From a user forum post:
I've noticed some interesting behavior when Liquibase compute MD5 sum of a changeset
It perform some kind of space trimming but it treat not only leading and trailing spaces, it checks also the spaces in the middle of a string: Liquibase delete duplicate spaces so the strings
will give us equal checksums.
This is perfect except one thing: if there is a spaces within a string literal, for example the following two statements
will give us the equal checksums, but in this case this is incorrect in my opinion