ODBC. Value from DOUBLE rounded and trimmed to 1st digit

Hello.
My small, simple, VisualStudio C++ application queries database via ODBC driver/Data Source.
This database contains the “val” table containing fields of DOUBLE and DECIMAL type.

< < < Problem < < <
Value from DECIMAL field is displayed properly but value from DOUBLE field is wrong: is rounded and trimmed so only the first digit remains.
> > >

I have performed some investigation.

a. When I query my database directly, via “MySQL Client(MariaDB...) as user, it works well:
“select * from val;” gives proper answers from both type fields: DOUBLE and DECIMAL.

b. When I examine my database with HeidiSQL it also looks good.

c. But when I debug my application with VisualStudio I get CRecordset record with some errors.
GetFieldValue(...DECIMAL field...) returns string with the proper value (it is OK) but
GetFieldValue(...DOUBLE field...) returns string with error as above.
Hence my application displays wrong values.

Here are some examples for DOUBLE field:
a, b: (proper value: directly from db) - > c: (from/via ODBC driver/Data Source):
3000 - > 3
12000 - > 1
21.27 - > 2
175.68 - > 2
19.68 - > 2
So value is rounded and only the first digit remains.

d. When I roll the ODBC connector back to version 3.1.11 (the database was created at):
( mariadb-connector-odbc-3.1.11-win64.msi )
everything works well, for both field types: DOUBLE and DECIMAL.
But with this ODBC connector versions: 3.1.20, 3.1.19, 3.1.17 it works wrong again, as above.

e. When I change (with HeidiSQL) the type of field from DOUBLE to DECIMAL, forth and back, the problem follows: works good for DECIMAL and wrong for DOUBLE. And this is stable.

Environment:
Win10 64bit, fresh installation: locale still (and only)=English(United Kingdom) but I am in Poland.
mariadb-10.11.6-winx64.msi (UTF8 chosen)
mariadb-connector-odbc-3.1.20-win64.msi
Data Source filled only with necessary, minimal options set (like tcp, user, Database, Timeouts, ConnCharSet: utf8mb4), remaining: left unchecked.

The database created years ago (could be that at locale=polish). Now I import it after total, fresh Win10 and applications reinstallation.

Collation seems to has none influence: was mainly utf8mb3_unicode_ci, all changed to utf8mb4_unicode_ci.

Have you ever seen such a tricks

Thank you very much for any help.
Regards
Robert