So, I think Martin[gz]'s analysis is good; I'm ok with us doing what
your patch does *as long as* we don't lose motivation to fix these
other things:
- so, any comparisons doing implicit casting should error (that is,
we should say that u'foo' != 'foo'), which will need to be its own
patch [it will be huge].
- we need to *really visibly* log the unicode exception
- we need to not permit any unicode exceptions on PQM, because its
meant to only get clean things.
I think on balance, that I wouldn't do what you've done, myself - I'd
just find and fix the root causes anytime the test suite did get
aborted: there can't be many such root causes.
So, I think Martin[gz]'s analysis is good; I'm ok with us doing what
your patch does *as long as* we don't lose motivation to fix these
other things:
- so, any comparisons doing implicit casting should error (that is,
we should say that u'foo' != 'foo'), which will need to be its own
patch [it will be huge].
- we need to *really visibly* log the unicode exception
- we need to not permit any unicode exceptions on PQM, because its
meant to only get clean things.
I think on balance, that I wouldn't do what you've done, myself - I'd
just find and fix the root causes anytime the test suite did get
aborted: there can't be many such root causes.
HTH,
Rob