[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: .changes files as first-class objects



"Adam D. Barratt" <adam@adam-barratt.org.uk> writes:
> On Sun, 2010-01-17 at 22:40 -0800, Russ Allbery wrote:

>> If is_lab() returns true for a lab without the changes directory, can
>> Lintian then fix the lab on the fly to have that directory?

> We don't currently have the means to do so.  In fact, we don't have a
> means to update the structure of a lab itself, rather than the package
> data contained within it.

> So far as I can see, the last time such a change was made was in 2004,
> when udeb support was added.  At that point, what is now Lab::is_lab()
> (and was at the time still part of the frontend process) was updated to
> require the udeb top-level directory to exist; the upgrade path for
> static labs appears to have been "add the udeb directory".

I'm okay with doing it this way, personally.  It looks like if you tell
Lintian to re-setup the static lab using the existing directory, the right
thing happens.

> We could take this opportunity to add a format specification to the
> top-level lab folder to allow such cases to be handled, either by
> reusing the existing $LAB_FORMAT or separating the two format versions.
> A lab without such a format file would be assumed to need upgrading to
> the current version; either is_lab() or Lab::setup() could handle the
> upgrade, ensuring that the lab's structure is up-to-date.

> Of course, it's possible I'm over-thinking this. :-)

We could have setup() for a static lab automatically recognize this case
and create the new directory.  Adding a format is probably overkill for
just checking whether all the directories but one are present and adding
the missing one if so.

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>


Reply to: