Inconsistent callout behaviour between JIT and not-JIT

[addisoncrump]

Discovered with #322.

It appears that in some cases, the callout behaviour between JIT'd regex and non-JIT'd regex diverges in the presence of the autocallout and no_start_optimize flags.

sh-5.2$ xxd callout_crash 
00000000: 0000 0000 0400 015c 283f 216e 293f 3e3f  .......\(?!n)?>?
00000010: 3e3f 3e3e                                >?>>
sh-5.2$ ./pcre2_fuzzer callout_crash 
Encountered failure while performing match errorcode comparison; context:
Pattern/sample string (hex encoded): 283f216e293f3e3f3e3f3e3e
Compile options 00110004 never_backslash_c,auto_callout,no_start_optimize
Match options 00002000
Non-JIT'd operation emitted an error: callout error code (-37)
JIT'd operation did not emit an error.
1 matches discovered by JIT'd regex:
Match 0 (hex encoded): 3e3e

The sample regex here is (?!n)?>?>?>>.

Not sure if this is considered a bug, but may indicate that there's a deeper bug associated with these flags.

[PhilipHazel]

The code in the fuzzer aborts with error -37 if there are more than 100 callouts (to save wasting resources). The difference is that the interpreter and the JIT generate different numbers of callouts. I'm not sure why this is, but I'm guessing it's to do with different optimizations. I don't think this is a bug.

[addisoncrump]

If that's the case: should we ignore cases where the error code is callout error code?

[PhilipHazel]

Yes, I think that makes sense.

[addisoncrump]

Fixed in 895ce2e.