Comment 2 for bug 44725

The way ignores work currently, as I understand them, is that .bzrignore gets read in when it is needed. It is cached internally until unlock() is next called, when the cache is cleared.

If we allow manipulating of ignore definitions through the WorkingTree, we will need to write the definitions back out when unlock() is called. What happens if .bzrignore has been edited between the locking and unlocking?