[regexp] Handle zero-length matches in experimental engine

Cq-Include-Trybots: luci.v8.try:v8_linux64_fyi_rel_ng
Bug: v8:10765
Change-Id: I9e23fb5ba8a21aa1ddf96ab7936058f671bcc9ae
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2397895
Commit-Queue: Martin Bidlingmaier <mbid@google.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#69736}

src/regexp/experimental/experimental-compiler.cc

@@ -4,6 +4,7 @@
 
 #include "src/regexp/experimental/experimental-compiler.h"
 
+#include "src/regexp/experimental/experimental.h"
 #include "src/zone/zone-list-inl.h"
 
 namespace v8 {
@@ -34,8 +35,10 @@ class CanBeHandledVisitor final : private RegExpVisitor {
   static bool AreSuitableFlags(JSRegExp::Flags flags) {
     // TODO(mbid, v8:10765): We should be able to support all flags in the
     // future.
-    static constexpr JSRegExp::Flags allowed_flags = JSRegExp::kGlobal;
-    return (flags & ~allowed_flags) == 0;
+    static constexpr JSRegExp::Flags kAllowedFlags = JSRegExp::kGlobal;
+    STATIC_ASSERT(!ExperimentalRegExp::kSupportsUnicode ||
+                  (kAllowedFlags & JSRegExp::kUnicode) == 0);
+    return (flags & ~kAllowedFlags) == 0;
   }
 
   void* VisitDisjunction(RegExpDisjunction* node, void*) override {

src/regexp/experimental/experimental-interpreter.cc

@@ -6,6 +6,7 @@
 
 #include "src/base/optional.h"
 #include "src/base/small-vector.h"
+#include "src/regexp/experimental/experimental.h"
 
 namespace v8 {
 namespace internal {
@@ -84,15 +85,33 @@ class NfaInterpreter {
   // `matches_out`.  The search begins at the current input index.  Returns the
   // number of matches found.
   int FindMatches(MatchRange* matches_out, int max_match_num) {
-    int match_num;
-    for (match_num = 0; match_num != max_match_num; ++match_num) {
+    int match_num = 0;
+    while (match_num != max_match_num) {
       base::Optional<MatchRange> match = FindNextMatch();
       if (!match.has_value()) {
         break;
       }
 
       matches_out[match_num] = *match;
+      ++match_num;
+
+      int match_length = match->end - match->begin;
+      if (match_length != 0) {
+        SetInputIndex(match->end);
+      } else if (match->end == input_.length()) {
+        // Zero-length match, input exhausted.
         SetInputIndex(match->end);
+        break;
+      } else {
+        // Zero-length match, more input.  We don't want to report more matches
+        // here endlessly, so we advance by 1.
+        SetInputIndex(match->end + 1);
+
+        // TODO(mbid,v8:10765): If we're in unicode mode, we have to advance to
+        // the next codepoint, not to the next code unit. See also
+        // `RegExpUtils::AdvanceStringIndex`.
+        STATIC_ASSERT(!ExperimentalRegExp::kSupportsUnicode);
+      }
     }
     return match_num;
   }

src/regexp/experimental/experimental.h

@@ -41,6 +41,8 @@ class ExperimentalRegExp final : public AllStatic {
   static int32_t ExecRaw(JSRegExp regexp, String subject,
                          int32_t* output_registers,
                          int32_t output_register_count, int32_t subject_index);
+
+  static constexpr bool kSupportsUnicode = false;
 };
 
 }  // namespace internal