[libcxx-commits] [libcxx] [libc++][regex] Correctly adjust match prefix for zero-length matches. (PR #94550)

[Konstantin Varlamov via libcxx-commits]

================
@@ -114,5 +114,53 @@ int main(int, char**)
         assert(i == e);
     }
 
+  {
+    // Check that we correctly adjust the match prefix when dealing with zero-length matches -- this is explicitly
+    // required by the Standard ([re.regiter.incr]: "In all cases in which the call to `regex_search` returns true,
+    // `match.prefix().first` shall be equal to the previous value of `match[0].second`"). For a pattern that matches
+    // empty sequences, there is an implicit zero-length match between every character in a string -- make sure the
+    // prefix of each of these matches (except the first one) is the preceding character.
+
+    auto validate = [](const std::regex& empty_matching_pattern) {
+      const char source[] = "abc";
+
+      std::cregex_iterator i(source, source + 3, empty_matching_pattern);
+      assert(!i->prefix().matched);
+      assert(i->prefix().length() == 0);
+      assert(i->prefix().first == source);
+      assert(i->prefix().second == source);
+
+      ++i;
+      assert(i->prefix().matched);
----------------
var-const wrote:

Not sure if this repetition warrants another helper function (calls would look something like
```cpp
check_prefix(i->prefix(), /*matched=*/true, /*length=*/1, /*first=*/source, /*second=*/source + 1);
```
).

https://github.com/llvm/llvm-project/pull/94550