refs https://github.com/TryGhost/Team/issues/1306
- This removes the limitation described in commit ff46449ad6
- The only edge case is that when a publication changes their timezone, it will have maximum 24 hours where the member last_seen_at could be incorrect
refs https://github.com/TryGhost/Ghost/pull/14197
- Moved from updating the last_seen_at value "at most every 24h" to "at most every UTC day".
- It will simplify explaining the following behavior: a publication is set in UTC-10, a user visits at 2pm on Monday and at 1pm on Tuesday, the last_seen_at value is still Monday.
- There is no way to go around the above issue due to the technical constraint of updating the `last_seen_at` value at most once a day.
- This might create database write spikes at midnight UTC
refs https://github.com/TryGhost/Team/issues/1306
- Contains all services that listen on member events
- Only contains the last-seen-at-updater service for now
- Listens for `MemberViewEvent` events to update the `member.last_seen_at` timestamp
- Updates after 24hours of the last timestamp to avoid too many writes
- Also updates when the value is NULL
- This is using the existing `last_seen_at` value to avoid an SQL query when no writes are required
refs https://github.com/TryGhost/Toolbox/issues/203
- without `--all`, c8 ignores files that should be included in the
coverage score but aren't used in tests
- this means we have artificially high scores in places where this isn't
used
- this commit adds `--all` where previously missing
- where this fails `--check-coverage`, that has been removed for now
refs https://github.com/TryGhost/Toolbox/issues/214
- After disconnecting Stripe API the `_configured` flag stayed as `true`, causing behaviors as if Stripe was still connnected.
- The `api.configure` method was never reachable when disconnecting Stripe API, thus causes hanging "configured === false" state inside of the StripeAPI wrapper
refs https://github.com/TryGhost/Team/issues/1374
We cannot update payment details for members which we don't know about,
so returning and giving a successful response to Stripe is the correct
thing to do.
- these are some of our first packages here and use `master` in their
repository link
- we've since switched the repo to use `main` but these links were not
updated
- this commit updates the links
refs: 23b383bedf
- @tryghost/error constructors take an object, not a string - the expectation is that message, context & help should all be set
- This does the bare minimum and just ensures message is set correctly
refs: 3a1c51672c
- @babel/eslint-parser was added to enable support for static members on classes but this causes other problems as it fails on eslint ugprades etc
- a better solution is to set ecmaVersion to 2022, as our default is 2018 which does not support static, yet 2022 does
no-issue
The new tests added do not meet the default minimum of 90% and we do not
want to dedicate more time to writing tests for this package right now.
refs https://github.com/TryGhost/Team/issues/1322
We no longer restart the Members service based on the Stripe service
being updated, which meant that if it was initially configured with
missing URL's and later Stripe connected, it would not get the new
config until a server restart. This moves the last of Stripe config into
the Stripe service, so that all things concerning Stripe can be handled
in one place and updated together.
refs https://github.com/TryGhost/Team/issues/1319
Rather than allowing our code to attempt to speak with Stripe and error
there, we opt to fail fast and throw validation errors so that our API
will respond with a 422, the existing pattern for validation errors.
refs: https://github.com/TryGhost/Toolbox/issues/203
- Without the all flag only files touched by tests are counted
- This shows our unit-testing picture more clearly
- Will get this change rolled out across all repos and packages ASAP, but for now this one seemed most important
no-issue
There is a bug in the event repository code where filters are not
correctly applied, this results in the api verification being extremely
trigger happy. For now we're removing the logic, with the plan to revert
this commit once the event repository is fixed.
no-issue
We were incorrectly mixing transactional and non-transactional
operations. An e2e test in Ghost will be merged shortly which caught
this problem.
