bzrlib.tests.test_import_tariff.TestImportTariffs.test_simple_local failure

We need some kind of general approach for the case where we think it's reasonable for more code to be imported when particular plugins are installed. Indeed, even more generally, some underlying bzr behaviour may be changed by plugins....

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

On 12/15/2010 7:43 PM, Martin Pool wrote:
> We need some kind of general approach for the case where we think it's
> reasonable for more code to be imported when particular plugins are
> installed. Indeed, even more generally, some underlying bzr behaviour
> may be changed by plugins....
>

That is true, but at the same time, the check is also there explicitly
to tell us that something is loading something we don't think we'll need.

In the case of bzr-guess, it wasn't making use of patience-diff for this
command. So the test *actually caught* that we wanted bzr-guess to be lazy.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk0KQU8ACgkQJdeBCYSNAAMeHwCglZuAUhP0027z5umw2lG9bXqB
ZVEAn1lMJiBb9l+rTkuPjD1wADDkOEw0
=jJl8
-----END PGP SIGNATURE-----

On 17/12/2010 3:42 AM, "John Arbash Meinel" <email address hidden> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 12/15/2010 7:43 PM, Martin Pool wrote:
>> We need some kind of general approach for the case where we think it's
>> reasonable for more code to be imported when particular plugins are
>> installed. Indeed, even more generally, some underlying bzr behaviour
>> may be changed by plugins....
>>
>
> That is true, but at the same time, the check is also there explicitly
> to tell us that something is loading something we don't think we'll need.
>
> In the case of bzr-guess, it wasn't making use of patience-diff for this
> command. So the test *actually caught* that we wanted bzr-guess to be
lazy.

Yes, indeed, it's great that it trapped it and made us think "is it really
reasonable to be forcing patiencediff to be loaded?"

If, as in this case, we don't want it loaded, it's fine, we just update the
plugin. Of course we normally want to aim for this.

But there might be some other cases where it actually is reasonable for the
plugin to change the behaviour that we would have expected to see without
it. One case that came up elsewhere is a plugin that adds information to
the output of `bzr st` and tends to break all the builtin tests for it.