Implementation for PowerShell/PowerShell-RFC#391. A few implementation decisions: 1. Check the package manager list only once per process (**same as CMD**). So, changing the package manager list in registry doesn't affect a PowerShell session that is already running. 2. When detecting user/system PATH value changes, only check new items PREPENDED or APPENDED to the original string (**same as CMD**). The detection is made simple, which is not intended to detect complex changes to the user/system PATH. 3. The PATH update feature is disabled when the `Environment` provider is not available, or when the current session is restricted. - Both `Environment` provider and `IsSessionRestricted` are checked only once per PowerShell session. - Use `ConditionalWeakTable` to cache whether the feature is enabled or not for a given session. - We cannot use `thread-static` variable for the caching because it's possible for Runspace to reuse threads.
Pester Testing Test Guide
Also see the Writing Pester Tests document.
Running Pester Tests
Go to the top level of the PowerShell repository and run: Start-PSPester
inside a self-hosted copy of PowerShell.
You can use Start-PSPester -Tests SomeTestSuite* to limit the tests run.
Testing new powershell processes
Any launch of a new powershell process must include -noprofile so that
modified user and system profiles do not causes tests to fail. You also must
take care to call the development copy of PowerShell, which is not the first
one on the path.
Example:
$powershell = Join-Path -Path $PsHome -ChildPath "pwsh"
& $powershell -noprofile -command "ExampleCommand" | Should Be "ExampleOutput"
Portability
Some tests simply must be tied to certain platforms. Use Pester's
-Skip directive on an It statement to do this. For instance to run
the test only on Windows:
It "Should do something on Windows" -Skip:($IsLinux -Or $IsMacOS) { ... }
Or only on Linux and OS X:
It "Should do something on Linux" -Skip:$IsWindows { ... }
Pending
When writing a test that should pass, but does not, please do not skip or delete
the test, but use It "Should Pass" -Pending to mark the test as pending, and
file an issue on GitHub.