| Commit message (Collapse) | Author | Age | Files | Lines |
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
* Rework the [webservice] section to be more useful to restish.
* Get rid of the resource_type_link
* Add and test some helpers.
* Move the root resources to a different module.
|
| |
|
|
|
|
| |
stuff around. We no longer need APIValueError or IResolvePathNames. Also,
refactor the creation of the REST server so that it could be used with other
WSGI frameworks.
|
| |
|
|
|
|
| |
RESTful. We have to use httplib2 to get DELETE and PUT.
This is a prototype and could still use much cleanup.
|
| |
|
|
|
| |
* Put the domain name on the BadDomainSpecificationError
* Give the IListManager an __iter__()
|
| |
|
|
|
| |
REST architecture, at the expense of a few features, and less support. So far
so good though.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
multiversioned web services. In concept, that's a great idea, but in (current
as of 0.9.18) practice it sucks because it /forces/ us to adopt multiversions
when really I could care less. After gobs of painful experimentation, this
passes all the tests. I guess that means it's right <wink>.
Another, more reasonable fix is to re-order when logging is initialized. This
is moved to the start of initialize_2() because in the test environment, we
don't want to initialize the loggers until after the test configuration has
been pushed. Otherwise, because of the changes to support the FHS, logging
will go to the wrong place. This really only affects tests, since in
operational mode, initialize always happens immediately one after another.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
* Give IMailingList a .domain attribute which looks up the IDomain for its
.host_name. This cleans up a lot of code.
* Add a test for the 'confirm' email command.
* Suppress blank lines in email command responses.
* Make the IDomainCollection a utility.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
means commit() is now transaction.commit() in doctests.
Add tests of extended domain creation, which exposes the need to unlock the
database (via transaction.abort()) so that the foreground test process doesn't
lock the background REST server process.
|
| |
|
|
|
|
|
|
| |
Add an IDomainManager and a global domain manager which can be gotten by
adapting the global config object.
Add an IDomainCollection interface for exposing the domain manager onto the
API.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add get_mailing_lists() which is used just for the web interface. Because of
a bug in lazr.restful, this cannot be a generator.
Similar change in IDomainSet.
Instrument IListManager to be vended through the api.
The REST server must be run in a separate process since SQLite does not like
objects created in one thread to be used in another thread. Note that this
breaks the domain.txt test, but domains really need to be in the database
anyway.
|
| |
|
|
|
|
|
| |
* Get rid of in-Python adapter registration in favor of ZCML. Eventually, I'd
like to get rid of the ZCML.
* Set the view_permission to None, but it's still not right.
* Add IDomainSet and an adapter from Configuration to IDomainSet.
|
| |
|
|
|
|
|
|
|
|
|
| |
This lets you run a bit of Python at various stages of the initialization
process.
Move the warnings filters to the top mailman package __init__.py so you always
get it. lazr.restful is too chatty.
The environment variable $MAILMAN_CONFIG_FILE now overrides -C. Hmm, I think
we really want it the original way.
|
| |
|
|
|
|
| |
* Fix a typo in the runner
* Add support for 404 Not Found exceptions
* Suppress sha module deprecation warnings from lazr.restful
|
| | |
|
| |
|
|
| |
useful information. It's GET only atm.
|
| | |
|
|
|
* Use the source version of lazr.restful until it's cheeseshopped.
* IHasGet -> IResolvePathNames
* path_override -> None; we don't need this
* Simplify publication. We don't need anything fancy, except that what sucks
is that it seems like we need to unwrap the security proxy in callObject()
* Adaptation to IAbsoluteURL still doesn't work right.
|