regexp: repair extracting positions from string match

racket/src/regexp/demo.rkt

@@ -71,6 +71,8 @@
       '((3 . 4)))
 (test (rx:regexp-match-positions "(?m:^.\n)" "a\nb\nc\n" 2 6 #f #"\n")
       '((2 . 4)))
+(test (rx:regexp-match-positions "(?:(?m:^$))(?<=..)" "ge \n TLambda-tc\n\n ;; (extend Γ o Γx-s\n extend\n\n ;;" 29 #f #f #"\n")
+      '((46 . 46)))
 
 (test (regexp-replace* "-" "zero-or-more?" "_")
       "zero_or_more?")

racket/src/regexp/match/extract.rkt

@@ -59,11 +59,11 @@
                    (bytes->string/utf-8 bstr-in #\? (- (car p) delta) (- (cdr p) delta))))
             null)))
 
-(define (byte-index->string-index str pos)
+(define (byte-index->string-index str start-pos pos)
   ;; We assume that pos is on a code-point boundary in the
   ;; UTF-8 encoding of str. Find out how many code points
   ;; are before the index.
-  (let loop ([lo-pos 0] [lo 0] [hi (min (string-length str)
+  (let loop ([lo-pos 0] [lo 0] [hi (min (- (string-length str) start-pos)
                                         (* pos 6))])
     (cond
       [(= lo hi) lo]
@@ -71,7 +71,7 @@
        (if (= lo-pos pos) lo hi)]
       [else
        (define mid (quotient (+ lo hi) 2))
-       (define len (string-utf-8-length str lo mid))
+       (define len (string-utf-8-length str (+ start-pos lo) (+ start-pos mid)))
        (define mid-pos (+ lo-pos len))
        (cond
          [(= mid-pos pos) mid]

racket/src/regexp/match/main.rkt

@@ -349,7 +349,7 @@
             ;; boundary, and everything from `ms-pos` to `ms-end` must
             ;; still be in `lb-in`. So, find `ms-pos` in the original
             ;; string, and take it from there.
-            (define ms-str-pos (byte-index->string-index in (- ms-pos start-pos)))
+            (define ms-str-pos (byte-index->string-index in start-offset (- ms-pos start-pos)))
             (define delta (lazy-bytes-discarded-count lb-in))
             (byte-positions->string-positions bstr ms-pos me-pos state
                                               #:start-index (- ms-pos delta)