| Age | Commit message (Collapse) | Author |
|---|
|
preparing for #5664 with some test improvements i ran into this issue
This commit includes a fix and the test improvements. In particular it
adds BrowserIntegrationTest#login - so there is no need to go through the signup procedure everytime you want a user to be logged in.
|
|
Hiding them using two mechanisms in case one fails:
.hidden class - bootstrap hides them then
style='display:none' - so they are hidden even if css load fails
|
|
|
|
The client_side_validations gem is not maintained anymore and the validations
were not working lately. So instead of trying to fix it I started working on
independent validations for the password as it can't be validated on the
server due to SRP.
So far these validations are very primitive. They require 8 characters length
and a matching confirmation.
|
|
input or button can be used
|
|
Feature/api quota support + current_user null pattern
|
|
cost -> rate
quota -> storage
|
|
The changes to the configuration required some non minor changes to the platform and also added some flexibility we don't require yet - and thus some new possibilities for errors.
So instead we still use the allow_..._certs and ..._cert_prefix options.
They basically provide the framework in which service levels can operate.
The service level configuration will not include the cert prefix anymore.
It only states if the service level is rate limited or not.
This avoids conflicts between the two configuration options.
I also removed the anonymous service level entirely.
It was also turning a boolean decision (do we provide anonymous eip or not) into something way more complex. Instead I added the AnonymousServiceLevel class to handle the corner cases for people who are not logged in.
Furthermore i renamed the UnauthenticatedUser to AnonymousUser so it matches the Anonymous Service Level nicely. It's also shorter and more intuitive.
|
|
|
|
Null Pattern for current_user - use it to get rid of some conditionals
|
|
:api/service will return a hash of the current users service level
This is failiing if the user is not logged in. Instead it should return the service description for an anonymous user.
|
|
they have response.json as a dict instead of response.json()
|
|
|
|
|
|
|
|
now the webapp_login test looks nice and clean. soledad next.
|
|
|
|
nagios_test.run takes a function and executes it.
If it returns nothing or 0 and OK nagios message is printed.
If it returns sth. else this will be printed a a warning
If it raises an exception that will result in a CRITICAL report.
This way we can keep the nagios things outside the test cases and just write simple functions that either return 0, a warnign or raise a meaningful exception
|
|
|
|
|
|
it may not have been required before the RackTest support class.
|
|
Also renamed help to support so it's harder to confuse it with documentation
|
|
cleaned up all the engine stuff that was never really used.
Afterwards there is not that much left that makes it into the toplevel.
|
|
prevent recursive loadign of test/factories.rb ** can be empty.
|
|
|
|
some message tests are failing for me right now.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* store the last screenshot in tmp
* store the html and the end of the server log in tmp
Filenames are build from the test class name and the tests name.
|
|
Tests fail on machines that default to other available languages.
Also move setting capybara driver to be javascript driver into the BrowserIntegrationTest.
|
|
response headers (in particular, 'X-Minimum-Client-Version'). It must now be placed in config/provider/provider.json
|
|
* set locale based on request header
* enforce locale path prefix when current locale is not the default
* note: don't use root_path anymore, instead use home_path
|
|
|
|
We add a class to the html element based on the detected os and use that to pick which download link should be visible.
If we detect an os that is not supported we display a deactivated download link instead with all alternatives.
|
|
|
|
|
|
|
|
|
|
This test would fail sometimes on
assert_equal '/', current_path
I believe it was a timing issue. page.has_content? will wait for the content to show up. So afterwards the current_path should always be correct.
|
|
We use port 3003 for the integration test server.
This test takes a few seconds (~8) now. Most of this is startup time of the server. A second run still takes 2 seconds like before.
|
|
|
|
This way we can track down couch version related issues on travis.
|
|
* first setup couch similar to what we'll have on the platform
* then run migrations as admin
* then drop admin privileges
* then proceed with the normal test script
|
|
|
|
|
