Using Selenium and Chrome in C#

This is a C# example project that shows how to attach to a current running Chrome instance (or start a new one) with GUI/browser or without GUI (headless) and then open a new tab and load a website. It also works when there are multiple Chrome instances running (or not) on different remote debugging ports. This can be GUI or headless instances.

This project uses appsettings.json as default where all settings are loaded from. You can use a parameter if you want to load settings from a different .json file. This can be handy when you want to start two instances using different settings. It also writes all output to a logfile. Logging is done by Serilog and can be controlled by changing the settings.

Important to know is that Chrome uses a profile located at some directory which contains all Chrome settings and user preferences.
Chrome can also listen on a specific port so you can attach to it from for example C# using the Selenium NuGet package and control what is happening in the browser. This way you can automate webtasks and/or test if webpages are loaded correctly and are functioning as expected.
Profile location and port can be set by manually starting Chrome using arguments like --remote-debugging-port and --user-data-dir. See startChromeWithRemoteDebugging.bat for an example how to use. If there is no Chrome available it will be auto-started by the example.

The example first checks if there is a Chrome instance running and listening on the specified DebuggingPort, if not a new instance on that port will be started which automatically reuse the correct profile when using the same port as before.
If it's the first time a new profile will be created by Chrome. If this is the case Chrome may ask some one-time questions. The answers will be persisted in the profile.
If there are any Chrome processes running on different port than the DebuggingPort from the settings you can optionally have them automatically closed. You can control this with ChromeInstancesNotListeningOnSpecifiedDebuggingPort:

• AutoClose
• DontClose
• Ask

If there is already a Chrome instance running on the specified DebuggingPort and it is in a different mode (GUI or headless) than specified in settings by RunHeadless you can control what to do with this instance by setting ChromeInstancesNotListeningOnSpecifiedDebuggingPort:

• AutoCloseAndRetry
• DontClose
• Exit

If you want to run headless you can specify extra arguments by using ExtraHeadlessArguments. For example:

	"ExtraHeadlessArguments": [ "no-sandbox", "disable-gpu" ]

For more possible arguments see https://chat.openai.com/share/45483f44-4fd8-4631-8d2a-067b81bda534

If you want to use a specific version of the chromedriver executable you can specify (only!) the path containing it. Which browser is used is specified by ChromePath. Setting both can help when specific versions are needed. It's important that both version match!! Luckily ChromeDriver will report if there is a mismatch.

The profile is either located under the application direction or found in the Windows temp path. This can be controlled by using UseTempPathForProfile.
In this example project the profile location is dependent on the DebuggingPort, so it supports running multiple Chrome instances on different ports.
You can override the path yourself by specifying OverrideProfilePath. If set to empty string "" or NULL it will be ignored.

If ChromeDriver is connected to Chrome it will open the website specified in WebsiteToOpen. After the specified number of seconds ExtraWaitingTimeAfterWebsiteLoadedInSeconds it will take a screenshot (controlled by TakeScreenshot) and/or dump the contents of the webpage (controlled by DumpContents). The screenshot (.png) and contents (.html) files can be found in the executable directory.

There are also some helper batch files for Windows:

• checkIfChromeIsListeningOnRemoteDebuggingPort.bat: checks if Chrome is running and if it's listening on a specific port.
• startChrome.bat: start Chrome normally
• startChromeWithRemoteDebugging.bat: starts Chrome on a default (or given) remote debugging port by using a specific profile (using the Windows temporary path)
• killTaskUsingPort.bat: kills a task which is LISTENING on a given port using a TCP connection. This script is also used by the example to kill a specific Chrome instance.

Be sure your regular Chrome browser version (see Help > About) matches the ChromeDriver version.

Building yourself

There is a build script buildAll.bat which builds the source to several platforms (Windows x86/x64, Linux arm/arm64/x64 and OS x64). If you are owner of this repository you can use the GitHub helper batch files publishNewRelease.bat and uploadFilesForRelease.bat to also publish these builds (it supports uploading files >25MB). Update at least the solutionfolder, version and TAG_NAME variables at the top of the scripts.

Installation / requirements / dependencies

All platforms:

• Chrome browser (must match ChromeDriver version)
• ChromeDriver (must match Chrome version)
• JDK

Windows:

• GitHub gh CLI tool (needed for the GitHub helper batch files)

Linux:

• install-dependencies-Linux-x64-Ubuntu.sh
• if user is root then set "ExtraHeadlessArguments": [ "--no-sandbox" ] in appsettings.json

Know limitations / pending issues

• If your Chrome instance saves the tab session and contains multiple tabs pointing to some websites, the example sometimes fails to attach to Chrome and won't continue opening a tab. It will fail with an connection timeout. Some workarounds:
  • before the connection timeout occurs click one or more of the tabs to force load it.
  • do not save the tab session in Chrome by switching from Continue where you left off to Open the New Tab page in Settings > On startup.
  • run in headless mode. See appsettings.json
  • use a dedicated Chrome profile which is only used for a specific purpose (like this example project or testing a website)
• On Linux you can get bind() failed: Cannot assign requested address (99) which can be ignored.
• On Linux when running as root you can get a fatal error:

Error attaching ChromeDriver: unknown error: Chrome failed to start: exited abnormally.                      
  (chrome not reachable)                                                                                                    
  (The process started from chrome location /opt/google/chrome/chrome is no longer running, so ChromeDriver is assuming that
 Chrome has crashed.)

Solution is to set "ExtraHeadlessArguments": [ "--no-sandbox" ]