Re: [Monotone-devel] Changes in nvm.basic_io.inventory

[Stephen Leake]

Thomas Keller <address@hidden> writes:

> Stephen Leake schrieb:
>> I'd like to get nvm.basic_io.inventory merged into the main line.
>> Emacs DVC is working well with the current implementation. It would be
>> good to get this into the released version.
>
> Have you been able to sort out all of the remaining issues, f.e.
> incomplete output in a directory-restricted case (renames)? 

Test cases that were mentioned in this list are in
tests/automate_inventory_path. Some of them are failing; so no, they
are not yet sorted out. I'll see if I can understand why they are
failing.

>> What remains to be done?
>> 
>> There are tests in tests/automate_inventory_path/ that are marked with
>> FIXME: I guess we should fix those, or at least change them to
>> expected failures.
>
> Well, this is the point where I'm a bit unsure - shall we replace an
> incomplete, buggy format then with another less incomplete (not
> complete), less buggy (not bugfree) format? Or is this more a "anything
> is better than the current solution"-deal?

Since it allows Emacs DVC to work, I think it is better. But perhaps
we need to hear from more people?

To recap: Emacs DVC is an Emacs front-end for distributed
configuration management systems, including bzr, monotone, git, and a
couple others. The user interface is similar to pcvs. The monotone
interface requires basic_io inventory.

I had talked about putting in options to not output known or ignored
files; I still think that's worthwhile, but there are other things in
Emacs DVC that are more urgent.

>> The format for outputing the root directory is still problematic. I
>> think the simplest solution is to simply not output it; in Emacs DVC,
>> we just drop that directory from any display.
>
> I don't think of it as a good idea. You're effectively hiding two kind
> of information by this: pivot_root-renames and attr sets/drops on the
> workspace root. 

Ah, I forgot those. So we should add test cases for them. And I see
that there is a FIXME to that effect in
tests/automate_inventory/__driver__.lua.

> So IMHO we really should find a way to display it somehow. Even if
> we display it as a non-standard item, but we should not omit it.

I think the path should be ".". And we also need to fix fs_type and
status; they are currently "none", "missing".

I'll work on that also.

-- 
-- Stephe