Empty concat causes vector error

[jbclements]

Using an empty concatenation as part of an SRE pattern can cause an out-of-range vector reference.

(irregex-search '(or (:) "was") "abc")

pajaro2:/tmp/irregex (git)-[master]- clements> scheme
Chez Scheme Version 9.5.4
Copyright 1984-2020 Cisco Systems, Inc.

> (import (irregex))
> (irregex-search '(or (:) "was") "abc")
Exception in vector-ref: 3 is not a valid index for #(0 0)
Type (debug) to enter the debugger.
>

[ashinn]

Wow... how did you even find this example? :)

We can rewrite these cases to (? (or "was")) but I want to figure out why this confuses the number of states.

[jbclements]

Well, as you can imagine, I was using Z3 to construct a full proof of correctness of the code...

Uh, no. It was basically just stupidity and/or luck. I think I wasn't yet sure how to specify an empty match.

With that said, I suppose a simple tester that enumerates all small SREs might well have found it. XSmith, maybe? Dunno.

[sjamaan]

It looks like somehow the states corresponding to the second or argument get entirely dropped:

> (sre->nfa '(or (:) "was") 0)
#(0 -1 () () () ((3 . #f) (0 . #f)) () ())

[sjamaan]

OK, turned out to be a pretty simple fix

[sjamaan]

Could you review it @ashinn, just to make sure it is indeed the correct fix?