Regex.Split() with \G stops after first item

[sharpjs]

Description

When the \G anchor is used in a (?<=...) construct, the Regex.Split(string, string) method stops splitting after the first item.

This bug is also exposed in PowerShell 7.1.0: PowerShell/PowerShell#14112

Example

Regex.Split("aabbccdd", @"(?<=\G..)(?=..)")

Case	Result
Expected:	{ "aa", "bb", "cc", "dd" }
.NET 5.0	{ "aa", "bbccdd" }
.NET Core 3.1	{ "aa", "bb", "cc", "dd" }

Reproduction

To reproduce, clone sharpjs/Net50RegexSplitBug and run tests.

Configuration

.NET 5.0.100
Windows 20H2 (OS Build 19042.630) x64
Not specific to this configuration; easily reproducible using PowerShell 7.1.0 on Linux.

docker run -it --rm mcr.microsoft.com/powershell:preview

[regex]::Split("aabbccdd", "(?<=\G..)(?=..)")

Regression?

Yes, in .NET 5.0.100.

[ghost]

Tagging subscribers to this area: @eerhardt, @pgovind, @jeffhandley
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Description

When the \G anchor is used in a (?<=...) construct, the Regex.Split(string, string) method stops splitting after the first item.

This bug is also exposed in PowerShell 7.1.0: PowerShell/PowerShell#14112

Example

Regex.Split("aabbccdd", @"(?<=\G..)(?=..)")

Case	Result
Expected:	{ "aa", "bb", "cc", "dd" }
.NET 5.0	{ "aa", "bbccdd" }
.NET Core 3.1	{ "aa", "bb", "cc", "dd" }

Reproduction

To reproduce, clone sharpjs/Net50RegexSplitBug and run tests.

Configuration

.NET 5.0.100
Windows 20H2 (OS Build 19042.630) x64
Not specific to this configuration; easily reproducible using PowerShell 7.1.0 on Linux.

docker run -it --rm mcr.microsoft.com/powershell:preview

[regex]::Split("aabbccdd", "(?<=\G..)(?=..)")

Regression?

Yes, in .NET 5.0.100.

Author:	sharpjs
Assignees:	-
Labels:	area-System.Text.RegularExpressions, untriaged
Milestone:	-

[stephentoub]

Thanks for the repro. This is a one-line fix. I'll put up a PR tonight.