As of SVN revision 1176, pcre2_get_startchar() may return an arbitrary,
undefined result.
Here is the pcre2test example input:
/x/utf
\x80\=startchar
This pcre2test input used to return
Failed: error -22: UTF-8 error: isolated byte with 0x80 bit set at offset 0
but with SVN 1176 it returns an arbitrary value, for example
Failed: error -22: UTF-8 error: isolated byte with 0x80 bit set at
offset 2156822670
This happens because at pcre2_match.c pcre2_match() returns early at
line 6337
return PCRE2_ERROR_UTF8_ERR20; /* Isolated 0x80 byte */
before setting the match_data->startchar further below.
It looks like the problem was introduced with
SVN Revision 1094
Implement support for invalid UTF in the pcre2_match() interpreter.
Without further testing, the same problem seems to be present for JIT
matching at around line 6215.
Ralf
