS module users (including those on any previous release) are broken atm unless they happen to have cache for the npm install -g tsx step in the TS SDK: #maintainers message
Options are either A) open an issue on the tsx repo and hope they fix it or B) do an engine release ASAP
Thought of an option C, which is a wrapper of the TS SDK that works as a git ref and can be used in the meantime before we do the release: #maintainers message So unless there's any objections I say we go with that (I'll post to the TS channel and such) and just do the engine release ASAP early next week
Sounds good to me! Pretty unlucky but seems better than rushing a release on a Friday. If we did that I'd want to do a hotfix, since my OTel PR is merged, but not the version compat. check PR yet
Yeah exactly I'm just ruling out doing all of that by EOD essentially, doesn't quite reach the threshold of fire to call for that if there's a workaround
Also, we can postpone https://github.com/dagger/dagger/pull/7315 for the subsequent release in the interest of getting it all out faster, was nice to have but not worth blocking on given it will take some time to review
That being said, I did fill it out the PR description and took it out of draft, so if anyone wants to review it, it's ready ๐
a few prs/issues still in the milestone need reviews/work: https://github.com/dagger/dagger/milestone/48
@hot rivet i took https://github.com/dagger/dagger/issues/7490 out of the milestone, though someone should find time to do it eventually ๐ข
i keep thinking i will get to it, and it never seems to happen
going to cut the release soon
sending PR in a bit
What is the issue? I have observed that sometimes when calling Dagger the first time (engine gets started) the connect step takes an excessive amount of time. ref: https://discord.com/channels/7076...
awesome, will pull this fix into the release
tagged v0.11.7
hitting some weird instability in the new engine: #infrastructure message
going to keep going, and get all the sdks out
retriggered the jobs, it could be a flake (doesn't seem to be)
fyi @brazen wave @bitter seal
Will be online in a few, does this happen at all locally when provisioning the new engine version the first time?
we seem to hit it in the test/lints, looks like it's at the end of a job related to that
we bumped to the latest buildkit version - could maybe be related to that
or maybe something service-y again
@hot rivet @pine oyster we should be good to update the playground
dagger-for-github here: https://github.com/dagger/dagger-for-github/pull/132
yeah this is definitely weird, doesn't seem service related: https://github.com/dagger/dagger/actions/runs/9469086782/job/26086942090?pr=7616
we seem to not be getting the end of the initialize span somehow
Seeing if I can repro locally, pulled the improve releasing PR locally and running those sdk lint jobs, no hangs so far
turned on the cloud token for those "waiting steps", kinda curious what the traces look like for them
For this typescript failure: https://github.com/dagger/dagger/actions/runs/9469257837/job/26087498583?pr=7616
Why are we running golangci lint in the middle of it? exec golangci-lint run -v --timeout 5m 
No idea if related, but certainly odd
ah we're linting the typescript runtime
ohh okay, that makes sense
Actually I just hit an attempt locally of running call sdk typescript lint that seems to hang, looking at it
Playground: https://github.com/dagger/dagger.io/pull/3785 cc @hot rivet
@frail idol when you're around, noticed that the go/php sdks are having some issues - context and fix in https://github.com/dagger/dagger/pull/7631
Fixes #7630.
Introduced in 8421636 (#7554).
This caused the https://github.com/dagger/dagger-go-sdk read-only copy to include the full dagger/dagger repo, instead of the filtered one.
Done!
again, will take an action to work out why our provisioning tests didn't catch this
am i good to retag the go and php onto that merged pr? requires a force-tag push, but since there are no changes to those sdks it should be "alright" ah you mentioned you'd do it, nvm
The PHP SDK was good: https://github.com/dagger/dagger-php-sdk/releases/tag/v0.11.7
The Go SDK had to be force updated: https://github.com/dagger/dagger-go-sdk/releases/tag/v0.11.7 - this is good now
wonderful ๐
The problem is that the Go package did not pick up this new version: https://pkg.go.dev/dagger.io/dagger@v0.11.7
Also, we seem to have published a bunch of other packages that does not look right - like this one: https://pkg.go.dev/github.com/dagger/dagger
we don't publish these - go picks them up automatically
Do we expect this to be resolved? https://github.com/dagger/dagger/issues/7630
indeed - it might refresh - that would be nice
We now have the right tag in place
Trying to pull it explicitly in case that fixes it
I am still seeing the same error as reported in that issue. Moving the discussion there: https://github.com/dagger/dagger/issues/7630#issuecomment-2163374092
A separate issue which I have noticed is that PHP SDK v0.11.7 did not get published. Publishing now manually:
TL;DR we will not be able to fix the v0.11.7 Go SDK release. Publishing v0.11.8 as soon as we address some of the edge cases that we discovered sounds most reasonable to me.
if we publish v0.11.8 - we will be making an intentionally broken release - are we alright with that? cc @brazen wave @bitter seal
there is a legitimate timeout of testdev on main
Where does the breakage come from?
Probably an old version, v0.11.5 or so
could we re-release just Go "0.11.7" as 0.11.8? or are we keen to keep the versions in lock step
We could release v0.11.8-1 or something (v0.11.8-rc1 would also work)
That's a prerelease
Then we don't do a whole new release, just one new tag
might not get picked up by Go versioning (not sure)
And it'll be replaced by v0.11.8 when it arrives
Also, I was not able to get the PHP SDK to force update to v0.11.7 . Still not sure what is happening with that one, but both SDKs are affected. The Go one is definitely far more important to get fixed.
The other issue with doing an engine released now, is we just merged a large pr that refactored a lot - I was hoping to give it more time to settle to avoid just hitting similar issues in several days again
Yes, we could try tagging v0.11.8-rc1 for Go only. Even if it will use Engine v0.11.7, that should be OK.
It should upgrade automatically in the same way, we don't have a version conflict when we do v0.11.8, it sounds good to me
We could push it directly to the go-sdk repo as well (same for php)
OK, how about this?
- let's retract
v0.11.7explicitly since it's broken - once this is merged & sync'ed to dagger-go-sdk, we can tag
sdk/go/v0.11.8-rc.1ondagger/daggerand let the workflow do the rest
WDYT?
OK, let me put together a PR that implements this.
Fyi, I'm going to be at a tech meetup in a few minutes, so will have slower response times
PHP SDK is now fixed, v0.11.7 showing as expected https://packagist.org/packages/dagger/dagger
working on the mixed up tags/branches thing now