Currently pgagent doesn't handle unicode correctly.
CharToWString function corrupt multibyte characters because it processes
string one byte at a time:
148 std::string s = std::string(cstr);
149 std::wstring wsTmp(s.begin(), s.end());
WStringToChar function does not take into account that there can be
_multi_byte character on wcstombs output and create buffer with
size = wcslen:
157 int wstr_length = wcslen(wchar_str);
158 char *dst = new char[wstr_length + 10];
Also pgagent do not setup locale with setlocale(), without it all
wcs/mbs functions cannot handle multibyte strings.
For example:
=== step code ===
select 'это проверка кириллицы в теле запроса pgagent'
=================
=== postgres log ===
2021-02-05 23:19:05 UTC [15600-1] postgres@postgres ERROR: unterminated quoted string at or near "'" at character 8
2021-02-05 23:19:05 UTC [15600-2] postgres@postgres STATEMENT: select '
====================
Please see attached patch.
I only test it on GNU/Linux and can't test it on Windows, sorry.
--
Sergey Burladyan