Failing tests merging bzr.dev to brisbane-core

Andrew Bennetts andrew.bennetts at canonical.com
Tue Mar 17 22:47:21 GMT 2009


Andrew Bennetts wrote:
[...]
> 
> Uncommenting the code you commented out makes the test_fetch_to_knit3 tests all
> pass again.

So the key question is how do you want to perform the upcasting in various
scenarios.  For example, when the source or target is remote, I don't think we
want to use InterDifferingSerializer; the existing streaming fetch logic should
Just Work fairly efficiently, and InterDifferingSerializer is currently pretty
heavily reliant on its base class, InterKnitRepo, which is certainly not
generally applicable.  That's why you got a failure;
_eliminate_revisions_not_present is not implemented on most repository formats,
nor should it be.

Also, there's a bit of duplication of logic between fetch.py's Inter1and2Helper
and InterDifferingSerializer's _fetch_batch.  It would be good to remove that
duplication.  Perhaps by having just one code path for upcasting, or am I too
optimistic?

-Andrew.




More information about the bazaar mailing list