[Spread-users] problem re-using private names with 3.17.0

[Guido van Rossum]

> I just want to make one thing clear.
> 
> If you disconnect and immediately connect with the same name, it is
> very likely to get the not-unique rejection and this is perfectly
> normal and is the intended outcome (and not a bug).

Yes, we know this.  Our test catches the error, sleeps for 0.1 second,
and tries again, up to 20 times.  That used to be enough.  Has the
timing of this changed?

> Upon disconnection, Spread passes a safe message to let everyone knows
> that this guy disconnected and only upon delivery of this message it
> actually clears this guy out. In the meantime, connecting with the
> same name will be rejected. Usually this should take a small amount of
> time, very likely less than a second in local area networks.
> 
> From the previous description I understand it takes sometimes over 5
> minutes on 3.17.0 which means to me that probably Spread never
> clears the name at some situations. If this is correct, it is
> absolutely a bug.

Glad you agree. :-)

>    Cheers,
> 
>    :) Yair.
>    
> Tim> [Yair Amir, on Jeremy's connect/disconnect/connect woes]
> >> ...
> >> That was my first reaction to Jeremy's message.
> >> However, reading it more carefully, you see that he means
> >> that Spread does not allow him to use the same private name
> >> AFTER the previous process that used the name already disconnected.
> >>
> >> If this is true, it is definitely a bug. I know Jonathan
> >> is looking into this.
> 
> Tim> Great.  In context, we have a Spread-based app with a large
> Tim> test suite, and the setup and teardown methods for the
> Tim> individual tests routinely reuse the same set of private names,
> Tim> establishing a new connection(s) in the setup and disconnecting
> Tim> in the teardown.  Other parts of the tests rely on those names.
> Tim> The tests all passed under 3.16.2, but started suffering
> Tim> massive "Connection rejected -- name not unique" failures when
> Tim> Jeremy tried 3.17.0.  I'm surprised that he can still write C
> Tim> code <wink>.

--Guido van Rossum (home page: http://www.python.org/~guido/)