* FIX: Fix race condition when resolving tag and category hashtags
If the category hashtags were resolved first and then tag hashtags, then
the tags would overwrite the categories. Similarly, if the category
hashtags were resolved last it would overwrite even hashtags which ended
with '::tag'.
* DEV: Add test
* DEV: Fix test
* FIX: Improve category hashtag lookup
This commit improves support for sub-sub-categories and does not include
the ID of the category in the slug, which fixes the composer preview.
* FIX: Sub-sub-categories can be mentioned using only two levels
* FIX: Remove support for three-level hashtags
* DEV: Simplify code
Adds a new rake task `plugin:checkout_compatible_all` and
`plugin:checkout_compatible[plugin-name]` that check out compatible plugin
versions.
Supports a .discourse-compatibility file in the root of plugins and themes that
list out a plugin's compatibility with certain discourse versions:
eg: .discourse-compatibility
```
2.5.0.beta6: some-git-hash
2.4.4.beta4: some-git-tag
2.2.0: git-reference
```
This ensures older Discourse installs are able to find and install older
versions of plugins without intervention, through the manifest only.
It iterates through the versions in descending order. If the current Discourse
version matches an item in the manifest, it checks out the listed plugin target.
If the Discourse version is greater than an item in the manifest, it checks out
the next highest version listed in the manifest.
If no versions match, it makes no change.
If any value, including nil, is passed in as an argument the default
won't be set, so we need to handle when a non-Array value is passed in
to the `generate_thumbnails!` method.
This is a very expensive process, and it should only be required in exceptional circumstances. It is possible to run a similar recovery using `rake uploads:recover` (5284d41a8e/lib/upload_recovery.rb (L135-L184))
Because of how the dropdown was structured, as long it was in the DOM, all clicks outside the widget would rerender it.
This commit introduces `widget-dropdown-body` that handles the `clickOutside` callback and is rendered conditionally, so it won't get called when the dropdown is closed.
* FIX: CookText may be gone before promise resolves
Bug introduced in 293467a37ae33c5fd8fd40bbfc05bffa578b90ed.
* DEV: Drop the window.requireModule in cook-text
It was introduced in 2017 in 232311aa8c8e879c98dc061ac1755ee4d0dd9b0f but doesn't seem to be necessary anymore.
`OptimizedImage#filesize` calls `Discourse.store.download` with an OptimizedImage as an argument. It would in turn attempt to call `#original_filename` and `#secure?` on that object. Both would fail as these methods do not exist on OptimizedImage, only on Upload. We didn't know about these issues because:
1. `#calculate_filesize` is not called often, because the filesize is saved on OptimizedImage creation, so it's used mostly for manual filesize recalculation
2. we were using `rescue nil` which swallows all errors
Newer versions of Normalize remove the `border-collapse: collapse;` property and fallback to browser defaults. This commit restores that property because we're using it in quite a few places.
Previously, basic-topic-list had its own implementation of topic-list-item on mobile, which made it more difficult to maintain and extend. The visible difference was that the basic-topic-list implementation had no large avatar on the left. This commit adds a new hideMobileAvatar parameter to topic-list-item and topic-list, and sets it to `true` for the basic-topic-list component.
`Nokogiri::HTML.fragment` is a huge hack (a comment in the source code
admits this). The current behavior of `Email::Styles` is to try to
emulate `fragment` using nokogumbo, but it misses some edge cases. In
particular, meta tags in a email template don't make it through to the
final email.
Instead of treating the provided HTML as an indeterminate fragment, this
commit makes `Email::Styles` treat the HTML as a complete document. This
means that the generated HTML for an email will now always contain top
level structure (a doctype, html, head and body tags).
This new behavior is behind a hidden site setting for now and defaults
off.
Previously we would include this section, unfortunately
1. It is usually elided in gmail
2. It can make the emails longer and more confusing
3. Omission is a feature, it means people need to visit site to get context
It is a second attempt to this update. First one was reverted here https://github.com/discourse/discourse/pull/8618
We noticed a problem that `like` counter had incorrect colour on mobile.
I added a missing rule to the bottom of the file (that rule existed in normalize-3 and was removed in normalize-8)
```
button,
input,
optgroup,
select,
textarea {
color: inherit;
}
```
Previously, while generating the topic page's canoncial url we used the current post number. It will create invalid canonical path if the topic has whsiper posts. Now we only taking the visible posts for current page index calculation.
The previous fix (f43c0a5d857d34) wasn't working for images that were already uploaded.
The "metadata" (eg. 'for_*' and 'secure' attributes) were not added to existing uploads.
Also used 'Upload.get_from_url' is the admin/site_setting controller to properly retrieve
an upload from its URL.
Fixed the Upload::URL_REGEX to use the \h (hexadecimal) for the SHA
Follow-up-to: f43c0a5d857d34
When uploading an image as a site setting, we need to return the "raw" URL, otherwise
when saving the site setting, the upload won't be looked up properly.
Follow-up-to: f11363d446
08044b4f regressed emoji auto complete logic since we (I) forgot to add the
space into the not capturing group at the beginning.
This meant that
`hello 👍t` would not trigger an autocomplete to pick skin tone.
* Change S3Helper::DOWNLOAD_URL_EXPIRES_AFTER_SECONDS to 5 minutes, which controls presigned URL expiry and secure-media route cache time.
* This is done because of the composer preview refreshing while typing causes a lot of requests sent to our server because of the short URL expiry. If this ends up being not enough we can always increase the time or explore other avenues (e.g. GitHub has a 7 day validity for secure URLs)
Previously we were using `$elem.find(...).not($elem.find(...))`. This took approximately 2ms on my machine with a test post.
This commit switches to using a native querySelectorAll method, which takes less than 0.5ms on the same test post.
