refs 2f1123d6ca
Usage of the raw Error class has been deprecated in favour of our own
errors, which are more descriptive and have built in HTTP status codes.
This also updates the same errors to use @tryghost/tpl for the error
messages, which is the new pattern we are following in order for us to
deprecate the i18n module.
no issue
- i18n is deprecated in favour of `tpl`
- normalized method syntax so `add` matches the rest of the controller's methods (fixed a complexity warning but was not the primary intention)
refs 2f1123d6ca
refs 6f1a3e1774
- As per refed commits, we are removing deprecated use of `new Error()` in the codebase
- Exposed few internal from commands module methods for easier testing, otherwise it was turning into neverending mocking show
refs 20aea8c819
refs bf587d4055
- The 1000 ms limit was not enough to pass consistently for some slow unit tests. Doubling the previous threshold to be on the safe side of the slow CIs.
- We should limit this back down to 1000ms or less during next cleanup!
refs bf587d4055
- The change broke regression tests and due to this change being out of
scope of the cycle dicided to rever it and increase the time limit on
unit tests instead
- We should focus on refactoring moving slow unit tests in the next
cleanup period
refs https://github.com/TryGhost/Team/issues/696
The userAuth spam prevention logic is reused, but a new piece of
middleware has to be created so that we can use a custom lookup key to
conatin the member email.
We must also add json parsing middleware to the route so that the brute
middleware can read the email.
The express body-parser middleware handles multiple instances on the
same route, so this doesn't cause problems upstream.
https://github.com/expressjs/body-parser/blob/1.19.0/lib/types/json.js#L99-L103
no issue
- Two affected unit tests were triggering the "slow unit test" error on CI which gave a clue they were either:
- written badly
OR
- were not unit tests at all
- The latter proved to be true and the tests were moved where they belong - into the regression test suite with much more forgiving running itme constraints.
- Unit tests should run quick ALWAYS as that's the first thing that gets triggered before pushing improvements to main. Faster - better.
refs https://github.com/TryGhost/Team/issues/527
refs https://github.com/TryGhost/Ghost/issues/10790
- Frontent has to have as few as possible coupling points with the Ghost Server API. By design that point has been a "proxy.api" property that will become more and more constraint in the future based to limit the surface of frontend interaction with servers's API
- Removing `.../server/api` requires in favor of using a proxy decreases direct coupling
refs https://github.com/TryGhost/Team/issues/542
The members-api has been updated to use the passed `created_at`
property when creating members for the status change events, as well as
using the subscription start_date for members paid subscription events
This is a duplicate of 7eb821b3f5 but
with updated dependencies.
closes https://github.com/TryGhost/Team/issues/743
Unlike tags, a label has a unique constraint on its `name`. So saving a new label on member with the same name as existing label fails with error due to unique constraint error.
- adds id for new label to match existing label if they are the same name, which avoids creating a new label
refs https://github.com/TryGhost/Team/issues/542
The members-api has been updated to use the passed `created_at`
property when creating members for the status change events, as well as
using the subscription start_date for members paid subscription events
refs https://github.com/TryGhost/Team/issues/880
The aggregate for `paid_delta` was incorrect as it did not handle the
case where an event went from paid->comped or from comped->paid. This
resulted in an overcount for paid members.
refs https://github.com/TryGhost/Team/issues/860
- Slow unit tests cause longer waiting time to deliver code to main. Before this fix the test was taking a whooping 6s on average
- The main cause of the delay was a downstream's package (got) default retry logic that was taking up a lot of time. The test is meant to check a "500" response handling not a retry logic, so disabling it for this specific test makes sense
refs https://github.com/TryGhost/Team/issues/860
- `nock` has been used in other similar unit tests. It was proven to have easy to understand and use API. By not making up custom code through express/httpServer stuff we reduce the amount of boilerplate code significantly - makes tests a lot more readable
- Bonus, removes couplint dependency to "shared" express module
refs https://github.com/TryGhost/Team/issues/860
- Slow unit tests cause longer waiting time to deliver code to main. Before this fix the test was taking a whooping 6s on average
- The main cause of the delay was a downstream's package (got) default retry logic that was taking up a lot of time bypassing the retry logic present in the default scheduler itself
refs https://github.com/TryGhost/Team/issues/756
When running the tests it was possible for this middleware to be
instantiated before the settings cache, resulting in an undefined
'session_secret' setting being passed. This would cause tests to fail.
Tracking this down proved difficult, so the fix was made here, by
instantiating the express-session middleware only once a request needs
to use it, we can be confident that Ghost has completely started.
refs 2f1123d6ca
refs 6f1a3e1774
- As per refed commits, we are removing deprecated use of `new Error()` in the codebase
- This bit cleans up the rest of `new Error()` usage in MEGA service
no issue
- Exposing internal methods out of the module is a non-standard practice. Adding `_` prefix allows to signal that this method is not for general use.
- When mega is refactored into a proper class this method will become exposed anyways
refs 2f1123d6ca
refs 6f1a3e1774
- As per refed commits, we are removing deprecated use of `new Error()` in the codebase
- This bit cleans up `new Error()` usage in MEGA service
refs 2f1123d6ca
refs 6f1a3e1774
- The use of new Error() has been deprecated. Refactoring the migration to use `createIrreversibleMigration` made most sense to have central error handling for migration which are not meant to be reverted.
no issue
- This mehod has an important `tableSpec` parameter which MUST be present when creating a new table migration. Having a description in form of the JSDoc somewhat helps this cause
- Next best improvement would be throwing an error if the parameter wasn't present, but that would require a bigger refactor backporting all usages of `addTable` method
no issue
- `RELEASE_TOKEN` is currently a GitHub personal token, but this
has some downsides:
- if the token ever expires and I'm unaware, it'll break the release
process
- GitHub Releases say the creator was `daniellockyer` even if someone
else actually did the release
- this commit switches over to using the built-in `GITHUB_TOKEN`, which
is owned by the `github-actions` app and should never expire
- aside from that, Ghost releases will be created by the neutral
`github-actions` account
refs https://github.com/TryGhost/Team/issues/841
When using our development tooling Ghost should always start, instead of
exiting with an error. This check for the WEBHOOK_SECRET env var was the
primary cause of Ghost erroring in development, so it's been switched
with a warning.
