HiI have a selenium cucumber project with some tests.The tests worked and the chromeDriver had launch when I run the -mvn clean compile test in my IDE.I created this .gitlab-ci.yml in my project's root:image: markhobson/maven-chrome:jdk-11 stages: - build - test variables: MAVEN_OPTS: "-Dmaven.repo.local=.m2/repository" cache: paths: - .m2/repository/ - target/ build: stage: build script: - mvn $MAVEN_OPTS clean package - mvn compile test: stage: test script: - mvn test When I run the pipeline, I face this error in the console:Unable to load browser: Could not start a new session. Possible causes are invalid address of the remote server or browser start-up failure. Build info: version: '4.0.0-beta-1', revision: '9c03156c2b' System info: host: 'runner-fa6cab46-project-25183469-concurrent-0', ip: '172.17.0.4', os.name: 'Linux', os.arch: 'amd64', os.version: '4.19.78-coreos', java.version: '11.0.10' Driver info: org.openqa.selenium.chrome.ChromeDriver What can I do for running the tests?

Thank you. You mean I should two versions of the configuration. one for GitLab (with headLess mode) and one for IDE (without headLess mode)? At least at first, while you are still setting it up, you should deactivate the headless mode in the IDE. When you are sure it works everytime, you can try setting-up the headless mode. Finally, setting-up gitlab-ci. Also, regarding the following line: driver.manage().window().maximize(); It is recommended to maximize the Google Chrome browser through ChromeOptions class as follows: options.addArguments("start-maximized"); In your code, that may look like this: WebDriverManager.chromedriver().setup(); ChromeOptions options = new ChromeOptions(); options.setHeadless(true); options.addArguments('start-maximized'); driver = new ChromeDriver(options);

If I understand well, now, your tests are working well in your IDE, in both windowed and headless mode? And in gitlab-ci, you do not have a "Unable to load browser" error anymore, but the "Unable to wait and click" one? If so, I can see two reason for that: - your test is not idempotent: the state of your page changes over time. To debug this, try to take a screenshot and to retrieve it as an artifact in gitlab-ci - your app needs more time to properly load its content, and in gitlab-ci, docker images are not powerfull enough to process it in time. Try to globally increase the timeouts before a test is considered failing To make sure the whole thing is properly working and to distinguish issues in tests from issues while trying to execute the tests, you can try to add a really simple "Hello world" test in your test-suite, which load the simplest possible page and check it is displayed by the browser. If that test pass, you know that Selenium is properly set-up in the CI. If not, you can continue working on setting it up.

chrome-driver is already in your pipeline thanks to the docker image markhobson/maven-chrome:jdk-11 Your CI will always require chrome to be executed in headless mode. If you consistently have an issue also in your IDE in headless mode, you should focus on that. Try taking and checking screenshots at each steps of your test

> I noticed that the browser window zooms in on the screen and the link that I need to click on that is disappear!!! Please Look: It looks like this is a responsive issue. Chrome headless seems to have a lower definition than regular, so it apply css breakpoints to github. Testing github is part of your test-suite? Or is it a "Hello world" test to setting things up? To fix this, in the chrome options, try adding something like this: `options.addArguments("window-size=1920,1080")`. I'll let you look for some documentation if you need to. Forget about driver.manage().window().maximize(): in headless mode, you won't have any window at all, so it won't ever work.

I don't know Maven, so I don't know what does mvn compile or mvn test. But I assume your pipeline is ok and those two commands seems relevant to their respective jobs, so, that's looks good to me

run cucumber selenium tests in gitlab-ci

16 Replies

aurelien-reeves
Staff
4 years ago
Hi,

You're gitlab-ci.yml looks good

Can you provide more logs of your pipeline?

Do you properly run chrome in headless mode?

WebDriverManager.chromedriver().setup(); ChromeOptions options = new ChromeOptions(); options.setHeadless(true); driver = new ChromeDriver(options);
- rezza
  Occasional Contributor
  4 years ago
  aurelien-reeves
  Hi
  Thanks for replying.
  What is needed to providing?
  May I give all the pipeline's output?
  
  This is my file which has the driver configurations.
  And This is that file after adding the chromeDriver headLess option.
  As you can see, I added the chromeDriver headLess option.
  
  Before adding, everything is worked in IDE mode.
  But when I add the chromeDriver headLess option, I face this error in the testNG's report:
  
  feature (Login into account) cucumber.runtime.CucumberException: java.lang.AssertionError: Unable to wait and click on the WebElement, using locator: <Proxy element for: DefaultElementLocator 'By.linkText: Sign in'> at cucumber.api.testng.TestNGCucumberRunner.runCucumber(TestNGCucumberRunner.java:69) at cucumber.api.testng.AbstractTestNGCucumberTests.feature(AbstractTestNGCucumberTests.java:21) Caused by: java.lang.AssertionError: Unable to wait and click on the WebElement, using locator: <Proxy element for: DefaultElementLocator 'By.linkText: Sign in'> at org.junit.Assert.fail(Assert.java:89) at pageObjects.BasePage.waitAndClickElement(BasePage.java:59) at pageObjects.Login_page.clickOnSignInButton(Login_page.java:35) at stepDefinitions.LoginSteps.user_clicks_on_the_login_button_on_homepage(LoginSteps.java:18) at ✽.And User clicks on the login button on homepage(Login.feature:6) ... Removed 28 stack frames (Runs Cucumber Feature)
  
  I didn't run the test in GitLab mode. because first, I want to solve the problem in IDE mode.
  I'm so new, please guide me.
  Thanks in advanced.
  - aurelien-reeves
    Staff
    4 years ago
    For better readability, could you paste your code in the forum using the "insert code sample" snippet instead of JustPaste.it?
    
    And I suggest you run your tests not in headless mode in your IDE. Maybe something is actually failing but as it is in Headless mode, you can not see it.
    
    Can you remove the code about Firefox and IE to make it cleaner and easier to debug?
basnetyub
New Member
3 years ago
So what is happening here is google chrome is crashing during the run.

Chrome uses /dev/shm for runtime data which is 64MB by default under Docker. If this is not sufficient then this can cause Chrome to crash . Possible workarounds:
Increase the size of /dev/shm
Mount /dev/shm to the host's
Start Chrome with the flag --disable-dev-shm-usage

Forum Discussion

run cucumber selenium tests in gitlab-ci

16 Replies

Related Content

Selenium/Cucumber Integration

run cucumber selenium in gitlab-ci

Gitlab CI/CD license issue

Integrate TestComplete Tests into Gitlab CI/CD Pipeline

Running TestComplete through Gitlab CI and VirtualBox

Recent Discussions

Cucumber-Java question.

Multiple Step Definition match - feasibility of specifying priority

Playwright and Cucumber integration

Stuck with Cucumber.js - how do I run the same test on different pages?

Multiple Data type in cucumber