From c1c17305161dda72461e237c2058668aeafbcd5b Mon Sep 17 00:00:00 2001
From: Michael Stapelberg <michael@stapelberg.de>
Date: Sun, 11 Sep 2011 11:40:51 +0100
Subject: [PATCH] pcre: enable UCP, UTF-8 (if available), extend t/19-match

---
 src/regex.c            | 29 ++++++++++++++++++++++-------
 testcases/t/19-match.t | 30 +++++++++++++++++++++++++++++-
 2 files changed, 51 insertions(+), 8 deletions(-)

diff --git a/src/regex.c b/src/regex.c
index df01dae4..da8f91d8 100644
--- a/src/regex.c
+++ b/src/regex.c
@@ -21,16 +21,31 @@
  */
 struct regex *regex_new(const char *pattern) {
     const char *error;
-    int offset;
+    int errorcode, offset;
 
     struct regex *re = scalloc(sizeof(struct regex));
     re->pattern = sstrdup(pattern);
-    if (!(re->regex = pcre_compile(pattern, 0, &error, &offset, NULL))) {
-        ELOG("PCRE regular expression compilation failed at %d: %s",
+    /* We use PCRE_UCP so that \B, \b, \D, \d, \S, \s, \W, \w and some POSIX
+     * character classes play nicely with Unicode */
+    int options = PCRE_UCP | PCRE_UTF8;
+    while (!(re->regex = pcre_compile2(pattern, options, &errorcode, &error, &offset, NULL))) {
+        /* If the error is that PCRE was not compiled with UTF-8 support we
+         * disable it and try again */
+        if (errorcode == 32) {
+            options &= ~PCRE_UTF8;
+            continue;
+        }
+        ELOG("PCRE regular expression compilation failed at %d: %s\n",
              offset, error);
         return NULL;
     }
     re->extra = pcre_study(re->regex, 0, &error);
+    /* If an error happened, we print the error message, but continue.
+     * Studying the regular expression leads to faster matching, but it’s not
+     * absolutely necessary. */
+    if (error) {
+        ELOG("PCRE regular expression studying failed: %s\n", error);
+    }
     return re;
 }
 
@@ -43,8 +58,8 @@ struct regex *regex_new(const char *pattern) {
 bool regex_matches(struct regex *regex, const char *input) {
     int rc;
 
-    /* TODO: is strlen(input) correct for UTF-8 matching? */
-    /* TODO: enable UTF-8 */
+    /* We use strlen() because pcre_exec() expects the length of the input
+     * string in bytes */
     if ((rc = pcre_exec(regex->regex, regex->extra, input, strlen(input), 0, 0, NULL, 0)) == 0) {
         LOG("Regular expression \"%s\" matches \"%s\"\n",
             regex->pattern, input);
@@ -57,7 +72,7 @@ bool regex_matches(struct regex *regex, const char *input) {
         return false;
     }
 
-    /* TODO: handle the other error codes */
-    LOG("PCRE error\n");
+    ELOG("PCRE error %d while trying to use regular expression \"%s\" on input \"%s\", see pcreapi(3)\n",
+         rc, regex->pattern, input);
     return false;
 }
diff --git a/testcases/t/19-match.t b/testcases/t/19-match.t
index 4d38c41f..e4fc6ec0 100644
--- a/testcases/t/19-match.t
+++ b/testcases/t/19-match.t
@@ -139,7 +139,7 @@ sleep 0.25;
 
 # two windows should be here
 $content = get_ws_content($tmp);
-ok(@{$content} == 1, 'two windows opened');
+ok(@{$content} == 1, 'window opened');
 
 cmd '[class="^special[0-9]$"] kill';
 
@@ -148,5 +148,33 @@ sleep 0.25;
 $content = get_ws_content($tmp);
 is(@{$content}, 0, 'window killed');
 
+######################################################################
+# check that UTF-8 works when matching
+######################################################################
+
+$tmp = fresh_workspace;
+
+$left = $x->root->create_child(
+    class => WINDOW_CLASS_INPUT_OUTPUT,
+    rect => [ 0, 0, 30, 30 ],
+    background_color => '#0000ff',
+);
+
+$left->_create;
+set_wm_class($left->id, 'special7', 'special7');
+$left->name('ä 3');
+$left->map;
+sleep 0.25;
+
+# two windows should be here
+$content = get_ws_content($tmp);
+ok(@{$content} == 1, 'window opened');
+
+cmd '[title="^\w [3]$"] kill';
+
+sleep 0.25;
+
+$content = get_ws_content($tmp);
+is(@{$content}, 0, 'window killed');
 
 done_testing;
-- 
2.39.5