Merge lp://qastaging/~jelmer/bzr-builder/stacked-on into lp://qastaging/bzr-builder

> stacked=(bzr_version_info >= (2, 3, 0, 'dev', 5)))

That comparison is wrong because 2.3.0.beta.5 is lexicographically lower. Sorry. (Maybe we should change our encoding to avoid this.)

You can use a more complex expression, or just assume 2.3.0.* is enough.

On Wed, 2011-01-12 at 02:51 +0000, Martin Pool wrote:
> Review: Needs Fixing
> > stacked=(bzr_version_info >= (2, 3, 0, 'dev', 5)))
>
> That comparison is wrong because 2.3.0.beta.5 is lexicographically
> lower. Sorry. (Maybe we should change our encoding to avoid this.)
Thanks, that's a good point.

> You can use a more complex expression, or just assume 2.3.0.* is enough.
I'm not sure if it's worth adding a more complex expression here, so
I've updated it to simply assume that 2.3.0 is enough.

Another alternative would have been to check for one of the private
methods that was added by the commit-on-stacked branch, but I'm worried
that will quietly degrade performance again if it ever gets removed.

Cheers,

Jelmer

On 11 January 2011 21:49, Jelmer Vernooij <email address hidden> wrote:
> On Wed, 2011-01-12 at 02:51 +0000, Martin Pool wrote:
>> Review: Needs Fixing
>> > stacked=(bzr_version_info >= (2, 3, 0, 'dev', 5)))
>>
>> That comparison is wrong because 2.3.0.beta.5 is lexicographically
>> lower.  Sorry.  (Maybe we should change our encoding to avoid this.)
> Thanks, that's a good point.
>
>> You can use a more complex expression, or just assume 2.3.0.* is enough.
> I'm not sure if it's worth adding a more complex expression here, so
> I've updated it to simply assume that 2.3.0 is enough.

Works for me.

I wonder what the easiest way in Python to introspect "does callable a
accept a parameter b" is? (There is probably no non-Turing-complete
way to check for kwargs, but for statically declared arguments it
seems possible in principle.)

> Another alternative would have been to check for one of the private
> methods that was added by the commit-on-stacked branch, but I'm worried
> that will quietly degrade performance again if it ever gets removed.

I agree, that would be poor.

--
Martin

On Wed, 2011-01-12 at 04:04 +0000, Martin Pool wrote:
> On 11 January 2011 21:49, Jelmer Vernooij <email address hidden> wrote:
> > On Wed, 2011-01-12 at 02:51 +0000, Martin Pool wrote:
> >> Review: Needs Fixing
> >> > stacked=(bzr_version_info >= (2, 3, 0, 'dev', 5)))
> >>
> >> That comparison is wrong because 2.3.0.beta.5 is lexicographically
> >> lower. Sorry. (Maybe we should change our encoding to avoid this.)
> > Thanks, that's a good point.
> >
> >> You can use a more complex expression, or just assume 2.3.0.* is enough.
> > I'm not sure if it's worth adding a more complex expression here, so
> > I've updated it to simply assume that 2.3.0 is enough.
> Works for me.
>
> I wonder what the easiest way in Python to introspect "does callable a
> accept a parameter b" is? (There is probably no non-Turing-complete
> way to check for kwargs, but for statically declared arguments it
> seems possible in principle.)
It looks like inspect.getargs() can do that, although it has some
caveats (e.g. if the function is changed to accept *args, **kwargs it
will break).

In this particular case we can't check for new arguments, as no new
arguments to public methods were added - it's just a bug fix we need to
make sure is present in bzrlib.
(https://code.launchpad.net/~jameinel/bzr/2.3-commit-to-stacked/+merge/42698)

Cheers,

Jelmer