- due to schema changes between versions, we need to completely nuke the
DB between these tests
- this is definitely not the best way to do it but I'll fix properly next week
no issue
`show_title_and_feature_image` leads to more intuitive logic in themes and we can use `posts` rather than `posts_meta` as there are no longer row-length issues with MySQL 8.
- removed original add-column migration that was never in a release
- added new add-column migration that puts
`show_title_and_feature_image` column with a default of `true` on the `posts` table
- renamed property and default value everywhere
- bumped `@tryghost/admin-api-schema` to allow the new property through at the API level
refs @TryGhost/Product#3551
- PostsImporter would convert the HTML from the import file into
Mobiledoc, even if the post was written in Lexical
- As a result, the imported posts would have both mobiledoc & lexical
fields populated, which prevents the post from being updated in the
Lexical editor
- Added a check to see if the post was written in Lexical, and if so,
skip the HTML > Mobiledoc conversion
closes https://github.com/TryGhost/Product/issues/3557
- bumped `@tryghost/admin-api-schema` to allow passthrough of the new property in API requests
- updated output mapper to ensure property always returns a boolean rather than `null` in the case where `posts_meta` doesn't exist for a page
- updated `PostsService.copyPost()` to include the new property when copying
- updated `checkResponse` test util and snapshots to expect `hide_title_and_feature_image` property in page API responses
- fixed pages e2e test so it doesn't inadvertently modify the match object breaking later tests
closes https://github.com/TryGhost/Team/issues/3550
We want to allow an option to hide the title and feature image on a
per-page basis, to do that we need somewhere to store the setting value.
The existing `posts_meta` table is the simplest candidate, especially as
this is a single setting and we don't have a desire to introduce many
such settings.
- added migration that adds the `hide_title_and_feature_image` column to
the `posts_meta` table with a `boolean` data type and a default value of
`false` (matches behaviour of all existing pages)
- updated schema file for initial database creation
- removed property from API output via serializers to keep migration PR
minimal
refs https://github.com/TryGhost/DevOps/issues/39
- up until now, we've had a CI job which does a really basic test for
migrations, but it barely functions and misses bugs all the time
- this commit removes that and switches to an actual test suite for our
migrations, so we can ensure they function as expected
- also removes the env var hack I came up with for those migrations
tests
- this should lead to safer migrations and faster tests
refs https://github.com/TryGhost/Team/issues/3544
- Bumps required packages to fix an issue on posts that still contains base64 images.
<!-- Leave the line below if you'd like GitHub Copilot to generate a
summary from your commit -->
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at 5fb728e</samp>
Updated `@tryghost/helpers` dependency to fix some helper issues. This
improved the accuracy and security of the `{{reading_time}}` and
`{{encode}}` helpers in Ghost themes.
refs https://github.com/TryGhost/Team/issues/3544
- Bumps required packages to fix an issue on posts that still contains base64 images.
<!-- Leave the line below if you'd like GitHub Copilot to generate a
summary from your commit -->
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at 5fb728e</samp>
Updated `@tryghost/helpers` dependency to fix some helper issues. This
improved the accuracy and security of the `{{reading_time}}` and
`{{encode}}` helpers in Ghost themes.
fixes https://github.com/TryGhost/Team/issues/2937
Bumps juice to 9.1.0:
- Support for 'auto' width and height attributes
- Fixed a bug with counter-reset styles
- Dependencies updates
refs https://github.com/TryGhost/Ghost/pull/17065
- This is a follow up cleanup work after introducing even mapper when working on collections refresh mechanism.
- This module manages interception of the Model events and maps/dispatches Domain events that could later be used in different libraries.
fixes https://github.com/TryGhost/Team/issues/3541
The email preheader, which is only present in the html version of an
email, is also included in the plaintext version of all emails. This
results in all text being duplicated twice in plaintext emails.
We setup the relations and NQL expansions/replacements here rather than in the
repository, we want to keep all the bookshelf code together in one place.
When we end up wiring this to the database, this generator will also ensure
uniqueness by appending/incrementing a number on the end of the slug. Long term
it would be good to offload this to a shared slug service, this could also
ensure that slugs are unique globally or between multiple tables, if desired
This follows our existing patterns and protects us from th eslint max line rule
when wiring up dependencies. The rule is designed to keep logic out of glue
code, and this file won't have logic, just lots of deps.
refs https://github.com/TryGhost/Team/issues/3169
- The "current" part of the payload should always have the "id" included to make the data structure easier to use for the event consumer
refs https://github.com/TryGhost/Team/issues/3169
- With the new architecture all possible dependencies should be injected rather than required directly - this applies to the DomainEvents module as well.
refs https://github.com/TryGhost/Team/issues/3169
- When a not yet mapped event is produced by Ghost model the posts belonging to a collection are recalculated in a brute-force way. This causes a lot of strain on the database and times out some of the tests. Right now we only care about a handful of events, so limited the ones we listen to and will be uncommenting the rest when we have an optimal mapping/handling for them.
refs https://github.com/TryGhost/Team/issues/3169
- To be able to apply NQL filtering on Collection Posts the dates should be serialized to be ISO Date Strings instead of raw Dates. Otherwise, NQL filtering fails to compare Date with a Date String.
refs https://github.com/TryGhost/Team/issues/3169
- When handling a single generic event the code becomes riddled with if statements to detect correct "data" that is being passed with the event. Switching to have a domain event per model event helps solving this problem and makes code more readable.
refs https://github.com/TryGhost/Team/issues/3169
- To make the coupling to Ghost's model events as loose as possible added a bridge that maps model events to domain events. These domain events it what the collections module can subscribe to to make necessary updates.
refs https://github.com/TryGhost/Team/issues/3169
- Ghost model events should not be crossing emitted to outside libraries. The wrapper code is a "glue code" that can still listen to model events but transforms them to specific format that is useful for the collections module
- This hooks up the Ghost model events and transforms them into a "CollectionEvent" format.
refs https://github.com/TryGhost/Toolbox/issues/602
- the new regex allows for the filename to contain seconds, which would have
helped prevent a bug we had with migrations being in the wrong order
- mentioned filename ordering in the migration review comment to bring
that to the forefront of the mind
refs https://github.com/TryGhost/Ghost/issues/15725
This pull request adds a new configuration option for the Mailgun email
provider that allows the user to set the maximum number of recipients
per email batch via a new config option `bulkEmail.batchSize`
fixes https://github.com/TryGhost/Ghost/issues/17125
- both collections migrations have the same timestamp so they're ordered
based upon their slug name, and the one that creates
`collections_posts` is ordered before the one that created
`collections`
- as a result, the migration fails because the table does not exist
- this commit pushes the `collections_posts` migration after the other
one by bumping the timestamp
no issue
Keep an eye on Ghost's changelog (https://ghost.org/changelog/) for the
full feature announcement.
<!-- Leave the line below if you'd like GitHub Copilot to generate a
summary from your commit -->
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at ee03e4d</samp>
This pull request removes the `signupCard` feature flag and enables the
signup card for all users in the Lexical editor. This is part of the
Lexical editor beta release, which is a new editor for Ghost that
provides a better writing experience and more features for content
creators.
fixes https://github.com/TryGhost/Ghost/issues/16120
- if you set `sodoSearch.url` to `false` in config, it'll currently
crash because we're not correctly handling the types correctly
- the first part of the fix is ensuring the value is a string so we can
call `.includes` on it
- second, the `false` value is passed into the output as a string, so we
should detect if we passed a false value and early return with an
empty string if so
- credits to https://github.com/ltoinel for the fix inspiration
- in the migration tests we need to boot Ghost and then kill it
afterwards
- because there was no easy way to do this, the workflow waits for 20s
and then kills the last process ID
- aside from being a terrible idea, it means we're also just arbitrarily
waiting for 20s, which burns time when it takes shorter to boot Ghost
- this commit implements an environment variable that will kill the
server once it has run the whole boot process, and then fixes the
workflow to use that
refs https://github.com/TryGhost/Team/issues/3504
- Sentry was never setup and we don't use it
- Styles have been moved to inline JS styles (no separate css file generated)
- App version was never used
- Improved current script tag detection
closes https://github.com/TryGhost/Team/issues/3493
- Fixed pages not saving on force revision. As a side effect, it broke
admin navigation as it doesn't manage to create a new revision upon going back to the pages list.
- This was simply caused by a missing option in the API endpoint config.
---
<!-- Leave the line below if you'd like GitHub Copilot to generate a
summary from your commit -->
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at b646916</samp>
This change enables the `pages` endpoint to handle page revisions by
adding the `save_revision` permission. This is part of a pull request
that adds page versioning and restoring functionality to Ghost.
refs: https://github.com/TryGhost/Toolbox/issues/595
We're rolling out new rules around the node assert library, the first of which is enforcing the use of assert/strict. This means we don't need to use the strict version of methods, as the standard version will work that way by default.
This caught some gotchas in our existing usage of assert where the lack of strict mode had unexpected results:
- Url matching needs to be done on `url.href` see aa58b354a4
- Null and undefined are not the same thing, there were a few cases of this being confused
- Particularly questionable changes in [PostExporter tests](c1a468744b) tracked [here](https://github.com/TryGhost/Team/issues/3505).
- A typo see eaac9c293a
Moving forward, using assert strict should help us to catch unexpected behaviour, particularly around nulls and undefineds during implementation.
refs TryGhost/Ghost#3494
- By default, the post scheduler runs as user_id = 1, which is the
original owner of the site
- If ownership has been transferred to a different user, it's possible
that there is no user with id = 1
- In this case, the scheduler would fail to publish a post, because
updating the post using user_id = 1 failed a foreign key constraint in
the post_revisions table
- This commit fixes the issue by checking if the contextUser exists, and
if not, replacing it with the current owner of the site
closes https://github.com/TryGhost/Team/issues/3499
- bumps `@tryghost/kg-default-nodes` and `@tryghost/kg-lexical-html-renderer` to fix missing `kg-image` class on the `<img>` element of rendered image cards
refs https://github.com/TryGhost/Toolbox/issues/592
- it turns out that `TRUNCATE` in CI takes ~300ms for all tables, but
`DELETE FROM` takes ~30ms
- whilst truncating is generally known to be faster, I believe it's only
faster on large tables
- this saves 90% of the time it takes to reset the DB in MySQL
closes https://github.com/TryGhost/Team/issues/3423
- For convenience we need a way to fetch posts that belong to a certain collection. This change adds support for `collection` query parameter: `/?collection=` which can be either an id or slug of the collections we are trying to fetch.
- When posts are fetched by collection we ignore any filters passed along in query parameters as collection is a "filter" by it's very nature.
no issue
This was a bit of an oversight from our feature built at the retreat. We
didn't take revisions into account for pages at all, but luckily it made
revisions without issues regardless.
It just wasn't accessible and users weren't able to restore via ADMIN
because the API didn't serve them at all.
This wires up the revisions relation to be served by the API so we can
retrieve it in Admin.
We've got some fairly simple diffing logic here to update the collections which
a post is in, the bulk of the changes here are to support the return of a DTO
rather than Bookshelf Model. This also helps improve the architecture because
we are step closer to removing infrastructure concerns (HTTP Response Headers)
from the business logic layer.
For now there is a crappy EventString which can be passed back to the
controller which can then handle any HTTP related concerns, although long term
these should be actual events like PostPublished or PostUpdated.
This prepares us to return a DTO rather than BookshelfModel to the serialiser
layer. When passing a BookshelfModel, the serialisation layer uses the model to
read from when building computed properties. By stripping values out in the
toJSON method it means that the DTO will be missing them and the computed
properties won't be able to be calculated. Instead we return ALL values to the
serialisation layer, and then strip out the ones that weren't requested in the
"clean" step.
This also inadvertently fixes the issue with `reading_time` requiring the
`html` field to be requested, we can now request just `reading_time`, as well
as have it included by default.
refs https://github.com/TryGhost/Team/issues/3423
- When querying for posts that belong to a collection we should be returning full post information just like we do for Posts API.
no issue
- We need to send information about Stripe being enabled or disabled in live mode to analytics
- This hooks up the Domain events listeners in the analytics service and processes this information accordingly
no issue
- Withing the `SettingsBREADService` we want to call the `connect` method of the Stripe Service, which then dispatches the `DomainEvent`
- In order to ensure that the Stripe service is connected in 'live' mode, we compare the public key and make sure no live key was set before so to avoid multiple triggers of this call.
refs https://github.com/TryGhost/Toolbox/issues/592
- this commit extracts the regression tests into a separate workflow
- this means they run in parallel and reduce the time we have to wait
for DB tests in general
- also fixes a test that was reliant on being run after the E2E
tests (!)
refs https://github.com/TryGhost/Toolbox/issues/592
- async-await makes the code easier to read
- also performs a small optimization to only load the foreign_keys
pragma once for SQLite
refs https://github.com/TryGhost/Toolbox/issues/592
- we should reset the URL service to avoid event listeners piling up and
slowing down CI due to the number of events it has to process
refs https://github.com/TryGhost/Toolbox/issues/592
- heads up, I'm not really sure about this fix
- when we're wrapping `setTimeout`, time stops and mysql2 starts doing
weird things because we then shift time and it hits timeouts
- apparently `shouldAdvanceTime` should fix this by automatically
incrementing time along with the system clock
- given the problem is quite difficult to hit, I could just be seeing a
lack of this due to some other factor
- also removed unnecessary sinon sandbox creation as this is superfluous
closes https://github.com/TryGhost/Team/issues/3425
- Index collection is needed to support one of the usecases we have in the near future where we'd hold all posts that would be displayed on the "index" page.
closes https://github.com/TryGhost/Team/issues/3431
- We don't currently have a clear usecase to use the new pattern of updating posts as nested resource (of a collection). To simplify the API we are sticking with the approach of controlling where the post belongs to only through the Posts Admin API.
refs https://github.com/TryGhost/Team/issues/3274
Co-authored-by: Sanne de Vries <sannedv@protonmail.com>
Co-authored-by: Ronald Langeveld <hi@ronaldlangeveld.com>
Co-authored-by: Jono Mingard <reason.koan@gmail.com>
Co-authored-by: Sag <guptazy@gmail.com>
refs https://github.com/TryGhost/Toolbox/issues/592
- we don't need to receive the refreshed model back afterwards so we can
save an SQL select per insert by disabling auto-refresh
- this saves about 18000 DB queries for the MySQL tests, which should
help reduce test time due to the overhead
no issue
- The class should not rely on being passed a specific dependency, but rather needs to communicate with types what structure and method it needs to function correctly.
- Replaced the specific dependency to `sentry` with a generic definition of what is expected.
no issue
- In order to listen to `DomainEvents` for `MilestoneCreatedEvents` we need to add a `DomainEvents` listener and handler to the Segment analytics service.
- For better readability and to be more consistent with how code is currently written in Ghost, I refactored the service index file and split the two types of event listener into separate classes which is much cleaner and easier to test.
Regression tests are not written to ensure coverage of code, in fact they are
barely written at all anymore, instead we write unit and e2e tests. Because of
this the coverage is constantly dropping as the codebase grows. This causes
significant pain and suffering for developers and slows down development.
closes https://github.com/TryGhost/Team/issues/3325
Awaited `DomainEvents.allSettled()` to ensure domain event is fully
processed before asserting member was successfully updated
refs https://github.com/TryGhost/Team/issues/3145
Updates pintura integration to be switched on by default for all sites by adding a migration to update the default value for the setting.
refs https://github.com/TryGhost/Team/issues/3376
fixes b4a97d084f
- The in-memory stores are not cleaned up when the Ghost instance is "shallow restarted" between test suite runs, causing the initialization of built-in collections to run multiple times. The initialization should ever add the collections once.
refs https://github.com/TryGhost/Team/issues/3376
- When the Ghost instance is initialized it has to have a set of built-in collections. With these changes Ghost starts with a "featured posts" collection - available to be used right away.
closes https://github.com/TryGhost/Team/issues/3170
- When resources that are related to automatic collection filter are updated the posts in collections should be updated as well.
- This change adds a super-basic way to track changes in post/tag/author resources and updated all automatic collections when any of those resources change. In the future we can optimize the update process to be more performant, but it's good enough for current needs
refs https://github.com/TryGhost/Team/issues/3148
- Collections should never ever operate on pages, so limiting them at the fetch stage is the bes possible place, so they don't enter the system when an automatic collection is created.
refs https://github.com/TryGhost/Team/issues/3170
- This implementation allows to create an automatic collection with a filter defining automatically populated posts that belong to a collection
- To populate collection using a filter the API client can send a `filter` property along with a collection request
- Filter values are compatible with the filters used in Content API (https://ghost.org/docs/content-api/#filter)
refs https://github.com/TryGhost/Team/issues/3269
- fixes continuation of list sequence when a non-list-node separates list nodes bringing rendered output in line with editor depiction
refs https://github.com/TryGhost/Team/issues/3337
Moved current email customization functionality that is behind the
`makeItRain` to its own flag (`emailCustomization`) and removed the now
redundant `makeItRain` flag
fixes https://github.com/TryGhost/Team/issues/3331
This adds attribution tracking to the signup form. It sends a newly
created url history when sending the signup API call, this url history
will get translated to a proper attribution and saved on the backend. We
send a history with only a single item that contains the referrer
source, medium and path of the Embed form.
This also makes some changes to the E2E tests so that the tests run
in an https environment instead of about:blank.
refs https://github.com/TryGhost/Team/issues/3260
- We need a way to remove posts form collections without fetching the whole collection's content. This API method allows to remove posts from manual collections by collection id and post id.
- As a response it returns up to date collection state without the removed post.
refs https://github.com/TryGhost/Team/issues/3260
- We need a way to append posts to collections without sending over all of the posts that are already in the collection
- The API would receive post_id and collection_id as required fields and will optionally take in sort_order to control the ordering in the manual collection
no issue
- Removed the postHistory flag from labs
- Post History will be saved and displayed for all lexical posts,
regardless of whether the lexicalEditor flag is currently set
- Post History will still not be displayed for any mobiledoc posts
- With this change, the logic is simplified as we don't have to worry
about flags, but only the content in the given post (mobiledoc vs
lexical)
- If someone toggles the lexicalEditor flag on, creates a new post, then
toggles the lexicalEditor off, we still want Post History to work for
the existing lexical post
refs https://github.com/TryGhost/Team/issues/3329
- product card images have explicit width and height attributes in the markup now, and the image needs auto height for that reason
- without it, the images will lose its original aspect ratio for themes without the auto height images
closes https://github.com/TryGhost/Team/issues/3324
- When the recipients batch size is larger than the limit in addition to logging the error we need extra data to figure out what exactly is inside those `2000` or `3000` records causing faulty behavior.
- This change grabs all available models and dumps them into a file inside of the `content/data` folder. The code is temporary and should be removed once the problem is narrowed down
refs https://github.com/TryGhost/Team/issues/2808
Updated the test to ensure that the date assertions do not unexpectedly
fail if the dates used are computed precisely at the start of a second
(no milliseconds `.000Z`)
refs https://github.com/TryGhost/Team/issues/3234
Added an e2e for the update check script to detect potential breakages
in the script due to uninitialised dependencies in the isolated
execution environment
fixes https://github.com/TryGhost/Team/issues/3296
Adds a new `signupForm` feature flag, that will enable/disable the new embeddable signup form code generation.
Since the new flag shares its name with a new config value (that contains the script location), this also fixes the feature helper to only use a config with the same name if it is a boolean.
closes https://github.com/TryGhost/Team/issues/3259
- API output mappers (soon to be serializers) are meant to work based on allowlist set of output properties. Having the allowlist early on will allow to track the API evolution consistently.
- this will allow us to see which set of tests are consuming the most
amount of time in CI
- in order to split apart the commands, I've had to override the
coverage thresholds for integration+regression tests in order to keep
c8 happy
- also sprinkled some more labels into the workflows to make things
clearer to read
refs https://github.com/TryGhost/Team/issues/3248
The current test fixtures didn't include any hidden Tiers, so I've added
a new fixture to test the filtering of hidden Tiers. It's not enabled by
default to avoid breaking the existing tests.
refs https://github.com/TryGhost/Team/issues/3167
- This is scaffolding for collections API. Contains wiring for service wrapper, e2e test, and a browse endpoint
- Adds basic implementation of the GET /collections endpoint to build up upon
- Note, there are no permissions in this version as they will be added in later stages of development with migrations etc
refs https://github.com/TryGhost/Team/issues/3234
The update check was failing to run due to recent changes in the tiers
service. This service now needs initialising before the update check can
be run.
refs https://github.com/TryGhost/Team/issues/3248
- API queries for tiers are now using the TiersRepository with internal caching
- the repository had a bug with it's `toPrimitive()` method which meant the cached tier objects had very few properties
- the Tier object has all properties as private getters except for standard `events` property which meant the spread operator didn't have anything to spread into the object resulting in all tiers having a shape like `{events: [], active: true, type: 'paid', id: 'abcd'}`
- the `getAll()` method uses nql to match against the cached tier objects but with them not being fully populated it wasn't able to match and so returned an empty array
---
- changing the spread to use `tier.toJSON()` means we're populating all of the tier data properly allowing filter matches to work
refs https://github.com/TryGhost/Team/issues/3224
When a product has a slug that is a single letter, checking if a user
had access to view a post associated with that product would cause a 500
error. The underlying cause of this issue is
https://github.com/TryGhost/NQL/issues/20 This fix circumvents this
issue by providing a value that the nql lexer will not error out on
refs https://github.com/TryGhost/Team/issues/3151
- adds route in Admin for opening new settings screen in react
- adds new config for passing `adminX` to Admin
- loads adminX settings from package in Admin UI when new route is opened
Co-authored-by: Peter Zimon <zimo@ghost.org>
refs: https://github.com/TryGhost/Team/issues/3139https://github.com/TryGhost/Team/issues/3140
- Added duplicate post functionality to post list context menu
- Currently only a single post can be duplicated at a time
- Currently only enabled via the `Making it rain` flag
- Added admin API endpoint to copy a post - `POST ghost/api/admin/posts/<post_id>/copy/`
- Added admin API endpoint to copy a page - `POST ghost/api/admin/pages/<page_id>/copy/`
refs https://github.com/TryGhost/Toolbox/issues/515
Tiers are very frequently queried and we want to reduce the number of DB calls
we're making. We can store the Tiers in-memory, using the existing in-memory
repository patterns, but still persisting writes the the database.
We also have to update our test helpers, because they were bypassing the
repository for writes, but using it for reads resulting in an invalid cache
We want to cache access to Tiers, and it's easier to do that in the
TierRepository. So we update a heavy user of Tiers to use the Tier
service so it can take adv of caching. The serializers are a big
offender for making calls to fetch Tiers.
- this dependency seems a pretty heavy one to require upon boot and
given most sites don't need it to function as normal, this saves
several MB of RAM per instance
closes https://github.com/TryGhost/Team/issues/3176
We were waiting for the welcome email to send before responding to the
client that setup is complete, this was causing the client to hang when
running `ghost install local` as mail isn't configured by default.
no issue
- the lexical lib file makes use of `jsdom` but there was no explicit dependency for it in `package.json` meaning we were relying on it being incidentally depended on through another package which is brittle
These versions use the latest version of @tryghost/errors, which uses
the correct import for @stdlib/utils-copy. This should hopefully stop
missing module errors when running locally.
We were incorrectly specifying the path of the file rather than relying on the
package exports, and this broke in the previous commit to a file rename. It
wasn't caught by CI because the browser tests are not run on pull requests
As discussed with the product team we want to enforce kebab-case file names for
all files, with the exception of files which export a single class, in which
case they should be PascalCase and reflect the class which they export.
This will help find classes faster, and should push better naming for them too.
Some files and packages have been excluded from this linting, specifically when
a library or framework depends on the naming of a file for the functionality
e.g. Ember, knex-migrator, adapter-manager
no issue
This pull request adds a new experimental feature flag `signupCard` to
the Ghost admin app, which enables a signup card component in the
lexical editor that's currently being worked on.
The feature flag can be toggled from the settings/labs
UI and is read from the server-side configuration.
Whilst Admin API Integrations had the permissions to create invites they were
blocked from doing so at the HTTP level. We've removed this restriction for
creating Invites as well as browsing Roles, because a Role ID is necessary to
create an invite. The code was also not setup to support Admin API Integrations
as it made assumptions about the existence of a User. That has been updated in
the permissions layer - so that the Invites are limited to Contributors,
Authors and Editors as well as at the email layer, which has has the copy and
from address updated to reflect the lack of a User creating the Invite.
refs 9d104c8511
- we've seen recurring instances where Ghost will hog memory after image
uploads
- we use `jemalloc` to try and help this, but it still seems to happen
- according to the sharp thread referenced in my commit above, memory
fragmentation can also be helped by reducing the concurrency within
sharp
- this is a bit of an experiment and we can revert if it causes issues
