Str.regexp raises "Invalid_argument: index out of bounds"

[vicuna]

Original bug ID: 7369
Reporter: @johnwhitington
Assigned to: @damiendoligez
Status: resolved (set by @xavierleroy on 2016-12-04T15:34:48Z)
Resolution: fixed
Priority: normal
Severity: major
Version: 4.03.0
Target version: 4.05.0 +dev/beta1/beta2/beta3/rc1
Fixed in version: 4.05.0 +dev/beta1/beta2/beta3/rc1
Category: otherlibs
Monitored by: @gasche

Bug description

#load "str.cma";;

let x = "if\( \|\t\|\n\|\f\|\t\|\).\( \|\t\|\n\|\f\|\t\|\)=\( \|\t\|\n\|\f\|\t\|\).*";;

val x : string =
"if\( \|\t\|\n\|\f\|\t\|\).\( \|\t\|\n\|\f\|\t\|\)=\( \|\t\|\n\|\f\|\t\|\).*"

Str.regexp x;;

Exception: Invalid_argument "index out of bounds".

Additional information

The backtrace is not very useful, just pointing to recursive invocations of emit_code:

Raised by primitive operation at file "str.ml", line 254, characters 4-49
Called from file "str.ml", line 385, characters 6-17
Called from file "str.ml", line 344, characters 6-17
Called from file "str.ml", line 413, characters 6-17
Called from file "str.ml", line 422, characters 2-14

[vicuna]

Comment author: @johnwhitington

I missed the top line. The error is in patch_instr at l254

(* Fill a reserved instruction slot with a GOTO or PUSHBACK instruction *)
let patch_instr pos opc dest =
(!prog).(pos) <- (instr opc (displ dest pos)) in

from a Group (n, r) call to emit_code at l385
from a Star r match at l344

Here are the input values to patch_instr for the whole run:

Str.regexp x;;

patch_instr prog length = 32, opc = 16, dest = 11, pos = 8
patch_instr prog length = 32, opc = 15, dest = 12, pos = 10
patch_instr prog length = 32, opc = 16, dest = 13, pos = 7
patch_instr prog length = 32, opc = 15, dest = 14, pos = 12
patch_instr prog length = 32, opc = 16, dest = 15, pos = 6
patch_instr prog length = 32, opc = 15, dest = 16, pos = 14
patch_instr prog length = 32, opc = 16, dest = 17, pos = 5
patch_instr prog length = 32, opc = 15, dest = 18, pos = 16
patch_instr prog length = 32, opc = 16, dest = 19, pos = 4
patch_instr prog length = 32, opc = 15, dest = 19, pos = 18
patch_instr prog length = 32, opc = 16, dest = 22, pos = 1
patch_instr prog length = 32, opc = 16, dest = 25, pos = 22
patch_instr prog length = 64, opc = 16, dest = 35, pos = 32
patch_instr prog length = 64, opc = 15, dest = 36, pos = 34
patch_instr prog length = 64, opc = 16, dest = 37, pos = 31
patch_instr prog length = 64, opc = 15, dest = 38, pos = 36
patch_instr prog length = 64, opc = 16, dest = 39, pos = 30
patch_instr prog length = 64, opc = 15, dest = 40, pos = 38
patch_instr prog length = 64, opc = 16, dest = 41, pos = 29
patch_instr prog length = 64, opc = 15, dest = 42, pos = 40
patch_instr prog length = 64, opc = 16, dest = 43, pos = 28
patch_instr prog length = 64, opc = 15, dest = 43, pos = 42
patch_instr prog length = 64, opc = 16, dest = 46, pos = 25
patch_instr prog length = 64, opc = 16, dest = 57, pos = 54
patch_instr prog length = 64, opc = 15, dest = 58, pos = 56
patch_instr prog length = 64, opc = 16, dest = 59, pos = 53
patch_instr prog length = 64, opc = 15, dest = 60, pos = 58
patch_instr prog length = 64, opc = 16, dest = 61, pos = 52
patch_instr prog length = 64, opc = 15, dest = 62, pos = 60
patch_instr prog length = 64, opc = 16, dest = 63, pos = 51
patch_instr prog length = 64, opc = 15, dest = 64, pos = 62
patch_instr prog length = 64, opc = 16, dest = 65, pos = 50
patch_instr prog length = 64, opc = 15, dest = 65, pos = 64

So it's trying to patch an instruction in the bytecode at position 64 in a bytecode program array 0...63.

[vicuna]

Comment author: @damiendoligez

See #826 ( #826 ).

[vicuna]

Comment author: @xavierleroy

GRP#826 merged, fixing this bug.