The version of Chromium we have in our images (120) is relatively
unstable and our system specs break regularly.
This patch makes sure Chrome is used instead for system specs.
Why this change?
We run on different runners depending on the scenario. We should use the
right number of parallel jobs for bundle install based on the number of
CPU cores the runner has.
This is so the CI output on GitHub actions isn't showing
tons and tons of unnecessary log data every time you want
to see the important thing, which is the actual test failure.
Why this change?
The output is too verbose and prevents us from quickly identifying tests
failures. Now that our tests are way more stable and less flaky, we can
drop the documentation format since we do not need it for debugging
purposes that often anymore
Before this commit, we had a yarn package set up in the root directory and also in `app/assets/javascripts`. That meant two `yarn install` calls and two `node_modules` directories. This commit merges them both into the root location, and updates references to node_modules.
A previous attempt can be found at https://github.com/discourse/discourse/pull/21172. This commit re-uses that script to merge the `yarn.lock` files.
Co-authored-by: Jarek Radosz <jradosz@gmail.com>
Why this change?
Our tests are more stable these days and there is little to no need for
us to be retrying on PRs which helps to increase confidence in our test
suite since flaky tests are raised earlier.
Why this change?
This regressed in 6e9fbb5bab because we
had a `request.xhr?` check before we decide to block requests. However,
there could not none-xhr requests which we need to block as well at the
end of each system test when `@@block_requests` is true.
This also reverts commit 6437f27f90.
Why this change?
On CI, we have been seeing flaky system tests because ActiveRecord is
unable to checkout a connection. This patch is meant to help us debug
which thread is not returning the connection to the queue.
Example of timeout issue: https://github.com/discourse/discourse/actions/runs/8012541636/job/21888013082
Why this change?
On CI, we have been seeing flaky system tests because ActiveRecord is
unable to checkout a connection. This patch is meant to help us debug
which thread is not returning the connection to the queue.
Why this change?
We have been seeing checkout timeouts happening on CI when using the
default of 5 seconds. This can happen in system tests when the server
has to process many requests using the same database connection.
Therefore, we will double the timeout for now and monitor if stuff
continues to timeout.
Why this change?
I have been investigating transaction related issues with our system
tests and I have a hard time figuring out what is causing the problem.
To help simplify our environment further, we will set the pool size in
the test environment to 1 so that it is impossible for us to be fetching
a different connection between the threads since they all share the
connection pool.
Also set `reaping_frequency` to `0` to ensure we don't reap any
connection ensuring the same connection is always used.
Why this change?
In https://www.postgresql.org/docs/current/non-durability.html, it is
recommended to create unlogged tables to avoid WAL writes which can help
speed at performance at the expense of durability. In the CI env, there is no need for durability at all.
Therefore, we are going to be creating unlogged tables by default.
Co-authored-by: Ted Johansson <ted@discourse.org>
Co-authored-by: Rafael dos Santos Silva <xfalcox@gmail.com>
Why this change?
We have been looking into a flaky system tests in one of our plugins
where the DB transaction flow can be messed up from time to time. Our
debugging effort is complicated by that fact that `test-prof` starts a
DB transaction in a `before(:all)` block which makes it hard to properly
log information. By allowing test-prof to be disabled completely, we
believe it will make it easier for us to isolate the problem we are
investigating.
What does this change do?
1. Avoid loading test-prof files if `PREFABRICATION` env has been set to
`0`.
2. Set `PREFABRICATION=0` for plugin system tests in Github actions
Why this change?
When running system tests on our CI, we have been occasionally seeing
server errors like:
```
Error encountered while proccessing /stylesheets/desktop_e58cf7f686aab173f9b778797f241913c2833c39.css
NoMethodError: undefined method `+' for nil:NilClass
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/path/pattern.rb:139:in `[]'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:127:in `block (2 levels) in find_routes'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:126:in `each'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:126:in `each_with_index'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:126:in `block in find_routes'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:123:in `map!'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:123:in `find_routes'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/journey/router.rb:32:in `serve'
/__w/discourse/discourse/vendor/bundle/ruby/3.2.0/gems/actionpack-7.0.7/lib/action_dispatch/routing/route_set.rb:852:in `call'
```
While looking through various Rails issues related to the error above, I
came across https://github.com/rails/rails/pull/27647 which is a fix to
fully initialize routes before the first request is handled. However,
the routes are only fully initialize only if `config.eager_load` is set
to `true`. There is no reason why `config.eager_load` shouldn't be `true` in the
CI environment and this is what a new Rails 7.1 app is generated with.
What does this change do?
Enable `config.eager_load` when `env["CI"]` is present
The regen_ember_5_lockfile script was actually just duplicating the ember3 lockfile without changes 🤦♂️. This commit fixes that, and updates the ember-version-enforcement workflow to detect lockfile issues in future.
Why this change?
`upload-artifacts` v4 introduces a breaking change where uploading
multiple uploads to the same name artifact is no longer supported.
However, we have external services reading these artifacts and the
change in the artifact name is breaking those services.
Why this change?
`github.job` returns the `job_id` per the docs but it doesn't actually
return the id of the job but instead returns the job's name strangely.
Per https://github.com/orgs/community/discussions/8945, there is no way
to get the `job_id` from the existing contexts in the actions run.
Therefore, we have to hit Github's API to fetch it. Not ideal but no
way around this.
Why this change?
The `tests` workflow runs many jobs. Each job when ran is given a unique
id. Since a job can be re-run, we do not want the test reports to
override each other so we differentiate it further by the `job_id` given
by `${{ github.job }}`.
Why this change?
Pull requests can introduce flaky tests into the mix and we do not want
to be hiing that during the pull request process. While this does mean
builds for PR will be less stable than the `main` branch without
retries, we do not foresee this to be a problem long term since the
monitoring of flaky tests on the `main` branch will mean that the number
of flaky tests will eventually be reduced.
What does this change do?
1. Introduce the `DISCOURSE_TURBO_RSPEC_RETRY_AND_LOG_FLAKY_TESTS` env
variable which will initialize `TurboTest::Runner` with the `retry_and_log_flaky_tests`
kwarg set to true when set.
2. Change the tests workflow run to set `DISCOURSE_TURBO_RSPEC_RETRY_AND_LOG_FLAKY_TESTS` only when
the build type is `backend` or `system` and the `github.ref_name` is
`main`.
It's very unlikely that something will be introduced which works under Ember 5 and not Ember 3. To reduce GitHub actions costs, flakiness, and visual noise, let's cut down the matrix so we're only using Ember 3 for the 'core frontend' job. All others can run under Ember 5.
What motivated this change?
Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence
in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we
have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers
assumed it was a flaky test.
What does this change do?
This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when
they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk
which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to
lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact
on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests
which have been logged for internal monitoring.
How is the change implemented?
1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner`
with the `retry_and_log_flaky_tests` kwarg set to `true`.
2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional
formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method.
The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the
tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`.
3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on
disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process.
In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
Loïc Guitaut
Alan Guo Xiang Tan
Martin Brennan
David Taylor
Jarek Radosz
Selase Krakani
dependabot[bot]
Gerhard Schlager