python: backward compatibility issue

[sim590]

The whole benchmark is broken by 768bfa5 since the SockAddr is not used by the benchmark and no "legacy" signature (with address as string) is still available in the python API for bootstraping 😒.

This is the case for benchmark, but also all python code which have already been written before 768bfa5. For instance, the pingpong script is broken. I think that some of ring's dht small swarm is using python bindings also. This is dangerous.

This kind of changes should be made more carefuly in the future since we never know who may use the python bindings also and we don't want to just drop support for API as fundamental as DhtRunner.bootstrap. I suggest, a legacy signature should be provided. If we want to deprecate it, then we could add a message on standard error specifying the matter, but anyhow I don't think it's good idea to remove this signature before a long time (couple of weeks or months).

[aberaud]

it's a bug, this is exactly the kind of things that will be detected automatically with unit tests

[sim590]

Oh. I just noticed that you in fact kept the legacy signature available. I didn't look close enough at first. So, yeah, like you say, this is a bug 😉

[aberaud]

Fixed by 99f8c29

[sim590]

@aberaud: But now, the bootstrap signature bootstrap(string, string) is not present anymore which was in part what I was complaining about. Don't you think that this canonical signature should always be available?

[aberaud]

I think bootstrap(string, string) was always present (as you noticed above) and still is, but was not working because of a conflict with the now-removed signature.

[sim590]

@aberaud. I really have trouble reading github's diffs.