bugfix: matching $ must count characters from end, not bytes

1 files changed, 20 insertions, 2 deletions

diff --git a/booki.c b/booki.c
index ffe055a..2bd30bf 100644
--- a/booki.c
+++ b/booki.c
@@ -1,3 +1,4 @@
+#include <assert.h>
 #include <getopt.h>
 #include <stdbool.h>
 #include <stdio.h>
@@ -631,6 +632,23 @@ char* next_book(char* current_pos) {
     return current_pos;
 }
 
+char* characters_from_end(char* str, int len, int number_of_chars) {
+    assert(len >= number_of_chars);
+    int retchars = number_of_chars;
+    char* newptr = str + len - 1;
+    unsigned char ch;
+    int total_bytes = 0;
+    while (number_of_chars > 0) {
+        ch = (unsigned char) *newptr;
+        // if we're at the first byte of a unicode point, adjust
+        if (ch < 0x80 || ch >= 0xC3)
+            number_of_chars--;
+        newptr--;
+        total_bytes++;
+    }
+    return newptr + 1;
+}
+
 /*** search ***/
 bool match_string(const char* pattern, const struct es text) {
 
@@ -652,8 +670,8 @@ bool match_string(const char* pattern, const struct es text) {
         // text must match the pattern starting from pattern + 1
         valid = comparable(pattern + 1, text.ptr, pattern_length - 1);
     else if (tail_match) {
-        // text starting from (pattern + 1) from the end must match pattern (without $)
-        valid = comparable(pattern, text.ptr + (text.len - pattern_length + 1), pattern_length - 1);
+        // text starting from (pattern - 1) characters(!) from the end must match pattern (without $)
+        valid = comparable(pattern, characters_from_end(text.ptr, text.len, pattern_length - 1), pattern_length - 1);
     }
 
     // we only need to compare while remaining text is