With two concurrent requests, it's possible to create two ballot lines
when only one of them should be created.
The reason is the code validating the line is not thread safe:
```
if ballot.amount_available(investment.heading) < investment.price.to_i
errors.add(:money, "insufficient funds")
end
```
If the second request executes this code after the first request has
executed it but before the first request has saved the record to the
database, both records will pass this validation and both will be saved
to the database.
So we need to introduce a lock. Now when the second request tries to
lock the ballot, it finds it's already locked by the first request, and
will wait for the transaction of the first request to finish before
checking whether there are sufficient funds.
Note we need to disable transactions during the test; otherwise the
second thread will wait for the first one to finish.
Also note that we need to update a couple of tests because records are
reloaded when they're locked.
In one case, reloading the ballot causes `ballot.user` to be `nil`,
since the user is hidden. So we hide the user after creating all its
associated records (which is the scenario that would take place in real
life).
In the other case, reloading the ballot causes `ballot.user` to reload
as well. So we need to reload the user object used in the test too so it
gets the updates done on `ballot.user`.
I haven't been able to reproduce this behavior in a system test. The
following test works with Rails 5.0, but it stopped working when we
moved to system tests in commit 9427f014. After that commit, for reasons
I haven't been able to debug (reintroducing truncation with
DatabaseClaner didn't seem to affect this test, and neither did
increasing the number of threads in Puma), the two AJAX requests
executed here are no longer simultaneous; the second request waits for
the first one to finish.
scenario "Race conditions with simultaneous requests", :js do
allow_any_instance_of(Budget::Ballot::Line).to receive(:check_sufficient_funds) do |object|
allow(object).to receive(:check_sufficient_funds).and_call_original
object.check_sufficient_funds
sleep 0.3
end
["First", "Second"].each do |title|
create(:budget_investment, :selected,
heading: california,
price: california.price,
title: title
)
end
login_as(user)
visit budget_investments_path(budget, heading_id: california.id)
within(".budget-investment", text: "First") { click_link "Vote" }
within(".budget-investment", text: "Second") { click_link "Vote" }
expect(page).to have_link "Remove vote"
expect(Budget::Ballot::Line.count).to eq 1
end
Using `pluck("DISTINCT")` was raising a warning in Rails 5.2:
DEPRECATION WARNING: Dangerous query method (method whose arguments are
used as raw SQL) called with non-attribute argument(s): "DISTINCT
taggings.tag_id". Non-attribute arguments will be disallowed in Rails
6.0. This method should not be called with user-provided values, such as
request parameters or model attributes. Known-safe values can be passed
by wrapping them in Arel.sql().
Since there was only one other use of distinct, I've decided to change
both of them in the same commit, even if the second one wasn't raising a
warning.
These columns were causing Rails 5.2 to throw a warning when ordering by
them, as if they weren't valid column names:
DEPRECATION WARNING: Dangerous query method (method whose arguments are
used as raw SQL) called with non-attribute argument(s):
:"budget/investments_count". Non-attribute arguments will be disallowed
in Rails 6.0. This method should not be called with user-provided
values, such as request parameters or model attributes. Known-safe
values can be passed by wrapping them in Arel.sql().
This change also makes their names consistent with the rest of our
tables and columns.
Rails 5.2 crashes in the `db:create` task because it tries to run the
`after_initialize` block before the database is created.
The easiest way to solve it is to move the code out of the initializer
and calculate the API type definitions on demand. Note results are still
cached using a class instance variable (not to be confused with a class
variable), and so once definitions are obtained, they will remain
constant until the application is restarted, even in the development
environment.
These tasks are not needed for new installations, and in existing
installations they've already been executed when upgrading to version
1.1.
One of them also raises a warning in Rails 5.2:
DEPRECATION WARNING: Dangerous query method (method whose arguments are
used as raw SQL) called with non-attribute argument(s): "MIN(id) as id".
Non-attribute arguments will be disallowed in Rails 6.0. This method
should not be called with user-provided values, such as request
parameters or model attributes. Known-safe values can be passed by
wrapping them in Arel.sql()
I'm not sure why the code didn't work without this line, but it doesn't
seem to be necessary anymore (maybe after upgrading Ruby or Rails?).
I'm removing it now because... why not now? :) The Ruby interpreter is
raising a warning due to this line, and in Ruby 2.5 constant lookup has
changed slightly (although I don't think this line is affected by that
change).
Note about the change in the Setting model: Ruby actually ignores return
values in setter methods, so the line isn't necessary.
The main obstacle to extract this partial was probably the paths for the
flag and unflag actions.
Now that we use Rails 5.1 `resolve` method to handle nested resources,
we can use `polymorphic_path`.
Also note the code is a bit ugly because comments render a divider. We
should probably use a CSS border instead.
Co-Authored-By: taitus <sebastia.roig@gmail.com>
We were treating legislation proposals as if they were proposals,
omitting the "legislation" namespace, and so we were flagging/unflagging
proposals when we wanted to flag/unflag a legislation proposal.
The `respond_with` method is no longer part of Rails (it's now part of
the responders gem) and we barely use it. Using a template forced us to
use different criteria for different controllers.
This change will also make it easier to fix the flag/unflag actions for
legislation proposals. With the old code, we would have to add another
condition for the legislation/proposals controller.
We weren't using `foundation()` in these cases, so after flagging a
debate or a comment, we had to reload the page before we could unflag
it.
We're also adding a test for the fix in commit ea85059d. This test shows
it's necessary to filter the elements with JavaSctipt using `first()` if
we want the same code to work with comments.
Co-Authored-By: taitus <sebastia.roig@gmail.com>
We don't need to set this value. In commit f2ef27d3 I made a mistake
thinking `Globalize.locale` and `I18n.locale` should always be in sync,
but they're actually automatically in sync when `Globalize.locale` is
`nil`.
So the best way to avoid any issues is not to assign `Globalize.locale`,
and use `Globalize.with_locale` where necessary instead.
Originally, the code was shared between the index action and the search
action, but since commit fb6dbdf2 that's no longer the case. So in the
index action we don't need to check whether a user is a
moderator/manager/admin/official or not; they all are.
After upgrading the rubocop-performance gem, we're asked to change this
code:
milestones.select { |milestone| milestone.image.present? }.last
To:
milestones.reverse.find { |milestone| milestone.image.present? }
IMHO the original code is easier to read, so the performance gain isn't
worth it.
Recent versions introduce the `Layout/SpaceAroundMethodCallOperator`,
which we are going to use. We aren't upgrading to the latest rubocop
version because it conflicts with the version of Capybara we're using
and because it isn't supported by Hound.
Some rules have been renamed:
Layout/IndentAssignment is now Layout/AssignmentIndentation
Layout/IndentHeredoc is now Layout/HeredocIndentation
Layout/LeadingBlankLines is now Layout/LeadingEmptyLines
Layout/Tab is now Layout/IndentationStyle
Layout/TrailingBlankLines is now Layout/TrailingEmptyLines
Lint/StringConversionInInterpolation is now Lint/RedundantStringCoercion
Metrics/LineLength is now Layout/LineLength
Note after upgrading we get a new "offense" in the `StartWith` rule, so
we're changing the code in order to fix it.
These routes are solved in a different way because of an inconsistency:
we define `groups` and `budget_investments`; we should either use the
`budget_` prefix in all places or remove it everywhere.
We can now share code using `polymorphic_path` even with these models.
In the past, we couldn't use `polymorphic_path` in many places. For
instance, `polymorphic_path(budget, investment)` would return
`budget_budget_investment_path`, while in our routes we had defined
`budget_investment_path`.
With the `resolve` method, introduced in Rails 5.1, we can use symbols
to define we want it to use `investment` instead of `budget_investment`.
It also works with nested resources, so now we can write
`polymorphic_path(investment)`.
This makes the code for `resource_hierarchy_for` almost impossible to
understand. I reached this result after having a look at the internals
of the `resolve` method in order to get its results and then remove the
symbols we include.
Note using this method will not make admin routes compatible with
`polymorphic_path`. Quoting from the Rails documentation:
> This custom behavior only applies to simple polymorphic URLs where a
> single model instance is passed and not more complicated forms, e.g:
> [example showing admin routes won't work]
Also note that now the `admin_polymorphic_path` method will not work for
every model due to inconsistencies in our admin routes. For instance, we
define `groups` and `budget_investments`; we should either use the
`budget_` prefix in all places or remove it everywhere. Right now the
code only works for items with the prefix; it isn't a big deal because
we never call it with an item without the prefix.
Finally, for unknown reasons some routing tests fail if we use
`polymorphic_path`, so we need to redefine that method in those tests
and force the `only_path: true` option.
With chromedriver >= 80, the tests are freezing sometimes, particularly
when the same editor is loaded again.
We don't know whether it's a CKEditor issue or a chromedriver issue. In
the past we've had some errors related to CKEditor trying to load the
same instance twice and we aren't sure they have been fixed since we
could never reproduce them.
It could be a coincidence, though. If we modify the views so the only
content of the `<body>` tag is a textarea with the `html-area` class,
chromedriver freezes even if we only access the page once. So maybe
we're only detecting the problem on the second visit because the second
request is faster than the first one.
Since chromedriver no longer hangs after this change, we don't have to
force any chromedriver version anymore.
After upgrading to chromedriver 80, tests checking CKEditor's content
were causing chromedriver to hang. That's why we were configuring
webdrivers to use an older chromedriver.
Version 80 of chromedriver introduced several issues regarding frames.
Debugging shows in this case chromedriver froze when we used `setData`
and then `within_frame`. Since adding a `sleep` call made it work, we
think `within_frame` was being executed before `setData` had finished.
The fact that `setData` causes the browser to enter the frame having
CKEditor is probably the reason.
Even though the `setData` method provides a callback when it's finished,
configuring it so the rest of the Ruby code isn't executed until that
happens leads to complex code. Using Capybara's `set` to fill in the
editor is IMHO a bit easier to understand.
After this change, since we're using a method provided by Capybara
instead of executing asynchronous JavaScript code, we don't have to
check CKEditor has been filled anymore. The "Admin Active polls add"
test, which failed on my machine without that check, now passes.
