If users have a paste_local.ini, use that one instead of
the shipped paste.ini.
That way local changes will not interfere with the main
shipped config.
Also improve the help string (./lazyserver.sh -h) a bit.
All gmg's subcommands used to have a -cf option to set the
config file. Instead make this a gmg global option.
This means:
bin/gmg migrate -cf mediagoblin_2.ini
gets
bin/gmg -cf mediagoblin_2.ini migrate
Many (all?) gmg subcommands take a -cf option to change the
used config file. This options used to be created in each
subcommand's parse_setup. Add a helper function and use it
around.
All references to static, shipped content should go via
request.staticdirect().
So insert a TestingMiddleware in the chain, which tests for
this on all requests/reponses happening in the test suite.
It's a simple text search for a bogus usage pattern.
The name of the cookie was hardcoded in the unit tests and
while renaming the cookie this part was missed.
So now read the cookie_name from the config in the test.
This also tests for the config item existing!
From WebOb release notes:
"Accept.best_matches() is gone; use list(request.accept) or
request.accept.best_match(..) instead (applies to all Accept-*
headers) or similar with request.accept_language."
... seems that's what borked it...
* aaronw/bug601_blank_comments:
Make Comment posted! translatable.
Update english translation file.
Added a check to prevent blank messages from being posted.
Conflicts:
mediagoblin/i18n/en/LC_MESSAGES/mediagoblin.po
* aaronw/bug608_checkbox_labels:
No need to translate label name.
Add label tag to fix translation issue.
Update english translation file.
Checkbox label is being translated in the form definition -- translating it in the template is redundant and breaks <label> functionality.
Conflicts:
mediagoblin/i18n/en/LC_MESSAGES/mediagoblin.po
* remove max_age - A session cookie is better, because it's
a session thing, really.
* Call the cookie mediagoblin_csrftoken, much clearer.
* Use the SCRIPT_NAME for the path of the cookie, so that
the cookie is sent back to the right place only.
Alternatively the path= parameter could be removed, so
that it defaults to '/'.
* call the randomness function only once, instead of twice.
64 bits should be enough. If really more bits are needed,
increase the number.
* Just give the number as cookie. No point in md5 and
hexdigest in my view (those functions just make another
representation).
* getrandbits gets a bit count directly, simpler API
