[MERGE][Trivial] invalid version tuple attempt to raise AssertionError fails with TypeError

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Lukáš Lalinský wrote:
> On Wed, Sep 24, 2008 at 5:56 AM, Aaron Bentley <aaron at aaronbentley.com> wrote:
>> If the authors of Qbzr thinks that _get_trees_to_diff
>> is a useful function, it would be really nice if they told us so.
>> Making it public is good for other bzrlib users, and it's also good for
>> them, because then they don't have to worry that one day it will
>> suddenly break on them.
> 
> There is also another point of view, here are the two options I see:
> 
>  a) Somebody submits a patch to make it public, the function changes
> => QBzr needs to be fixed to support both versions
>  b) Somebody submits a patch in the future, and this patch changes the
> function => QBzr needs to be fixed to support both versions
> 
> As you can see, both cases have the same end result for QBzr, but case
> b) requires much less effort *now*, which is more important if you
> work on something in free time and want to spend as less time on it as
> possible.

There at least two other possibilities:
c) The method stops being used by bzr itself and is removed because it
is private.
d) Because it is private, the method's behaviour is changed in ways that
that cause QBzr to silently misbehave.

Also, you talk about supporting both versions, but if you request the
method be made public as you discover you want to use it, it may become
available quickly enough that you only need to support the public version.

Option a) may require more effort, but does put you in control-- you'll
be prepared for the change, not surprised by it.

There are no requirements for testing private symbols, so a public
symbol will tend to be less buggy.

>> And this is not counting the effort to actually submit the
>> patch to make it public. :)

That's true, but I don't think I actually said that the plugin author
had to make the change to bzrlib.  Even just mentioning you'd like it
made public would be better than nothing.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFI2jvu0F+nu1YWqI0RAnsGAJ9s/nqs62bz28zfGreVoI4c8o7KNgCfQ1FK
sRy+duiWjc2KOdUq3ZGsdiw=
=gk4b
-----END PGP SIGNATURE-----