Unit test edge cases around consecutive \b in patterns

Current behavior is if \b consumes the start of a string, consecutive \b
is not allowed to consume the end of the string.

Should also be unaffected by wildcard.

Bug: b/493760346
Change-Id: If743daa0f2ed0a0908e3ee25a7652438fa2603d0
Reviewed-on: https://gn-review.googlesource.com/c/gn/+/21620
Reviewed-by: Takuto Ikuta <tikuta@google.com>
Commit-Queue: Richard Wang <richardwa@google.com>
diff --git a/src/gn/pattern_unittest.cc b/src/gn/pattern_unittest.cc
index e7f4e4d..7ecede7 100644
--- a/src/gn/pattern_unittest.cc
+++ b/src/gn/pattern_unittest.cc
@@ -30,6 +30,7 @@
       // Path boundaries.
       {"\\b", "", true},
       {"\\b", "/", true},
+      {"\\b\\b", "", false},
       {"\\b\\b", "/", true},
       {"\\b\\b\\b", "", false},
       {"\\b\\b\\b", "/", true},
@@ -50,6 +51,9 @@
       {"*a*b*c*d*", "abcd", true},
       {"*a*b*c*d*", "1a2b3c4d5", true},
       {"*a*b*c*d*", "1a2b3c45", false},
+      {"*\\b\\b", "", false},
+      {"\\b\\b*", "", false},
+      {"*\\b\\b*", "", false},
       {"*\\bfoo\\b*", "foo", true},
       {"*\\bfoo\\b*", "/foo/", true},
       {"*\\bfoo\\b*", "foob", false},