Should be more robust.

Remove duplicated code.
See #2641.

See #2641.

converter() could be rather clone.getGene()
See #2641.

See #3431.

See #2465.

See #3431.

We want to regularly run all our tests (and not to rely only on branch names).

See #3430.

Vidjil parser prototype

See merge request !201

allows vidjil files to contain no clones

merge -- packageless docker, server functional tests

See merge request !225

Merge branch 'vidjil_parser_prototype' of gitlab.inria.fr:vidjil/vidjil into vidjil_parser_prototype

Closes #3418.

See #3418.

feature-a/3393-binreadstorage-use-bins-when-necessary into dev

Closes #3393 and #3395

See merge request !267

Feature a/update TODOs

See merge request !257

Changing the test order for browser_test.rb is not the best idea in the
scope of this feature. Implementing a ServerTest class that inherits
from BrowserTest changing only the test order allows us to maintain
behaviour for browser tests and still have a random order for server
tests.

since the test order is random, the successful login test can be
executed before the failure. So we need to logout if we are already
logged between the tests

Found another crappy test thanks to random order. In some cases, the
patient needed for the edit test was being deleted before the test could
execute. So selecting a specific patient for the delete test should make
things more robust.

Watir was clicking on the get_results link (in the centre or the rows)
which was causing issues with the tests. Clicking the first cell of the
line ensures that no link is being clicked.

Add some pre_process dummy data and add edit and delete test

Search for a patient that has not been edited since this test can
execute before the edit test.

Since the order of tests is no longer garranteed, we can't expect the
index of patients to be the same each time. So we need to select our
patient based on another field (like info). And use the sample_set_id to
retrieve it after the info has been edited.

since we have implemented a dataset, we currently cannot test the
database initialisation method.

Instead of having a test that needs to be run after everything, use
Ruby finalizer to close browser.
This means we can then run tests in a random order and ensure we don't
have any order dependant tests

for simplicity

with the new dataset, the situation changed in the test where we no
longer had any results to delete. But we have come back to a state where
there are results to delete, and now match the checklist in #3401 again.

Since we now have completed tasks in our samples, there are two buttons
that match our criteria to find the delete button. So we need to ensure
only the first one is clicked

add a data_file to sequence_file and add a status to the scheduler_task.
This ensures the task will not be run by an existing scheduler. And
having the data_file ensures the run icon will appear.