A lot of work was done in v9.1 to unify how character conversion policies work across all platforms. A new policy object (AbtCodePageConversionPolicy) was added to allow the user to customize conversion rules such as whether to skip the UTF-8 BOM or not.

As part of this work, Unix’s default policy was changed to transliterate mode, which means it will try and convert characters in the origin encoding to the closest possible matching character in the target encoding. This unifies it with the policy that existed on Windows.

Before 9.1, the default policy on Unix was strict meaning an error was thrown if an invalid character sequence was detected during conversion.

To restore the old behavior of strict conversion, change the system wide policy to strict by executing the following code below.