Code review comment for lp:~parthm/bzr/300062-better-handling-for-invalid-ignore-pattern

I'd like to see a really clean internal behaviour + a good user experience.

I don't agree with John that the user may have a bad ignore file
already: compilation will be breaking, and it will be throwing the
native RE error at the moment, so existing, broken files are fine to
error on. The point is to error:
 - cleanly
 - helpfully
 - in a way they can correct

(and correcting a .bzrignore file is easy)

I have a strong feeling that, particularly for add and status in newly
created trees, erroring is actually more user friendly that warning:
warnings can get lost and ignored, particularly when you're dealing
with hundreds or thousands of lines of output. However, I think any
improvement is better than what we have today, so we should just do
whatever Parth wants to do here, and iterate when we find out its
wrong ;).

In terms of internal code, I'd like it if whatever approach is taken:
 - does not need handling code per-code-path-using-ignores
 - is entirely encapsulated in ignores.py
 - doesn't add a new exception handler to every unversioned file
considered by add

This is totally doable, I think, for either warnings or erroring, and
not something I expect any of us to disagree on:)

-Rob