Match a glob token made only of **/ segments [patch]

TextFilter.Test/TextFilterTests.cs

@@ -1190,6 +1190,39 @@
 		Assert.IsTrue(TextFilter.IsMatch("Docs/README.md", "**/readme.md", TextFilterType.Glob, TextFilterMatchOptions.ByWholeString, TextFilterCaseSensitivity.CaseInsensitive));
 	}
 
+	[TestMethod]
+	[DataRow("**/")]
+	[DataRow("**/**/")]
+	public void GlobTokenOfOnlyGlobStarSlashMatchesItemsEndingInSeparator(string filter)
+	{
+		List<string> result = [.. TextFilter.Filter(["readme.md", "src/a.cs", "docs/", "a b"], filter)];
+
+		Assert.AreEqual(1, result.Count);
+		Assert.AreEqual("docs/", result[0]);
+	}
+
+	[TestMethod]
+	[DataRow("-**/")]
+	[DataRow("-**/**/")]
+	public void GlobExcludedTokenOfOnlyGlobStarSlashHidesItemsEndingInSeparator(string filter)
+	{
+		List<string> result = [.. TextFilter.Filter(["readme.md", "src/a.cs", "docs/", "a b"], filter)];
+
+		Assert.AreEqual(3, result.Count);
+		Assert.AreEqual("readme.md", result[0]);
+		Assert.AreEqual("src/a.cs", result[1]);
+		Assert.AreEqual("a b", result[2]);
+	}
+
+	[TestMethod]
+	public void GlobRequiredTokenOfOnlyGlobStarSlashMatchesOnlyItemsEndingInSeparator()
+	{
+		List<string> result = [.. TextFilter.Filter(["readme.md", "src/a.cs", "docs/"], "+**/")];
+
+		Assert.AreEqual(1, result.Count);
+		Assert.AreEqual("docs/", result[0]);
+	}
+
 	[TestMethod]
 	public void GlobLiteralSlashInPatternStillMatchesSlashInText()
 	{

TextFilter/TextFilter.cs

@@ -125,7 +125,7 @@
 	private static readonly TimeSpan RegexMatchTimeout = TimeSpan.FromSeconds(1);
 
 	[System.Diagnostics.CodeAnalysis.SuppressMessage("Performance", "SYSLIB1045:Convert to 'GeneratedRegexAttribute'.", Justification = "Not available in older frameworks")]
 	private static Regex RegexMatchAnything() => new(".*", RegexOptions.Compiled);
 
 	// Stands in for a pattern that has already run out of time once. The timeout bounds a single
 	// IsMatch call, so without this the same pathological pattern would pay it again for every word
@@ -300,7 +300,7 @@
 
 			return ExcludedTokenPrefixes.Contains(prefix)
 				? TextFilterTokenType.Excluded
 				: RequiredTokenPrefixes.Contains(prefix)
 				? TextFilterTokenType.Required
 				: TextFilterTokenType.Optional;
 		})
@@ -643,9 +643,16 @@
 			try
 			{
 				string? expandedToken = ExpandGlobClasses(filterToken);
-				globs = expandedToken is null ? null : [.. ExpandGlobStarSegments(MaskPathSeparators(expandedToken)).Select(pattern => caseSensitivity is TextFilterCaseSensitivity.CaseInsensitive
-					? Glob.Parse(pattern, CaseInsensitiveGlobOptions)
-					: Glob.Parse(pattern))];
+				// A token made only of "**/" segments expands to the empty string as well as to "*/".
+				// Glob.Parse throws on an empty pattern, which would mark the whole token unparseable and
+				// discard the valid expansions with it. An empty pattern could only match empty text, which
+				// never reaches the matcher, so drop it; a token with nothing left is still unparseable.
+				Glob[]? parsed = expandedToken is null ? null : [.. ExpandGlobStarSegments(MaskPathSeparators(expandedToken))
+					.Where(pattern => pattern.Length != 0)
+					.Select(pattern => caseSensitivity is TextFilterCaseSensitivity.CaseInsensitive
+						? Glob.Parse(pattern, CaseInsensitiveGlobOptions)
+						: Glob.Parse(pattern))];
+				globs = parsed is { Length: > 0 } ? parsed : null;
 			}
 			catch (Exception ex) when (ex is not OutOfMemoryException)
 			{