fixes https://github.com/TryGhost/Team/issues/481
This change fixes an issue when multiple images with the same name are
uploaded in parallel. The current system does not guarantee that the
original filename is stored under NAME+`_o`, because the upload for the
original file and the resized file are happening in parallel.
Solution:
- Wait for the storage of the resized image (= the image without the _o
suffix) before storing the original file.
- When that is stored, use the generated file name of the stored image
to generate the filename with the _o suffix. This way, it will always
match and we don't risk both files to have a different number suffix.
We'll also set the `targetDir` argument when saving the file, to avoid
storing the original file in a different directory (when uploading a
file around midnight both files could be stored in 2023/01 and 2023/02).
Some extra optimisations needed with this fix:
- Previously when uploading image.jpg, while it already exists, it would
store two filenames on e.g., `image-3.jpg` and `image_o-3.jpg`. Note the
weird positioning of `_o`. This probably caused bugs when uploading
files named `image-3.jpg`, which would store the original in
`image-3_o.jpg`, but this original would never be used by the
handle-image-sizes middleware (it would look for `image_o-3.jpg`). This
fix would solve this weird naming issue, and make it more consistent.
But we need to make sure our middlewares (including handle-image-sizes)
will be able to handle both file locations to remain compatible with the
old format. This isn't additional work, because it would fix the old bug
too.
- Prevent uploading files that end with `_o`, e.g. by automatically
stripping that suffix from uploaded files. To prevent collisions.
Advantage(s):
- We keep the original file name, which is better for SEO.
- No changes required to the storage adapters.
Downside(s):
- The storage of both files will nog happen parallel any longer. But I
expect the performance implications to be minimal.
- Changes to the routing: normalize middleware is removed
no issue
There are a couple of issues with resetting the Ghost instance between
E2E test files:
These issues came to the surface because of new tests written in
https://github.com/TryGhost/Ghost/pull/16117
**1. configUtils.restore does not work correctly**
`config.reset()` is a callback based method. On top of that, it doesn't
really work reliably (https://github.com/indexzero/nconf/issues/93)
What kinda happens, is that you first call `config.reset` but
immediately after you correcty reset the config using the `config.set`
calls afterwards. But since `config.reset` is async, that reset will
happen after all those sets, and the end result is that it isn't reset
correctly.
This mainly caused issues in the new updated images tests, which were
updating the config `imageOptimization.contentImageSizes`, which is a
deeply nested config value. Maybe some references to objects are reused
in nconf that cause this issue?
Wrapping `config.reset()` in a promise does fix the issue.
**2. Adapters cache not reset between tests**
At the start of each test, we set `paths:contentPath` to a nice new
temporary directory. But if a previous test already requests a
localStorage adapter, that adapter would have been created and in the
constructor `paths:contentPath` would have been passed. That same
instance will be reused in the next test run. So it won't read the new
config again. To fix this, we need to reset the adapter instances
between E2E tests.
How was this visible? Test uploads were stored in the actual git
repository, and not in a temporary directory. When writing the new image
upload tests, this also resulted in unreliable test runs because some
image names were already taken (from previous test runs).
**3. Old 2E2 test Ghost server not stopped**
Sometimes we still need access to the frontend test server using
`getAgentsWithFrontend`. But that does start a new Ghost server which is
actually listening for HTTP traffic. This could result in a fatal error
in tests because the port is already in use. The issue is that old E2E
tests also start a HTTP server, but they don't stop the server. When you
used the old `startGhost` util, it would check if a server was already
running and stop it first. The new `getAgentsWithFrontend` now also has
the same functionality to fix that issue.
refs https://github.com/TryGhost/Team/issues/2419
We use a job queue to ensure that webmentions can be processed outside of
the request/response cycle, but still finish executing if the processed is closed.
With this we're able to update the e2e tests to await the processing of the mention
rather than sleepign for arbitrary lengths of time, and we've reintroduced the tests
removed previously
- aa14207b69
- 48e9393159
no issue
When a site doesn't have any emails on boot, it doesn't schedule the email analytics job. With this change, the new email flow will also restart that job after an email has been created.
fixes https://github.com/TryGhost/Team/issues/2484
The flow only send the email to segments that were targeted in the email
content. But if a part of the email is only visible for `status:free`,
that doesn't mean we don't want to send the email to `status:-free`.
This has been corrected in the new email flow.
fixes https://github.com/TryGhost/Team/issues/2484
The flow only send the email to segments that were targeted in the email
content. But if a part of the email is only visible for `status:free`,
that doesn't mean we don't want to send the email to `status:-free`.
This has been corrected in the new email flow.
refs https://github.com/TryGhost/Team/issues/2482
This change adds a small sleep in between dispatching events in the
worker thread that reads the events from Mailgun. That should reduce the
amount of queries we fire parallel to each other and could cause the
connection pool to run out of connections.
It also reduces the amount of concurrent sending to 2 from 10. Also to
make sure the connection pool doesn't run out of connections while
sending emails, and to reduce the chance of new connections falling back
on a (delayed) replicated database.
closes https://github.com/TryGhost/Team/issues/2429
- sends email notifications to staff users when their site receives a Webmention.
- currently behind a flag, that can be toggled in the labs settings.
refs https://github.com/TryGhost/Team/issues/2482
This change adds a small sleep in between dispatching events in the
worker thread that reads the events from Mailgun. That should reduce the
amount of queries we fire parallel to each other and could cause the
connection pool to run out of connections.
It also reduces the amount of concurrent sending to 2 from 10. Also to
make sure the connection pool doesn't run out of connections while
sending emails, and to reduce the chance of new connections falling back
on a (delayed) replicated database.
refs https://github.com/TryGhost/Team/issues/2476
When upgrading from a Complimentary subscription with an expiry, to a paid Subscription of the same Tier, the Member was eventually losing access to the Tier when the complimentary subscription expires as the `expiry_at` on the mapping was not removed. This change fixes the code by setting expiry as null when a member upgrades their subscription to paid. This also adds 2 migrations to fix any side-effects on existing sites -
- Removed invalid expiry tier expiry date for paid members
- Restored missing tier mapping for paid members
refs https://github.com/TryGhost/Team/issues/2476
When upgrading from a Complimentary subscription with an expiry, to a paid Subscription of the same Tier, the Member was eventually losing access to the Tier when the complimentary subscription expires as the `expiry_at` on the mapping was not removed. This change fixes the code by setting expiry as null when a member upgrades their subscription to paid. This also adds 2 migrations to fix any side-effects on existing sites -
- Removed invalid expiry tier expiry date for paid members
- Restored missing tier mapping for paid members
This test is failing because the `sleep` isn't long enough. Removing this test
until we've refactored to use the jobs service, at which point we can remove the
sleep and wait for the job to be complete.
We were incorrectly handling a "no resource found" return value from the
ResourceService, instead of an object with `null` values, we were expecting a
`null` value - so we were considering all URL's to be pointing toward a
resource.
refs https://github.com/TryGhost/Team/issues/2466
Now that we're checking for resources at the URL and rejecting if
there isn't one found, we want to make sure that we can handle pages
which are not a resource.
The idea here is to make a HEAD request to determine whether or not
the page exists. We don't need the full response so HEAD saves us some
bandwidth and we allow both 2xx and 3xx status codes because Ghost has
redirects to add missing trailing slashes, which may not be present in
the URL we're passed.
refs https://github.com/TryGhost/Team/issues/2466
The existing implementation was a very basic check to get us to the
first milestone. By checking if the page points to a resource we can
know for sure the URL exists on the site.
refs https://github.com/TryGhost/Toolbox/issues/503
- The "last_seen" property is not used in routing calculations. Without it the routing service was triggering an expensive process on each user login.
refs https://github.com/TryGhost/Toolbox/issues/503
- There was an error thrown due to empty "model._changed" field
- When attached or detached events (e.g. tag.attached) are sent through, their models do not contain any _changed properties. This was taken into account when checking for route related resource changes
refs https://github.com/TryGhost/Toolbox/issues/503
- The listener was not covered during quick and dirty implementation. While in the area did some cleanup to the sitemap manager test
- One of the problems I've stumbled upon when adding a test is having multiple instances of SiteManager in the test, which in turn created multiple "subscribe" events and repeat handle executions. Fixed it by having just one site manager instance (a singleton) as that's the pattern that used in main codebase
refs https://github.com/TryGhost/Toolbox/issues/503
- The Dynamic URL service no longer generates "url.added" event when only a partial resource update happened - only non-url forming properties were modified. The sitemaps service still needs to know when to update the lastmod ("Last Modified") field associated with specific URL.
refs https://github.com/TryGhost/Toolbox/issues/503
- Tier's are sometimes dynamically generated and are present in the "_changed" properties, causing full URL regeneration. They have no effect on post's URL, so should not trigger URL regeneration.
refs https://github.com/TryGhost/Toolbox/issues/503
- Full URL regeneration process was happening even when only unrelated to URL generation fields were updated (e.g. 'plaintext' change in post does not affect the URL of the post). Stopping the "resource updated" event processing early circumvents full url regeneration inside of DynamicRouting, which can be quite heavy depending on routing configuration
- The URLResourceUpdatedEvent is supposed to be emmited whenever there's an update to the resource already associated with the URL and no url-affecting fields were touched.
