* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* test(): test dynamic max workers
* Clarify test description and improve CI
**What**
- Initiate the framework package (which is just a place to move things around for now)
- move the config loader and related resources as well as the `ConfigModule` type
- Create a ConfigManager singleton which prepare and store the config (later can be stored entirely in the container) and allow for easier test override
- re export the logger from the framework
- replace medusa config loader with the framework one
- `build` run type check on tests as well but `prepublishOnly` will not fail on build if tests are typed broken
FIXES FRMW-2607
FIXES FRMW-2609
FIXES FRMW-2614
FIXES FRMW-2618
what:
- runs module and api integration tests in matrix strategy
- v28 of jest comes with sharding support, which we can utilize when we upgrade
- splits the tests by number of matrix jobs and run them in parallel
- This introduced some flakiness in some specs, but couldn't reproduce locally. Those have been skipped for now.
- uses swc/jest for added performance
- Locally, a chunk took 90 seconds with babel and 30 seconds with swc.
- This translated to 2 mins saved per shard in CI, but haven't tested this enough.
Few things to keep in mind:
1. You need to set MEDUSA_FF_MEDUSA_V2 to true before running the tests to run with the v2 API
2. You can use the `breaking` function to differentiate between v1 and v2 differences. This can help us identify what was breaking pretty quickly afterwards
3. You will need to run specific tests for now instead of all if you want to target v2. I think that's fine though, as we don't really need these to run on every PR until we have feature parity (and by then, all tests would be both v1 and v2 compatible)
**note: Adrien**
- add a new way to load modules only to run their loaders comparable to the way to run the migrations only
- improve tests runner to cleanup the data properly as well as re running all loaders and core defaults
Co-authored-by: Adrien de Peretti <25098370+adrien2p@users.noreply.github.com>
What:
Workflow Engine API.
Endpoints for:
- List workflow executions
- Run a workflow
- Set async steps as success or failure
- Retrieve the details of a workflow run
**What**
Integration tests to validate the workflow compensation.
Also, fix the transaction state when the workflow is compensating and some steps does not have any compensation
* chore: Use caret for all Medusa deps
* Create wild-balloons-push.md
* Address PR feedback
* force build order
* add missing dep
* add missing dev deps
* addresses last comments
## What
Harmonize and clean-up yarn test commands convention.
## Why
Current strategy to filter integration tests from unit tests was cumbersome to maintain. Also, some packages declared `test:unit` which ended up being an alias to `test`.
## How
* Remove all occurrences of `test:unit`
* Rename all `test` commands under `/integration-tests/**/package.json` to `test:integration`
* In the root `package.json`, rewrite `test:integration` command to run `turbo run test:integration`
* In the root `package.json`, remove filtering on the `test` command
* Introduce a new `test:integration` task in the Turborepo config.
