We're calling this method after setting the map location with
`map_location = MapLocation.new if map_location.nil?`, so the condition
`map_location.present?` is always going to be true.
The only view that linked to this action was never used and so it was
deleted in commit 0bacd5baf.
Since now the proposals controller is the only one place rendering the
`shared/map` partial, we're moving it to the proposals views.
It was possible to remove a map location from a different proposal (even
one created by a different author) by modifying the hidden `id`
parameter in the form.
So we're making sure the map location we destroy is the one associated
to the proposal we're updating.
Since we're now using the `@proposal` instance variable in the
`destroy_map_location_association` method, we're calling that method
after loading the resource with cancancan.
We were rendering the whole sidebar again, which wasn't necessary since
most of it remains unchanged. This resulted in complicated code to pass
the necessary information to render the same map that was already
rendered. Furthermore, when users had been moving the map around or
zooming out, we were resetting the map to its default settings after
voting, which was potentially annoying.
This also fixes the wrong investments being displayed in the map after
voting; only the investments on the current page were displayed in this
case while the index displayed all of them.
This is the only part of the sidebar that needs to be re-rendered after
an AJAX request adding or removing investments to a ballot, so having a
separate view just for it will make it easier to simplify the code.
Since this test was added in commit 78c6a30424 the
"current_password" was filled with an incorrect value.
This test did not fail because in previous versions of the
"devise-security" gem if this "current_password" was not valid
a "self.valid?" was performed which caused the error we are testing.
In version 0.17 (as can be seen in the following commit
https://github.com/devise-security/devise-security/pull/340/commits/41a99b67fe0)
if the "current_password" is not valid, related errors are still added
but the "self.valid?" code is removed, which is what causes the
expected error "must be different than the current password"
to appear in the "new password" field.
By adding a correct "current_password" in the test, we avoid this validation
for the "current_password" (which no longer includes the expected error)
and follow the natural flow of devise that does end up intercepting the error
for the new password entered.
Newer versions have more capabilities and also fix lot of errors,
usability and accessibility issues.
Note that we're using an external gem to rails-assets.org as it's
not supports the rails-assets pipeline. This gems wraps the
original one and make the code work with the default rails assets
pipeline as usual.
We were already using it because Rails adds the `Hash#except` method for
Ruby 2.7 and earlier, but since the method wasn't part of Ruby itself,
the Rubocop rule only works with Ruby 3.0 and later.
These tests were failing with Ruby 3.0 because we were getting an error
when loading the `translations` association of the dummy class:
```
NameError:
uninitialized constant
DummyBanner::#<Class:0x000055630e4dccd8>::Translation
```
Specifying the `class_name` of the `translations` association solves the
issue.
It's causing annoying behaviour for desktop users when scrolling
the page to the bottom and there is more content below the
map.
The behaviour of touchable devices does not seem to be
affected by this change and keeps behaving the same.
We were using `server` on staging but `server1` and `server2` on
preproduction and production.
The reason behind it is we've always used one server on staging but
sometimes we've used several servers on preproduction and production.
However, this is a bit of a mess on installations which have only one
server on preproduction or production and need to use the `server` key
for the staging environments but `server1` for other environments.
So, in order to keep compatibility with existing Consul installations,
we're now allowing either `server` or `server1` on any environment.
It doesn't make much sense that by default we use one server on
production on two servers on preproduction.
Note we're keeping `server1` instead of using just `server` in order to
keep compatibility with existing installation.
I thought SelfClosingTag was enabled since it was on the list and the
`enabled: false` got lost among all the `enabled: true`. ERB Lint 0.1
and later will also add RequireInputAutocomplete as a default linter,
and we're not interested in it, at least for now.
So, just like we do for Rubocop, we're disabling all linters and
enabling the ones we use explicitly.
It has been detected that for the :pt-BR, :zh-CN and :zh-TW locales,
the translate button was being displayed, but when requesting the
translation, the remote translation validation failed due to:
'''
validates :locale, inclusion: { in: ->(_) {
RemoteTranslations::Microsoft::AvailableLocales.available_locales }}
'''
That available_locales method did not contemplate these 3 languages
in the format used by the application.
To solve this problem the api response is mapped to return all
locales in the format expected by the application.
Add remote translation model test to ensure that a remote translation
is valid when its locale is pt-BR.
Co-Authored-By: Javi Martín <35156+javierm@users.noreply.github.com>
TranslatorText isn't compatible with Ruby 3, so we need to use a
different gem.
The 'translator-text' gem response was an array of one or more objects.
Now with the 'bing_translator' gem the response is an array with one or
several translated texts.
We remove the concept of object in the code. And we also remove the
"create_response" method from the specs since it is no longer necessary
to emulate that object and we can simply use arrays with texts to emulate
the new response.
Since we changed the way we integrate coveralls in commit 8ed8cc8b9,
we're getting 6 additional checks displayed in our pull requests.
We don't need these checks, and they only add noise. The only reason we
use coveralls is to know the test coverage in our master branch.
So we're changing the code so coveralls only runs on the master branch.
There's also a chance that the test suite will be faster because it
doesn't need to keep track of the coverage, although I haven't noticed any
significant differences during my tests.
I haven't found a more elegant way to say that a certain step should
only be run on push on master, so I'm setting the environment variable
we were already using.
We were getting many errors when trying to run them, from uninitialized
constant `HTTP` to undefined method `headers`.
We might move these examples to the documentation repository in the
future, but we need to look for possible side-effects first.
