In UTF mode, Nested lookbehind with an optional quantifier: unexpected no match

[YuFuSweetie]

Nested lookbehind with an optional quantifier: unexpected no match

Environment

• PCRE2 10.47 (2025-10-21)
• Windows 11, 8-bit library, UTF+UCP enabled

Pattern

/(?<=(?<=A).?)\d/

Subject

A12

Result

Only 1 matches; 2 does not. I expected 2 to match as well.

[carenas]

pcre2test will only do 1 match, unless you request the "global" modifier:

  re> /(?<=(?<=A).?)\d/g
data> A12
 0: 1
 0: 2
data>

[YuFuSweetie]

Please enable the utf mode. From my testing, the issue only occurs when utf is enabled; disabling it seems to work fine. @carenas

[YuFuSweetie]

PCRE2 version 10.47 2025-10-21 (8-bit)
re> /(?<=(?<=A).?)\d/g
data> A12
0: 1

[carenas]

that was unexpected, and seems to be a problem in the interpreter that goes all the way back to 10.43 and the implementation of VREVERSE.

% ./pcre2test -ac
PCRE2 version 10.43 2024-02-16 (8-bit)
  re> /(?<=(?<=A).?)\d/g,utf
data> A12
--->A12
 +0  ^      (?<=
 +4  ^      (?<=
 +4  ^      (?<=
 +8  ^      A
 +9  ^      )
+10  ^      .?
+12  ^^     )
+12  ^      )
+13  ^      \d
+15  ^^     End of pattern
 0: 1
--->A12
 +0   ^     (?<=
 +4   ^     (?<=
 +4   ^     (?<=
 +8   ^     A

JIT is not affected

[NWilson]

Wow, we'll definitely need to sort this out. Thank you for explaining.