built on top of MR !91 (merged) that allows upgrade to recent libxml2 fixes the build of experimental providers.
This is a workaround.
The correct things to consider doing are :
- make -1LL a constant for Storage::MISSING_OBJECT_SIZE or the like
- return that instead of an exception in all providers
- make sure no provider dependent exception type leaks. the customer must not have to write new error handling code when changing cloud provider.
- potentially return stack trace in exception messages...
( current code just sent "Blob does not exist" message. to a call trying to write a new blob... )