The `release` method is meant to clean up a sticky
element before removal from the store.
The part of this that cleared CSS added due to
being in dialog mode was only run if still in that
mode.
In the template folder JS, we set the mode before
running `recalculate` so elements can be cleaned
in a different mode to that they were last in.
This meant some CSS wasn't being cleaned up.
This sometimes caused an effect where elements
that were added back to the DOM but not stuck
would still have CSS used when they were last
stuck.
This commit also includes the addition of some
missing code that adds back CSS to offset an
element when in a dialog stack.
Moving focus before letting the sticky JS set the
position of sticky elements means focus is shifted
to them when they are still in the page. This
causes the browser to scroll so the focused
elements are visible.
This moves when focus is set to after the sticky
JS has set position therefore avoiding the issue.
Makes sure the tabindex we add for focusing is
removed.
Also removes the outer fieldset from the radios
for new templates. We don't wrap form buttons in
fieldsets anywhere else and it doesn't add any
useful semantics to the form.
Adds some extra styles so <fieldset>'s show as
focused when they are.
Also includes replacement of `overflow: hidden` on
sticky elements with a clearfix. hiding overflow
clips the outline and the clearfix can be used for
containing any floats instead. (I'm assuming that's
why it was set here.)
Bigger hit areas are generally better (cf Fitt’s law[1]), as long as
they’re not ambiguous.
This commit enlarges the hit area of the edit links (using a border) so
they fill the same vertical space as the smallest possible row (going
to the full height might look weird because some of these rows get very
tall).
Bigger hit areas are generally better (cf Fitt’s law[1]), as long as
they’re not ambiguous.
This commit enlarges the hit area of the edit links (using a border) so
they fill the same vertical space as the row they sit in.
We were removing the border to ‘unfold’ the corner of the page. This was
causing the size of the element to shrink.
This meant that it you hovered the bottom 1px of the element it would
cycle in and out of the hover state as fast as the browser could render
it.
This commit fixes that by making the border transparent, rather than
removing it.
Sometimes people print stuff under where we’re folding the letter. It’s
annoying to not be able to see it.
This commit adds a little detail where, once you’ve sent the letter
you can unfolds the corner to see what’s underneath.
It’s better that we do this for all letters for discoverability.
This is a platform admin page which lists all letter brands and their
domain. Once the page to edit a letter brand has been created, each
brand will link to its edit page, but for now this is just a list of
brand names and domains.
it wouldn't show search if there were under a certain amount of letter
or email branding options - however we know there will always be more
than that amount so lets remove some complexity.
Also, rename the SearchTemplatesForm because it can search anything -
it just prompts you to search by name is all.
new code is copied stylistically from the email branding patterns.
Instead of `service.dvla_organisation`, there's now
`service.letter_branding` and `service.letter_branding_id`. However,
unlike email branding we're not currently showing a preview of the
logo. That can come later when we work out how we want to do it.
To match the call to action on the template page.
Karl said:
> I feel good about it. You're not really editing it so much as choosing between 2 options. It feels more human.
>
> I think it only started out as edit to match all the other sections of the letter – it certainly wasn't given any real consideration.
Form errors were being shown (such as a domain not being a valid), but
we weren't showing nicely formatted error messages if the database
failed to save a new row.
This has a form with 3 fields - the file upload field, logo name, and an
optional logo domain. Logos need to be uploaded in `.svg` format and we
then convert this to `.png` format and upload both file types to S3 as
well as saving the letter branding details in the database.
Some of the tests were calling methods that didn't exist on MagicMock
objects, which meant that they were always passing. For example, tests
using the format of MagicMock.called_once_with(...) were always passing
since a MagicMock does not have the method `called_once_with`. Changing
this to `assert_called_once_with` ensures that the tests fail if they're
called with the wrong arguments.
