We were duplicating the asset host and the URL host in all environments,
but we can make it so the asset host uses the URL host unless we
specifically set it.
Note that, inside the `ApplicationMailer`, the `root_url` method already
uses `default_url_options` to generate the URL.
In the rare case of CONSUL installations who have changed the asset
host, this change has no effect since they'll get a conflict in the
environment files when upgrading and they'll choose to use their own
asset host.
Until now, when editing a specific environment, other CONSUL
installations had to edit the original file, which made it harder to
upgrade.
Now it's possible to change the default CONSUL settings using custom
files, making it easier to upgrade to versions of CONSUL which change
the original environment files (which is very common when upgrading
versions of Rails).
Before this change log generated by OmniAuth was sent to the STDOUT. Now
developers will find those gems log within the `app/log` directory as expected.
In the dev seeds, we were using `Setting["url"]/proposals`, but we can
use `proposals_url` instead, similar to what we do in the rest of the
application.
We can do a similar thing in the sitemap. This way the sitemap will also
work on installations who haven't manually set the "url" setting.
Since we aren't using `Setting["url"]` anywhere anymore, we're removing
it.
This setting was mainly redundant, since we already had the
`server_name` in the secrets. Furthermore, `server_name` is
automatically configured when running the installer, while
`Setting["url"]` had to be manually set in the admin section the
application was installed.
Note we're using `ActionMailer::Base` setting to generate URLs. Sounds a
bit strange, but it's a standard way Rails provides to generate URLs
outside the context of a request.
We've been using the `url` Setting for a long time, but since then we've
added a few references to `root_url` to this file, so we're now adding
consistency. We're also removing a now unnecessary condition.
We were using `Setting["url"]` to verify the content belonged to the
application URL, but we can use `root_url` instead.
Note that means we need to include the port when filling in forms in the
tests, since in tests URL helpers like `polymorphic_url` don't include
the port, but a port is automatically added when actually making the
request.
This task was added in commit 2b9b78e38. According to the notes in the
pull request introducing the change:
> If you have had [map validations] in production, you might want to run
> the rake task votes:reset_vote_counter to reset the cached votes
> counter. Otherwise, until there is another vote for a proposal, the
> counter will not be updated, and there might be some votes which are
> not yet displayed
In short, this was a task that was introduced for a specific release and
had to be run manually. So we can remove it now.
This task was "temporarily" removed in commit 7b6619528. Since that was
done three and a half years ago, right after the dashboard was
introduced, I think it's time to make this "temporary" measure a bit
more permanent ;).
By using the Rails `button_to` helper (which generates a form), and adapting the
response to `html` and `js` formats, the feature works with or without javascript
enabled.
We were rendering the `new` action, but that action doesn't exist.
Before commit ec861ca8e, we were rendering the `edit` action of an
answer, which was confusing as well.
Note that, when adding an invalid document, `@answer.documents` contains
that invalid document (which is not present in the database). Since
we're rendering the index, this new document would appear in the list of
the documents that can be deleted; to avoid that, we're kind of
"reloading" the answer object in the component by finding the record in
the database. We aren't using `@answer.reload` because doing so would
remove the validation errors.
We were using the same logic in four different places, so we're creating
a new class to handle that logic.
Note that I didn't find a way to delegate the `content` method to a
`Admin::TableActionsComponent`, so we're delegating the `action` method
instead. That means we need to create a method returning an
`Admin::TableActionsComponent`. We also need to cache this object;
otherwise we were getting an error when calling `actions.action` from
the `Admin::Poll::Questions::TableActionsComponent`.
Note that the `create` action doesn't create an image but updates an
answer instead. We're removing the references to `:create` in the
abilities since it isn't used.
In the future we might change the form to add an image to an answer
because it's been broken for ages since it shows all the attached
images.
Adding, modifiying, and/or deleting questions for an already started
poll is far away from being democratic and can lead to unwanted side
effects like missing votes in the results or stats.
So, from now on, only modifiying questions will be possible only if
the poll has not started yet.
In this form, the only case where `poll` might be present without
`question.poll` being present to is going to be the `new` action. We can
assign the poll in the `new` action and get rid of the `poll` variable
in the form.
We need to update a couple of tests because a poll is created in the
tests with a timestamp that includes nanoseconds and in the form to edit
the time of the poll the nanoseconds are not sent, meaning it was
detected as a change.
Instead of having to add `beginning_of_minute` to deal with an issue
with Capybara filling datetime fields as mentioned in commit 5a0fde4048,
we can travel to the beginning of the minute so we don't have to take
the seconds into account.
The extra check to see the voter count has increased was redundant; we
already check the request has finished inside the
`vote_for_poll_via_web` method and we check all three voters are created
in the results table.
Updating the poll so it's in the past after starting the browser might
result in database inconsistencies while running the tests, so we're
using `travel_to` instead.
We were already saving it as a time, but we didn't offer an interface to
select the time due to lack of decent browser support for this field
back when this feature was added.
However, nowadays all major browsers support this field type and, at the
time of writing, at least 86.5% of the browsers support it [1]. This
percentage could be much higher, since support in 11.25% of the browsers
is unknown.
Note we still need to support the case where this field isn't supported,
and so we offer a fallback and on the server side we don't assume we're
always getting a time. We're doing a strange hack so we set the field
type to text before changing its value; otherwise old Firefox browsers
crashed.
Also note that, until now, we were storing end dates in the database as
a date with 00:00 as its time, but we were considering the poll to be
open until 23:59 that day. So, in order to keep backwards compatibility,
we're adding a task to update the dates of existing polls so we get the
same behavior we had until now.
This also means budget polls are now created so they end at the
beginning of the day when the balloting phase ends. This is consistent
with the dates we display in the budget phases table.
Finally, there's one test where we're using `beginning_of_minute` when
creating a poll. That's because Chrome provides an interface to enter a
time in a `%H:%M` format when the "seconds" value of the provided time
is zero. However, when the "seconds" value isn't zero, Chrome provides
an interface to enter a time in a `%H:%M:%S` format. Since Capybara
doesn't enter the seconds when using `fill_in` with a time, the test
failed when Capybara tried to enter a time in the `%H:%M` format when
Chrome expected a time in the `%H:%M:%S` format.
To solve this last point, an alternative would be to manually provide
the format when using `fill_in` so it includes the seconds.
[1] https://caniuse.com/mdn-html_elements_input_type_datetime-local
This was the only place where we used the `? <= field` format; we use
the `field >= ?` format everywhere else.
We're also adding a variable to make it easier to understand that the
value is the same in both cases and to make the line shorter ;).
The reason why we were displaying the ending date as "one second before
the actual ending" was that, when seeing that a phase ends at a date
like "2000-12-31 00:00", we might end up thinking that the phase will
finished at the midnight between December 31st and January the 1st,
while it actually ends at the midnight between December the 30th and
December the 31st.
This is particularly important because we use a date field to select the
date of a phase and if select December the 31st, it will be stored in
the database as "2000-12-31 00:00". So, instead, in this case we display
"2000-12-30 23:59", which is less confusing.
But now we're going to add support for setting a time on polls, which
means a certain poll might end at 15:30. In this case, displaying that
it ends at 15:29 doesn't make much sense.
We were displaying dates in two different formats in the same component,
leading to strange hacks like manually calling the `call` method or not
being able to use `render_inline` in the tests.
Since we're going to reuse one of these formats outside the budgets
section, we're splitting the component. We're also removing the
mentioned hacks.
This is consistent with the way we show the duration of a budget and its
phases. Since budgets are the section with the most recent changes in
the admin area, we're using it as a reference.
Note that, unlike budgets (which are shown to finish at the beginning of
their ending day), a poll has always been considered to finish at the
end of their ending day, so we're showing it this way.
We're also solving a minor usability issue. While it's pretty intuitive
that a poll starting on a certain date will start at the beginning of
the day, a poll ending on a certain date isn't clear about when it
finishes exactly: is it at the beginning of the day, or at the end of
the day?
So now we're making this point clear.
In some cases (e.g. after editing or creating a resource
with errors) the default back_link did not redirect to the
expected page.
Now we force the back_links to the index pages, so we
always get the desired redirect.
These routes aren't used since commits (2993ef8707, 88a7a29d27)
In parallel while these routes were being removed, the route file
was being split into different files, and they were included again
in the commit 1cd47da9d4.
