The --full-index option seemed to be causing caching issues on some
systems.
Since we don't know the reason why this option was added in the first
place, it might have some advantages. However, some people have reported
problems getting "version can no longer be found" errors for some gems
in this step, and documentation for Docker and Rails doesn't mention
this option at all.
Since we were creating a new answer in the form, we weren't getting the
errors associated to the answer the administrator was trying to create,
and so we were skipping the test.
Using the answer which contains the information about validation errors
fixes the issue and so we don't have to skip the tests.
I'd say this feature is actually tested in the "proposal polls specific
validations"; the empty test was probably added by accident in commit
4b8cc85c4.
This way the tests won't appear as "pending" when running the test
suite, and so we get rid of a lot of noise in the test results. There
doesn't seem to be a way to call `skip` without the test being marked as
"pending".
Note that in the globalizable tests we need to build a factory before
deciding whether an atribute is required or not (particularly for the
milestone factory, since milestone attributes are required depending on
the presence of other attributes). This isn't possible before we're
inside the test, so we can't add an `if:` condition to the test. So
we're adding the condition inside the test instead. A minor
inconvenience of this method is the test still runs even when the
condition is `false`.
We define the available locales in the test environment, so Spanish is
always available in this environment even if it isn't available in the
production environment.
This feature was only enabled for proposals five years ago, and it
hasn't changed since then. The pending test only gets in the way.
Implement. Or implement not. There is no pending.
They were marked as pending.
Note Capybara doesn't support finding a button by its `aria-labelledby`
attribute, so we're using the ugly `click_button "Yes"`, like we did in
commit fabe97e50.
It looks like it was disabled because it was failing sometimes for some
reason. I haven't found the reason, though; we're changing the test a
little bit to make it easier to read. Enabling it will let us find out
whether it still fails.
This file only has tests related to tags; if the model doesn't have
tags, we simply wouldn't include `it_behaves_as` in their tests instead
of including it and then skipping it.
The map feature was never implemented for debates (only for proposals
and budget investments) and it was crashing for debates because the page
didn't load the geozones. And we don't have a "geozone" field in the
debates form either.
So we're removing the map page alongside its (pending implementation)
tests.
CONSUL doesn't implement blank votes via web; the comment was based on
the code used in Madrid, which was actually very complex.
And the concept of "all city" was also specific to Madrid. Poll
questions aren't associated to a geozone, so the geozone will depend on
the poll they're associated to.
Note there's a difference between gitignore and dockerignore; in
.gitignore, files without a leading slash match files in every folder,
while in .dockerignore, they match files in the root folder
We used a slash as a prefix in some cases but not in other cases. Now
we're defining files and folders following the gitignore rules: files
starting with a slash are only ignored in the root folder, while files
not starting with it are ignored everywhere.
IMHO it makes sense to ignore all folders named `tmp`, `.bundle`, `log`,
`.DS_Store` or `.ruby-gemset` everywhere in the source code and not only
in the root folder.
We're also adding a trailing slash to all folders for consistency.
We can't cover the files for every editor out there, so we're removing
it in the name of neutrality. Developers using this editor to work with
CONSUL can add the `.idea` to their global gitignore configuration (on
GNU/Linux systems, for instance, the `~/.config/git/ignore` file).
We are use a display: block style for labels containing check boxes inside
them, and the label has a width of 100%.
This means that clicking on the blank space on the right of the label text
will check/uncheck the checkbox. To avoid this behaviour we modify the
"display" attribute of the labels.
In order to prevent unexpected behaviour in terms_of_service form labels,
we add specific css for this case when define a checkbox within the
.actions class.
