No subject


Tue Dec 4 09:57:09 GMT 2007


never triggered the next range recognition, so I went ahead and
did that (which reveals another subtle bug in seek). Now reading
again when a range is exhausted start reading the next range
without seek being required (as it was before that).

This is already more than what is used by bzrlib so I stop here
(but the TDD exercise was worth it).

<snip/>

    Andrew> I wonder a little if we'll eventually decide to use
    Andrew> the same test parameterisation framework as for
    Andrew> e.g. repository implementations, but this is quite
    Andrew> readable so I have no qualms about landing it as is.

I didn't here fr lack of experience in using them but I was
planning to use it on my https test server branch anyway.

    Andrew> [...]
    >> +class TestRangeFilMultipleRanges(tests.TestCase, TestRangeFileMixin=
):
    >> +    """Test a RangeFile for multiple ranges."""

    Andrew> I think a little more detail in this docstring (or
    Andrew> maybe setUp's docstring) about the exact layout of
    Andrew> this RangeFile would be good; the code in this setUp
    Andrew> is just long enough that it's hard to see what
    Andrew> self._file is like at a glance.

Done.


    Andrew> Is it worth doing:

    Andrew>         elif self._range_hint =3D 'single':
    Andrew>             max_ranges =3D total
    Andrew>         else:
    Andrew>             raise AssertionError("Unknown _range_hint %r" % (se=
lf._range_hint,))

    Andrew> ?

It doesn't harm and will be covered by a test once I use
--coverage there too.

Interesting journey !

            Vincent



More information about the bazaar mailing list