From c5be6485fb545894f64a8c000fa1f096413f45f2 Mon Sep 17 00:00:00 2001
From: Paul Mucur <mudge@mudge.name>
Date: Mon, 29 Mar 2021 11:56:53 +0100
Subject: [PATCH] Fix crash when scanning invalid regexp

GitHub: https://github.com/mudge/re2/issues/52

As re2 will report the number of capturing groups for an invalid regular
expression as -1, we need to check whether a pattern is OK or not before
using the number to initialize any vectors, etc.

This was causing a crash when using RE2::Scanner#scan with an invalid
regular expression as we would attempt to initialize vectors with a
length of -1. Instead, set the number of capturing groups to 0 and rely
on re2's FindAndConsumeN returning false so we return nil to the user to
indicate no match was made.
---
 ext/re2/re2.cc           | 8 +++++++-
 spec/re2/scanner_spec.rb | 6 ++++++
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/ext/re2/re2.cc b/ext/re2/re2.cc
index 41d5341..7372d85 100644
--- a/ext/re2/re2.cc
+++ b/ext/re2/re2.cc
@@ -1251,7 +1251,13 @@ static VALUE re2_regexp_scan(VALUE self, VALUE text) {
   c->input = new(nothrow) re2::StringPiece(StringValuePtr(text));
   c->regexp = self;
   c->text = text;
-  c->number_of_capturing_groups = p->pattern->NumberOfCapturingGroups();
+
+  if (p->pattern->ok()) {
+    c->number_of_capturing_groups = p->pattern->NumberOfCapturingGroups();
+  } else {
+    c->number_of_capturing_groups = 0;
+  }
+
   c->eof = false;
 
   return scanner;
diff --git a/spec/re2/scanner_spec.rb b/spec/re2/scanner_spec.rb
index 89fddf1..b0d1744 100644
--- a/spec/re2/scanner_spec.rb
+++ b/spec/re2/scanner_spec.rb
@@ -45,6 +45,12 @@
       expect(scanner.scan).to be_nil
     end
 
+    it "returns nil if the regexp is invalid" do
+      r = RE2::Regexp.new('???', :log_errors => false)
+      scanner = r.scan("Foo bar")
+      expect(scanner.scan).to be_nil
+    end
+
     it "returns an empty array if the input is empty" do
       r = RE2::Regexp.new("")
       scanner = r.scan("")