I have developed a framework with selenium + .NET
I'm running all test using Bambo agent. Everything is working fine, except one thing.
As you know, chromedrive can not set window size bigger than actual machine screen resolution.
My bamboo machine has set 1920x1080 resolution. To make sure that it is for real 1920x1080, I wrote a powershell script that is executed on agent berofe tests and print out resolution. Btw I have bamboo agent with Windows
But when webdriver start browser (chrome, firefox or IE) it limits the browser window size to 1044x788. Looks like the webdriver is getting wrong resolution from bamboo agent.
As you know webdriver is not allowing to set bigger size of window than you actual machine has. Thats why we limit browser size to 1044x788
There is no way to change size of window for bigger one. Things like: Manage().Window.Maximize() or Manage().Window.Size = new Size (1920x1080) or starting chrome with webdriver adding some options for chrome like: chromeOptions.AddArgument("--window-size=1920,1080"); it is always limited to 1044x788.
Did any one have similiar problem? Can I somehow "hack" webdriver to open browser with 1920x1080? Meaby somebody knows how webdriver is checking the screen resolution? If he is looking for resolution in some windows file then meaby I could inject 1920x1080 resolution there?
Or meaby there is some setting in bamboo agent that I missed?
Why is that important? Becouse at same point I faced problem that on this smaller browser some webelements sometimes are overlaping (figure out that on screenshots) and it case some test to fail.
When I run test on my laptop it always have this 1920x1080 resolution.
P.S. I know about Selenium Grid but my aproach works great for me, until now :)
Before second I wrote another powershell script and create a screenshot on bamboo machine, and the screenshot has 1024x768 resolution. So meaby it is coused somehow by machine..
I know the answer. Bamboo runs all tasks as windows service (not as a user). That's why I was not able to see the browser and the screen resolution was wrong. There are two solutions:
Set up selenium grid hub and node on that one machine with your user, and paste url of hub in you framework. Then bamboo service will trigger selenium grid hub that is set by your 'user' and you will be able so see browser when you will log into the virtual machine and it wil have normal resolution (I have select this solution)
I think that there could be another solution to set up bamboo to run tasks not as windows service but as a user (not sure if it is possible).
Related
I have a Java Application, which controls an automated GUI test in a FF-Browser via Selenium WebDriver Libraray. The Java App reads test cases from a database and executes them according to the code logic.
For instance, if the app reads in a Field, it'll search it by using the "findElement"-method from the Selenium framework. I do not use any test scripts for Selenium.
Currently this is happening on a local workingstation of an employee.
Now I want to move this whole environment into a Docker container.
Is it even possible to instantiate a Firefox Browser in a Container?
btw: I do not need to see the actual GUI of my browser.
And secondly:
There are several containers with selenium on dockerhub ready to use, but these do not fit my surroundings am I right?
As far as I know the SeleniumGrid expects testscripts and cannot be executed through runtime.
I open up a Linux VM (Debian:Jessie distribution) with Vagrant, in which then runs Docker.
I am still a beginner with Docker.
I couldn't find any question around here regarding my purpose.
Thanks in advance!
Is it even possible to instantiate a Firefox Browser in a Container?
Yes. The simplest way to do this is would be using the selenium images on Docker Hub.
There are several containers with selenium on dockerhub ready to use, but these do not fit my surroundings am I right?
If you think the Selenium images don't work for you because they are all based on Selenium Grid, you can use the StandaloneFirefox and StandaloneChrome images instead. These are individual instances, they do not use Selenium Grid.
BTW, the non-Debug Selenium images do not have a GUI. You mentioned you didn't need to see the browsers running so these should be fine. If you do need to see the browsers, the Debug images have a VNC server installed so you can run the image, connect with a VNC client, and watch the browsers run the tests.
I'm having trouble running some Selenium tests on a Jenkins Slave. To be specific, the display resolution that is used to run the tests is too small, causing some of the tests to fail.
To check the display resolution, we log the display height and width to the console, using:
driver.manage().window().maximize();
System.out.println("Window height: " + driver.manage().window().getSize().getHeight());
System.out.println("Window width: " + driver.manage().window().getSize().getWidth());
This returns:
Window height: 784
Window width: 1040
which seems like a very strange resolution to me. The desired resolution is 1920 x 1080.
The server that is used as a slave is a virtual machine (Windows Server 2012 R2). The Jenkins Slave is Connected via JNLP agent. The slave has the service running with Log On As "Local System" with the "Allow service to interact with desktop" option enabled.
So far we've tried a number of things like:
Connecting to slave VM using RDP and disconnecting to leave session open with desired display resolution
Using powershell to set the display resolution
Setting the default display resolution in the VM configuration
Setting the window dimensions using Selenium
And more...
All of these didn't resolve the issues. Suggestions are very welcome!
Finally managed to fix these issues after realizing that Jenkins does not necessarily need to run the slaves as a windows service. To start the slave, the JNLP agent can be downloaded from Jenkins and copied to the server. When running the JNLP file, you can select for the option to install the slave agent as a service.
Previously we had this option selected, that's why the slave was running as a service. After stopping and removing the service, we ran the JNLP file again and made sure to not select the option.
The solutions suggested that included disconnecting the remote desktop session and leaving the session open with a large resolution didn't work when running the slave as a service. They do work however when running the slave in the default way.
Make sure that the remote desktop session is not ended after a certain period of time:
Hope this helps someone!
Based on my experience with this you can't solve this problem programmatically. Your tests will run on the resolution which was used last time when you physically accessed this VM's display. For example, if I open VM on my big screen monitor and maximize it, tests will run on that resolution. But if I open it on my laptop screen and close RDP connection, tests will run on that smaller screen size.
I know it sounds strange, but I really couldn't find better solution. :D So now I must be careful to maximize VM display on my bigger screen before I close VM. You will probably dislike this answer, but remember it when you find yourself out of other solutions. ;)
Solution that worked for me is to run Chrome in a 'headless' mode (without GUI). It works with Jenkins Agent running as a service, when GUI is not available.
Here is the code sample of the web driver initialization:
var options = new ChromeOptions();
options.BinaryLocation = #"C:\Program Files\Google\Chrome\Application\chrome.exe";
options.AddArgument("Headless");
options.AddArgument("window-size=1920,1080");
driver = new ChromeDriver(#"<path>\Selenium.WebDriver.ChromeDriver.2.37.0\driver\win32", options);
I've been experienced the same issue. My seleniumn tests run under jenkins slave installed as windows service, using "Local System" account with "Allow service to interact with desktop" option enabled in windows 7. Some test cases were always failed due to incorrect display resolution.
I logged in to the console of the windows VM (EXSI Server + VMware Fusion), changed the resolution to 1400x900, and restart the windwos VM. Everything works well now.
I was asked a question :
Assume I have 4 machines and I need to execute a script in all the machines across all the browsers. How will I achieve that.
I told him the concept of Selenium Grid, where in we could set up a machine which acts like a hub, configure 3 more machines which would act like a node.
Using Desired Capabilities among others we could choose a browser type and version type in that and write a script.
But he asked me two things :
IN all the node machines how do you configure the Windows username and Password if the machine is locked. Can you write windows Authentication in the script.
Can I achieve testing different browser versions of same browser type in a single node?
Can I pass as a the browser type and browser version as a parameter
from hub to the node?
Can someone throw some light on these as I was unable to answer. Thanks.
Question 1: Is it really necessary for the machine to be unlocked for the test to start? The selenium node is a background process that listens for commands, and executes them on the browser, so I do not think this is necessary. If it is necessary due to your specific windows settings however, then no, you cannot do this from the selenium script obviously.
Question 2: Yes, you can test different browser versions of the same type on the same node. You can pass the browser name and version to the node. However, keep in mind that the node cannot know the location of the different browser versions, so you will also have to supply the path to the browser executable for your requested version
We have Xserver-less CentOS system for continuous integration. So no UI.
I want to run selenium test cases on it. I am using new Alpha version of Selenium2 which has WebDriver integration. It starts and ends browser by its own. Now, How do I start that in background? Something similar is possible with Internet Explorer?
I am able to run correctly on local machine. But it opens Firefox UI.
Are you able to run a virtual frame buffer? That way, all the GUI programs can believe there's graphics happening, even if there's no actual screen. You may wish to search for the term "headless". One example.
Another possibility may be to use Celerity - a ruby-based "headless browser". It uses the Watir interface (originally a competitor to selenium), which I understand to be merging with webdriver.
I'm executing selenium tests with testng, that are started on a remote system with Selenium RC via hudson (with ssh connection). The remote system is windows xp with MKS Toolkit installed, hence ssh. Tests are NOT executed as a windows service.
I've tried using both captureScreenshot and captureEntirePageScreenshot methods. The first one always produces a black image. The second one creates the correct screen shot but it only works on Firefox and our tests usually pass on Firefox and fail in other browsers, so it is crucial to capture screen shots for the other browsers (mainly IE and Safari). The tests are ran in parallel, with many browser windows open at the same time. I'm not certain if this is what's causing the problem. Any thoughts will be appreciated.
Unfortunately screenshots in Selenium have been problematic from the start in browsers that are not Firefox. This is something that we Selenium Developers have been working on for a while to correct.
The latest work has been updating Snapsie to work in IE. There is a blog post at http://blog.codecentric.de/en/2010/02/remote-screenshots-mit-selenium-und-dem-robot-framework/ that explains what has happened.
I have noticed that if the screen isn't active, i.e. the screensaver has kicked in, it can produce black screenshots.
Edit:
I just had a thought. You can always run Castro to video record your tests and then watch it play back. This is something SauceLabs use to run Selenium in the cloud.
Write a method for this and call that whereever you need to take the screenshot. Use the java.awt package which has been used in selenium. For example, check this site
After setting Windows Auto-Logon, and launching process not as Windows Service, I found how to solve the Remote Desktop with Black Screenshots problem of IEDriverServer.exe, by creating a batch file that disconnects RDP, instead of closing the RDP session with the regular X button:
%windir%\system32\tscon.exe %SESSIONNAME% /dest:console
See more details here:
https://stackoverflow.com/a/24529629/658497
(Although, I would prefer there was a way to run it as the default action, when terminating RDP session with X Windows button).