As mentioned on #1879, the dialog.deleteLater() call appears to be
causing crashes on some systems on Qt 6.3. Explicitly destroying the page
prior to the containing webview appears to work around the issue.
Issue was introduced in 7922f18296
when the styling was moved into base.scss. From what I can tell, there
are two locations where the current styling is required:
- the editor, to ensure the tags are shown at the bottom
- the change notetype screen, which requires this styling for a sticky
bar at top
I'm no CSS guru, so if someone thinks this could be solved in a better
way, please submit a follow-up PR.
Closes#1782
Re: https://forums.ankiweb.net/t/anki-2-1-53-release-candidate/20122/2
Autoloads in MathJax are asynchronous, and the caller is expected to
use asynchronous APIs when they are in use [1]. The editor uses the synchronous
tex2svg(), which throws a "MathJax retry" error when an autoload package
has not yet loaded. Attempting to use the package before it has loaded
appears to break future invocations as well, so the package fails to work
at all until a new webview is created.
The following HTML will reproduce the issue when added to a single card
in a new profile:
```
<strong>6 </strong>Every combination of <anki-mathjax>\boldsymbol{v}=(1,-2,1)</anki-mathjax> and <anki-mathjax>\boldsymbol{w}=(0,1,-1)</anki-mathjax> has components that add to _____.<br>
```
Ideally we'd switch the MathJax rendering to be asynchronous, but that
didn't work well when I tried it in #1862. For now I've just switched
to the full package, which adds about 130KB to the final minified JS
(2.76MB), and likely slows down editor loading somewhat.
[1] https://github.com/mathjax/MathJax/issues/2557#issuecomment-727655089
In v3, it's more informative to show the count of child decks separately,
since increasing the limit of the current deck does not increase the limits
of child decks. When we rework the decks list in the future, a tooltip
will hopefully provide an easier way for users to see where cards are
available, and where limits are being applied.
Closes#1868
Probably a race condition. `.get_row()` is not a reliable check for card
or note existence, because it returns from an outdated cache,
if the backend is currently blocked.
* Remove custom config and use tex-chtml-full for editor Mathjax
* Add mathjax package in /ts
* Share mathjax config between tex-svg and tex-chtml
* Use "[+]" in Mathjax config again
* Remove mention of MathJaxReady
* Satisfy eslint
* Move focus into input field, when input is shown
* Change trapFocusOut to move focus into available inputs
- This means that e.g. closing the HTML editor with focus in it will
focus the visual editor in turn
* Prevent Control+A unselecting tag editor when no tags exist
* Add ResizableImage.svelte in ts/editable
* Set image constrained via attributes instead of managed style sheet
* Implement new constrained size method
* Turn WithImageConstrained.svelte into normal ts file
* Introduce removeStyleProperties
Removes "style" attribute if all style properties were cleared
* Avoid --editor-width and use one variable set on container
* Disable shrinking if already smaller than shrunken size
* Add button to restore image to original size
* Don't allow restoring original size if no custom width set
* Bottom-center HandleLabel
* Satisfy svelte-check
* Skip unparsable pylib files with black
This was causing black to return a non-zero code, preventing the
subsequent isort from running.
* Organise imports on save
The old `media_files_did_export` hook has been kept around for use with
the legacy apkg exporter (an add-on uses it), and a new
`legacy_export_progress` hook has been added so we can get progress
from the new colpkg exporter until we move over fully to the new code.
