launchpad-dev team mailing list archive

[Gary Poster <gary.poster@xxxxxxxxxxxxx>]

On Aug 18, 2010, at 4:03 PM, Guilherme Salgado wrote:

> On Thu, 2010-08-19 at 07:25 +1200, Tim Penhey wrote:
>> On Thu, 19 Aug 2010 07:08:30 Francis J. Lacoste wrote:
>>> Hi,
>>> 
>>> I agree with Salgado, I don't think removing that checks is the right
>>> solution.
>>> 
>>> And I don't really think there is a problem either. create_view and
>>> create_initialized_view have a layer parameter. Why don't you use that
>>> parametere?
>> 
>> It isn't just in unit tests, but in page tests too.
>> 
>> the TALES expression branch/fmt:link/+edit
>> 
>> gets converted to:
>>  canonical_url(branch, view_name="+edit")
>> 
>> And we do use that type of TALES expression all over the place.
>> 
>> If that is on a bug page, then the current request implements the BugLayer.  
>> With my new restrictive change, the +edit view isn't defined for the BugLayer, 
>> so it blows up.
>> 
>> This is the guts of the problem, not the unit tests.
> 
> I see.  In that case, I'd be in favour of your second idea, or maybe
> even a simple mapping of rootsites to layers.  Mostly because even
> though the view check in canonical_url() is not doing the correct thing,
> it still catches 404 links.

Your second idea sounds fine.

I was hoping there was already some easy reasonable path to get what you need, but yeah, I dug around, and I don't see one.

Adding the info to allvhosts would fit into one structure we have already.  That's the most attractive thing I see from my perspective.  (I'd be tempted to make the allvhosts data structure become utility registrations eventually, but that's another discussion.)

Gary