On Thu, 16 Jan 2020 at 12:23:51 +0100, Paul Gevers wrote:
> It seems that
> due to the rename of the binary package, autopkgtest logic doesn't apply
> anymore. autodep8 recently acquired a new feature that enables you to
> tell autode8 what the real module name is that should be tested [0].
I think this also illustrates that the transitional package is probably
inappropriate in this case: the API of python3-watson-developer-cloud is
"you can import watson_developer_cloud and use it", but that isn't actually
true here. So the autopkgtest has correctly identified a bug?
If there were existing packages that depended on
python3-watson-developer-cloud, they would now be broken (unable to import
watson_developer_cloud despite still having the right dependency), and would
need source changes to update their imports to ibm_watson - at which point
they might as well also update their dependency to python3-ibm-watson,
making the transitional package somewhat pointless.
(However, according to dak, there are currently no such packages.)
smcv
