Sending GET request to the main RPC base URL and expecting HTTP response
code 400 had two flaws:
1. It was not verifying actual JSON-RPC interface availability
2. It did not allow implementing support for new requests
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Acked-by: Jo-Philipp Wich <jo@mein.io>
Since we're already have LuCI.session.getID() we should mirror the same
interface for retrieving the session token.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Extend the LuCI bootstrap procedure to scan for class files in
/www/luci-static/preload/. Any JavaScript file found there will be
required automatically before setting up the view, allowing to stage
code that should run on every page load.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
The LuCI.fspath() function allows constructing absolute filesystem paths
from path segments relative to the document root.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
- Make builtin classes available via `require` to allow view code to
request external and internal classes in a consistent manner without
having to know which classes are builtin and which not
- Make base classes request any used class explicitely instead of
relying on implicitly set up L.{dom,view,Poll,Request,Class} aliases
- Consistently convert class names to lower case in JSdoc to match
the names used in `require` statements
- Deprecate L.{dom,view,Poll,Request,Class} aliases
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Currently the ubus path that provide the webserver is hardcoded to be /ubus.
Change this to make it configurable from the luci config file.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Currently LuCI can be loaded only when placed in the root of the server as the cgi-bin path are hardcoded. Change the index.html to load the cgi-bin path from the current level.
Also add a new entry in the env Object to make the cgi_base path easily accessible.
This variable will be based on the position of /cgi-bin/luci.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Use stacks keyed by class id + symbol internally to not clobber the
prototype scope pointer when repeatedly calling super() in invoked
methods.
Ref: https://github.com/openwrt/luci/issues/3316#issuecomment-558531111
Fixes: 374c23cda ("luci-base: luci.js: properly handle nested Class.super() calls")
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Introduce a unique __id__ attribute per class and use that, together with
the symbol name being looked up to, to form a unique key which is used to
store the prototype scope pointer.
This fixes cases where a call to super() invokes a procedure which is
calling super() as well on a different class or symbol.
Fixes: #3316
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Turn the "Save & Apply" button into a combo button which allows to
toggle between apply with rollback and unchecked apply modes.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Store the cached system feature flags keyed by the current session id,
this ensures that the features are refreshed on login.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
This allows for some naive cache busting when the LuCI version changes,
to avoid loading outdated classes from cache.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
- Capture stack trace in L.raise() if passed type is not an Error instance
- Use L.ui.addNotification in L.error() to render the error message
- Prevent duplicate error reporting in the ui
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
The new function allows querying the presence of certain system features
such as dnsmasq or firewall availability or the compile time features
of hostapd and wpa_supplicant.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Calling abort() on a finished xhr object will clear properties such
as status which can confuse code relying on legacy LuCI request
helper functions.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Refactor L.require() to use L.raise() instead of L.error() to signal
class loading failures. This allows callers to handle class loading
errors in a graceful manner.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
