fix: broken CI #771

Merged
decentral1se merged 3 commits from fix/ci-02-15 into main 2026-02-15 16:34:01 +00:00
Owner

Trying to get the integration suite back in shape with 2 fixes.

c7cdf258a6: this test was flaking on CI and I realised the test itself is borked (I wrote it 🙃). Along the way I realised that there is no check for a borked tag on abra app deploy, which is now covered and tested. I believe the tag parsing can still explode if you have a deployed borked tag but I hope that use-case will go away now it's blocked from the deploy side.

e56e1157bd: Follows up on #753 and #761. @ammaratef45 I realised that we had made recipe.Ensure(...) for this update logic, which handles the various cases internally (e.g. if the user passed --offline or not). When I wrote f9ea7506d0 it started breaking a shit tonne of abra app new integration suite tests because a direct EnsureUpToDate deletes unstaged changes 😆 I switched from WarnF to Fatal because I reckon that if you can't update and you want to, something much worse is wrong and you'd want to know about it.

Trying to get the integration suite back in shape with 2 fixes. https://git.coopcloud.tech/toolshed/abra/commit/c7cdf258a6832aa978f2b72a066cb98019c17079: this test was flaking on CI and I realised the test itself is borked (I wrote it 🙃). Along the way I realised that there is no check for a borked tag on `abra app deploy`, which is now covered and tested. I believe the tag parsing can still explode if you have a deployed borked tag but I hope that use-case will go away now it's blocked from the deploy side. https://git.coopcloud.tech/toolshed/abra/commit/e56e1157bd8235da7c87785a9bf58f38707617c9: Follows up on https://git.coopcloud.tech/toolshed/abra/pulls/753 and https://git.coopcloud.tech/toolshed/abra/pulls/761. @ammaratef45 I realised that we had made `recipe.Ensure(...)` for this update logic, which handles the various cases internally (e.g. if the user passed `--offline` or not). When I wrote https://git.coopcloud.tech/toolshed/abra/commit/f9ea7506d00e752c84767232ef08f6430ef73b02 it started breaking a shit tonne of `abra app new` integration suite tests because a direct `EnsureUpToDate` deletes unstaged changes 😆 I switched from `WarnF` to `Fatal` because I reckon that if you can't update and you want to, something much worse is wrong and you'd want to know about it.
decentral1se added 3 commits 2026-02-15 16:10:47 +00:00
fix: ensure and fail for updated recipes
Some checks failed
continuous-integration/drone/push Build is failing
0ccf3d2b12
chore: i18n
All checks were successful
continuous-integration/drone/pr Build is passing
continuous-integration/drone/push Build is passing
227d37dc26
decentral1se merged commit 227d37dc26 into main 2026-02-15 16:34:01 +00:00
decentral1se deleted branch fix/ci-02-15 2026-02-15 16:34:01 +00:00
decentral1se added this to the Abra v0.13 project 2026-02-15 16:34:10 +00:00
decentral1se moved this to Done in Abra v0.13 on 2026-02-15 16:34:15 +00:00
Owner

Hmmm, can we update the error log message and refer the user to using the -o flag if failure to update the recipe was to be an expected outcome? (working on a local only network or something like that)

Not a use case that I expect many people to have but seems easy to do, what do you think?

Hmmm, can we update the error log message and refer the user to using the `-o` flag if failure to update the recipe was to be an expected outcome? (working on a local only network or something like that) Not a use case that I expect many people to have but seems easy to do, what do you think?
Author
Owner

Not a very satisfying answer I imagine but I trust that people will see a network related error and check --help and find --offline. I'm pretty wary of going too deep into improving error messages because it is hard to detect when the context happened to show the right message.

Not a very satisfying answer I imagine but I trust that people will see a network related error and check `--help` and find `--offline`. I'm pretty wary of going too deep into improving error messages because it is hard to detect when the context happened to show the right message.
Owner

you're right, I guess it also adds natural language strings that can't be tied to variable logic (like if we change the offline option behavior we will 99% of the time fail to update all strings that talk about it)

cool, it's more satisfying of an answer than you assumed 😆

you're right, I guess it also adds natural language strings that can't be tied to variable logic (like if we change the offline option behavior we will 99% of the time fail to update all strings that talk about it) cool, it's more satisfying of an answer than you assumed 😆
Sign in to join this conversation.
No description provided.