From: Justin Viiret <justin.viiret@intel.com>
Date: Wed, 18 Jan 2017 00:33:57 +0000 (+1100)
Subject: prefilter: workaround for \b in UCP and !UTF8 mode
X-Git-Tag: v4.4.0^2~4
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=cacf07fe9bf7fd396ca841f7ab8ae849500ca048;p=thirdparty%2Fvectorscan.git

prefilter: workaround for \b in UCP and !UTF8 mode

For now, just drop the assertion (which will still return a superset of
matches, as per prefiltering semantics).
---

diff --git a/src/parser/prefilter.cpp b/src/parser/prefilter.cpp
index ea58a134..f69362e4 100644
--- a/src/parser/prefilter.cpp
+++ b/src/parser/prefilter.cpp
@@ -295,6 +295,16 @@ public:
 
     Component *visit(ComponentWordBoundary *c) override {
         assert(c);
+
+        // TODO: Right now, we do not have correct code for resolving these
+        // when prefiltering is on, UCP is on, and UTF-8 is *off*. For now, we
+        // just replace with an empty sequence (as that will return a superset
+        // of matches).
+        if (mode.ucp && !mode.utf8) {
+            return new ComponentSequence();
+        }
+
+        // All other cases can be prefiltered.
         c->setPrefilter(true);
         return c;
     }