* Make null member access tests as string to avoid parsing errors
* Add PSNullConditionalOperators feature to testmetadata.json
* Add Parsing tests to be skipped
* Add tabcompletion tests
- Adds `LanguagePrimitives.IsNullLike()` method to account for `DBNull.Value` and `NullString.Value` so that they can be considered the same as a null value where sensible in PowerShell.
- Updates `-ne` and `-eq` binders to treat `DBNull.Value` and `NullString.Value` as equal to null/AutomationNull.
- Update code paths for comparing objects in LanguagePrimitives to ensure consistency with how the `-eq` and `-ne` binders work when calling LanguagePrimitives methods to do the comparisons.
- Make `LanguagePrimitives.IsNull()` and `LanguagePrimitives.IsNullLike()` public methods.
- Added tests for null behaviours in `NullRepresentatives.Tests.ps1`
.NET Core changes to return "shortest roundtrippable string" by default for the ToString() method of double and float types. This results in ToString() for double/float values sometimes return a string in 17-digit/9-digit precision format. This PR updated the double/float-to-string conversion in PowerShell to continue using the old precision specifier before the change in .NET Core 3.0.
Expands line continuance for pipelines to allow lines to continue automatically without backticks with the pipe symbol at the start of a line.
This adds to the existing functionality where pipes can be used to continue lines by placing them at the end of a line.
Fixes#7557
* Adds support for binary parsing in format echoing hex: `0b11010110`
* Works with all existing type suffixes and multipliers.
* Supports arbitrary length parsing with `n` suffix using BigInteger; details below.
* Adds `NumberFormat` enum to specify hex/binary/base 10 for the tokenizer, replacing old `bool hex`.
* Adds `n` suffix for all numeric literals to support returning value as a `BigInteger` if requested. This bypasses the issue of large literals losing accuracy when they cast through `double`.
* Adds tests for all new behaviours.
---
### Binary / Hex Parsing Implementation
* Mimics old sign bit behaviour for int and long types. Sign bits accepted for 8 or 16-bit Hex parsing, and 8, 16, 32, 64 for binary.
* i.e., `0xFFFFFFFF -eq ([int]-1)` and `0xFFFFFFFFFFFFFFFF -eq ([long]-1)`, but suffixing `u` creates `int.MaxValue` and `long.MaxValue`, respectively, instead.
* Sign bits higher than this are accepted for bigint-suffixed numerals:
* Hex: Bigint-suffixed hex treats the high bit of any literal with a length multiple of 8 as the sign bit
* Binary: Bigint-suffixed binary accepts sign bits at 96 and 128 chars, and from there on every 8 characters.
* Prefixing the literal with a 0 will bypass this and be treated as unsigned, e.g. `0b011111111`
* Specifying an `u`nsigned suffix (or combination suffix that includes `u`) ignores sign bits, similar to how parsing a hex string using `[Convert]::ToUint32()` would do so.
* Supports negating literals using `-` prefix. This can result in positive numbers due to sign bits being permitted, just like hex literals.
---
### Refactored numeric tokenizer parsing
**New flow:**
1. Check for `real` (`.01`, `0.0`, or `0e0` syntaxes)
1. If the decimal suffix is present, TryParse directly into decimal. If the TryParse fails, TryGetNumberValue returns `false`.
2. TryParse as `Double`, and apply multiplier to value. If the TryParse fails, TryGetNumberValue returns `false`.
1. Check type suffixes and attempt to cast into appropriate type. This will return `false` if the value exceeds the specified type's bounds.
2. Default to parsing as `double` where no suffix has been applied.
2. Check number format.
* If binary, manually parse into BigInteger using optimized helper function to directly construct the BigInteger bytes from the string.
* If hex, TryParse into `BigInteger` using some special casing to retain original behaviours in int/long ranges.
* If neither binary nor hex, TryParse normally as a `BigInteger`.
3. Apply multiplier value before attempting any casts to ensure type bounds can be appropriately checked without overflows.
4. Check type suffixes.
* If a specific type suffix is used, check type bounds and attempt to parse into that type.
* If the value exceeds the type's available values, the parse fails. Otherwise, a straight cast is performed.
5. If no suffix is used, the following types are bounds-checked, in order, resulting in the first successful test determining the type of the number.
* `int`
* `long`
* `decimal` (base-10 literals only)
* `double` (base-10 literals only)
* ~~`BigInteger` for binary or hex literals.~~ If the value is outside `long` range (for hex and binary) or `double` range (for base 10), the parse will fail; higher values must be explicitly requested using the `n`/`N` BigInteger suffix.
---
*This is a breaking change* as binary literals are now read as numbers instead of generic tokens which could potentially have been used as function / cmdlet names or file names.
Notes:
* Binary literal support was approved by the committee in #7557
* ~~The same issue is still under further discussion for underscore support in numeric literals and whether BigInteger parsing ought to be exposed to the user at all.~~
* ~~Supporting underscore literals is a further breaking change causing some generic tokens like `1_000_000` to be read as numerals instead.~~ Per @SteveL-MSFT's [comment](https://github.com/PowerShell/PowerShell/pull/7993#issuecomment-442651543) this proposal was rejected.
* ~~Removing underscore support or preventing standard parsing from accepting BigInteger ranges is a relatively trivial matter. It is my personal opinion that there is no particular reason *not* to hand the user a BigInteger when they enter a sufficiently large literal, but I will defer to the PowerShell Committee's judgement on this.~~
`TypeBuilder.GetInterfaces()` returns only the interfaces that was explicitly passed to its constructor, so we need to flatten the interface hierarchy in order to properly support inherited interfaces.
Fix#8028
This change adds support for specifying the underlying type for an enum:
```powershell
enum MyEnum : long
{
A = 0x0FFFFFFFFFFFFFFF
B
}
# or
enum MyByte : byte
{
A = 0x01
B = 0x02
C = 0x03
D
}
```
Ubuntu18.04 seems to default to C.UTF-8 for LANG (representing InvariantCulture) which results in a case-sensitive hashtable since CurrentCultureIgnoreCase doesn't work for that culture. Fix is to use OrdinalIgnoreCase instead.
With this commit, users who invoke a command with -Debug will no longer be presented with a prompt asking them if they want to enter a nested prompt, continue execution, or halt execution entirely. Instead, any messages sent to the debug stream will simply be sent to the debug stream and the script will continue execution.
When implementing interfaces, PowerShell incorrectly produces non-virtual get/set methods for interface-defined properties.
This commit adds a lookup method for interface-defined properties and marks get/set methods for properties with matching signatures virtual.
Consolidation of all Windows PowerShell work ported to PSCore6
* Added ps1 file import restriction. Refactored InvokeLanguageModeTestingSupportCmdlet to HelpersSecurity module
* JEA loop back fix. Debugger running commands in CL mode.
* Support for new AMSI codes. Changed to use AMSI buffer API. Unhandled exception fix.
* Fixes for module bugs while running in ConstrainedLanguage mode
* Untrusted input tracking work
* Configuration keyword bug fix, PSRP protocol version check for reconstruct reconnect, Sharing InitialSessionState in runspace pool.
* Restricted remote session in UMCI, Applocker detection collision, Help command exposing functions, Null reference exception fix.
* Added mitigation for debugger function exposure
- Adds y suffix that is used to specify a numeric literal as the sbyte data type.
- Can be combined with the existing u suffix as uy to specify the byte data type.
Support calling methods with ByRef-like type parameters in PowerShell, as long as the argument specified for the parameter can be implicitly/explicitly converted to the ByRef-like type.
We cannot create an instance of a ByRef-like type in PowerShell, but there are types that can be implicitly or explicitly converted to ByRef-like types, such as `T[] -> Span<T>`, `T[] -> ReadOnlySpan<T>`, `String -> ReadOnlySpan<T>`. `ArraySegment<T> -> Span<T>`, `ArraySegment<T> -> ReadOnlySpan<T>`. With changes in this PR, we can call methods with ByRef-like parameter types by providing the arguments of the types that can be cast to the target ByRef-like type.
**What is enabled?**
1. Invoking methods that have ByRef-like parameters, but the return type is not ByRef-like.
2. Invoking constructors with ByRef-like parameters via `[target-type]::new` syntax, but the `target-type` is not ByRef-like.
3. Accessing indexers that have ByRef-like parameters, but the return type is not ByRef-like.
This change enables index operations on objects that implement `ITuple` and other interfaces that have `DefaultMemberAttribute` declared, including slicing and negative indexing.
ByRef-like types are supposed to be used on stack only, so we need to fail gracefully when accessing properties, calling methods, or creating objects related to ByRef-like types.
Fix#7089. Native globbing on UNIX was turning absolute paths into relative paths when it shouldn't. The code suppresses generating a relative path for paths starting with '~'. It also need to do the same for paths starting with '/'.
- Make switch-statement report correct error position when it fails to evaluate the condition.
- Make for-statement report correct error position when it fails to evaluate the initializer.
- For the condition of `if/for/while/do-while/do-until` statements, the sequence point update is either duplicate in some cases (for `if/for/while`) which causes debugger to stop at the condition twice before moving forward, or missing (for `do-while/do-until`) which causes debugger to skip the condition. They are fixed.
The fix causes a problem in debugging. When stepping over the scripts in debugging mode, the debugger will stop at the switch statement condition expression twice at the very beginning -- one for evaluating the condition, and one for iterating the condition.
Measure-Object should handle `ScriptBlock` properties. Fixed by renaming `MshExpression` to `PSPropertyExpression` and making it public. Then in `MeasureObjectCommand`, lifting it up to the parameter level. Previously the implementation exposed the Property as a string and
wrapped it internally as a `PSPropertyExpression`. Now the parameter type is `PSPropertyExpression` directly allowing for both wildcard strings and `ScriptBlock`.
`PSPropertyExpression` now lives in a public namespace where it can be used by cmdlet and script authors to easily add the same type of functionality to their commands. I also modified `PSPropertyExpression` to handle hashtables properly as objects so
@{prop = 3} | measure-object prop
and
@{prop = 3} | measure-object {$_.prop}
will work the same. (Previously the example using just the property name would fail.)
After parsing `if () { }`, new lines are skipped in order to see if either `elseif` or `else` is present. When neither is present, we should resync back to the pointer before skipping those possibly available new lines, so that the new line tokens can be utilized by the subsequent parsing.
Convert ShouldBeErrorId to Should -Throw -ErrorId in PowerShell tests.
Get rid of try { } catch { } formula to assert that errors were thrown.
Small fixes in tests to obey the new Pester -Parameter syntax.
Refactor code to make it easier to maintain and a little faster. Changes are as follows:
1. Support finding a matching signature with variance. But make PowerShell prefer exact match over a match with variance.
2. The metadata signatures in `PSMethod<..>` are generated based on the array of method overloads in `MethodCacheEntry.MethodInformationStructures`, in the exact same order. So in `LanguagePrimitive.ConvertViaParseMethod`, when we try to figure out if there is a match using the metadata signatures in `PSMethod<..>`, we can get the index of the matching signature, and the same index should locate the matching method in `MethodCacheEntry.MethodInformationStructures`. Therefore, we don't need to compare signatures again in the actual conversion method, and instead, we can just leverage the index we found when figuring out the conversion in `ConvertViaParseMethod`.
- This gets rid of the reflection call `GetMethod("Invoke")` and the subsequent signature comparisons in the final conversion method.
- Also, when comparing signatures using `PSMethod<..>` in `ConvertViaParseMethod`, we can just use the generic argument types of each `Func<..>` metadata type, instead of calling `GetMethod("Invoke")` and then `GetParameters()`. This makes the code for comparing signatures simpler (the type `SignatureComparator`).
- Move `MatchesPSMethodProjectedType` from `PSMemberInfo.cs` to the type `SignatureComparator` in `LanguagePrimitives.cs`, as it's closely related to the signature comparison. Also, renamed it to `ProjectedTypeMatchesTargetType`.
- These changes make PSMethod-to-Delegate conversion a little faster, but no big improvement, as the true bottleneck probably is in delegate creation(?). Actually, the performance of this conversion is not critical at all at this moment because this feature should rarely be used in any hot script path. So this exercise is mainly for fun.
3. Remove `PSEnum<T>`. We can directly use enum types when constructing the metadata type `Func<..>`.
4. Remove the code that generates metadata signatures for generic method definitions (call `MakeGenericMethod` with fake types like `GenericType0`, `GenericType1`). This is because:
- We don't support convert generic method to delegate today, so may be better not spending time on preparing the metadata signature types for those methods.
- When the day comes that we need to support it, it's better to use generic argument types directly to construct the `Func<..>` metadata types. I left comments in `GetMethodGroupType` method in `PSMemberInfo.cs` to explain why that approach is better.
* Build Update
- Change `TargetFramework` to `netcoreapp2.1` and removed unnecessary `RuntimeFrameworkVersion` from `PowerShell.Common.props`
- Update dotnet SDK to 2.1.300-rc1-008662
- Update `TypeGen` target in `Build.psm1` to work with 2.1
- Rename macOS runtime to `osx-x64` as the old build logic expects 10.12 and breaks running on 10.13 system.
- Remove `PackageReference` to `System.Memory` as it's part of dotnetcore 2.1
- Update search for `crossgen` executable to find the matching version
* Test Update
- Update test tools `WebListener` to latest `asp.net core`
- Marked `AuthHeader Redirect` tests as `Pending` due to change in CoreFX
Add support for replacement lambdas when using the -replace operator.
Requires minimal changes to existing code by using the following overload:
Regex.Replace(string input, MatchEvaluator evaluator)
when a ScriptBlock is passed in as the replacement argument.
Also remove a couple of language tests which were actually duplicated
Change the one loop which loops through test cases to include an iteration number to remove test name duplication
Based on standard practices, we need to have a copyright and license notice at the top of each source file. Removed existing copyrights and updated/added copyright notices for .h, .cpp, .cs, .ps1, and .psm1 files.
Updated module manifests for consistency to have Author = "PowerShell" and Company = "Microsoft Corporation". Removed multiple line breaks.
Separate PR coming to update contribution document for new source files: #6140
Manually reviewed each change.
Fix#6073
Breaking-change: "0".."9" returns [char] previously in PowerShell Core (6.0.0, 6.0.1), now it returns [int]. After the change, the behavior is the same as in Windows PowerShell.
Underpinnings to make calling of Extension methods /Linq easier from PowerShell.
Enables the following that previously had to be done via reflection.
class M {
[int] Twice([int] $value) { return 2 * $value }
[int] DoubleSum([int[]] $values) {
return [Linq.Enumerable]::Sum($values, [M]::Twice)
}
}
Each PSMethod is created as with a unique type for the combinations of method signatures in the MethodInfos it represents.
PSMethod<T> where T is a MethodGroup<>, potentially recursive in the last template argument.
This way, we can determine by just looking at the type of a PSMethod if there exists a conversion from the PSMethod to a delegate.
Using the assembly name to hint at the source of the classes was
problematic in multiple ways.
This change stores the actual filename in an attribute on the assembly.
So for a given type, one can get the assembly this way:
[SomeType].Assembly.GetCustomAttributes() |
? { $_ -is [System.Management.Automation.DynamicClassImplementationAssemblyAttribute] } |
% { $_.ScriptFile }
This unifies file encoding across the inbox cmdlets to be UTF-8 without a BOM for all platforms. This is a breaking change as cmdlets on windows have a number of different encodings. This supports better interoperability with tradition Linux shells as we are using the same encoding.
Validate that files are created with UTF-8 encoding without BOM
Update tests to validate Encoding parameter to new type and create new tests for
parameter type validation.
[Breaking Change] The '-Encoding Byte' has been removed from the filesystem provider cmdlets. A new parameter '-AsByteStream' is now added to indicate that a byte stream is required as input, or output will be a stream of bytes.
- Rename powershell.exe to pwsh.exe
- Fixe appveyor.psm1
- Update MSI to include 'pwsh' in path and app paths
- Revert change for hyper-v powershell direct
- Update names in packaging.psm1.
- Fix check for SxS