The correct fix is to the application software which didn't
provide a base URI when building an InputSource. That'd be
some puppycrawl config loader code, I'd guess.
The patch Mark posted earlier works great for me. What do you think
when we just apply it ?
I think the two bugs in that puppycrawl code should get
fixed. Then there will be no problems at all.
- It's completely bogus to turn arbitrary warnings
into fatal errors as a default policy.
- And it's equally bogus to call an XML parser with
borked inputs, expecting it guess correctly about
the base URI to use when resolving.