Deprecate implprefix

[goodboy]

Pertains to starting work on #116.
I'll have a corresponding PR to pytest to ensure the warning is avoided - turned out it's a trivial change.

[nicoddemus]

Awesome, great work!

[goodboy]

Note I've also put up pytest-dev/pytest#3487 as requested to get this moving!

[fschulze]

Could you add a hint to the deprecation warning on what to replace the functionality with?

[fschulze]

Would it also be possible to throw a deprecation warning if a plugin provides a hook without using the marker? That is much more important, because people have to update their plugins first before the functionality is removed.

[fschulze]

I would even say that there should be a pluggy release which only deprecates hooks without the marker before deprecating implprefix itself.

[goodboy]

Could you add a hint to the deprecation warning on what to replace the functionality with?

Would it also be possible to throw a deprecation warning if a plugin provides a hook without using the marker?

@fschulze yes definitely to both of these.

I would even say that there should be a pluggy release which only deprecates hooks without the marker before deprecating implprefix itself.

@fschulze deprecating implprefix and the allowance of non-marked hookimpl functions is effectively the same thing (since there's no other way to collect hooks outside of implprefix scanning) so I'm not sure there's much point in 2 separate deprecation periods.

[fschulze]

The difference is, that I would see DeprecationWarnings for implprefix as long as devpi supports plugins without the marker. What I would like is to let plugins see the warning that they should use the marker, but ignore the implprefix for that time. I guess I can use the warnings module to add a filter for it though.

[RonnyPfannschmidt]

@fschulze see the related pytest pr for a solution

[goodboy]

@fschulze pushed up the changes you requested. Let me know if that's good enough!

[fschulze]

@tgoodlet Thanks! Now one knows what to look for in the documentation when this deprecation pops up.

Question about Travis:
Are the pluggy tests from this PR run at all? The envs checks, docs and benchmark don't look like the tests and the rest are tests against pytest.
Maybe this PR should make the deprecation warnings for implprefix into actual warnings in the tox.ini instead of errors when testing against pytest.

[goodboy]

Question about Travis:
Are the pluggy tests from this PR run at all? The envs checks, docs and benchmark don't look like the tests and the rest are tests against pytest.

@fschulze Not sure what you mean. The tests are run with different versions of pytest (integration sanity checking) but the pluggy tests are what's run; check the CI job logs.

Maybe this PR should make the deprecation warnings for implprefix into actual warnings in the tox.ini instead of errors when testing against pytest.

I actually think the error on warnings is good because it shows where I haven't fixed up the tests yet. We should only be getting warnings where the tests are written to expect them.

[goodboy]

Just waiting on CI after fixing up warning expectations in the tests then I'll merge as long as there's no more concerns 👍