| Age | Commit message (Collapse) | Author |
|---|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Implement JWT auth for signaling connections (hello v2)
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Joachim Bauch <bauch@struktur.de>
|
|
Signed-off-by: Joachim Bauch <bauch@struktur.de>
|
|
[Documentation] Set proper header for requesting JSON response
|
|
Signed-off-by: Aaron Segura <aaron@aaronsegura.com>
|
|
* Question is limited to 32k chars (not multibyte save, so an emoji is 3-4 chars)
* Json encoded options array is limited to 60k chars (not multibyte save, so an emoji is 3-4 chars)
* Options on creation are now validated to be strings
* Options array on creating needs to be at least 2 strings long
* Options on voting are now validated to be integers
* System messages `poll_voted` do not contain the actor anymore to be okay with the hidden result mode
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Since https://github.com/nextcloud/talk-ios/pull/736 talk-ios targets iOS 12 as minimum iOS version.
Signed-off-by: Marcel Müller <marcel-mueller@gmx.de>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Signed-off-by: Vitor Mattos <vitor@php.rio>
|
|
Implement password policy checker
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
Signed-off-by: Joas Schilling <coding@schilljs.com>
|
|
The real participant mode is specific to Talkbuchet CLI and intended
only for development. This mode does not use Talkbuchet.js at all, and
just uses the real Talk app in the Nextcloud server to join the room
and, optionally, the call with a real participant.
Although it would provide the most accurate results in most cases it is
not suitable for load tests of the server or other clients, as the use
of real participants limits the number of participants that can be
created by the system running the CLI without saturating the CPU (unless
a remote Selenium server with plenty of available sessions is used).
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
The variables need to be exposed through a function, as the CLI has to
copy the elements defined by Talkbuchet.js to the global scope when
loading it, but if a variable is later overwritten the copied one would
not reflect it.
Note that CLI does not directly use the exposed publishers, subscribers
and virtual participants; they are just exposed in case they need to be
used by custom scripts.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Siege and virtual participant modes are independent one from the other,
so at any given time the CLI only provides the global functions that
belong to the mode currently active. By default the CLI starts in siege
mode.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
In order to add support for the virtual participant mode in a following
commit the Talkbuchet wrapper class was split in a base class with the
common code and a subclass with the code specific to siege mode.
Similarly, the global helper functions were also split between the
common ones and the siege specific ones, which will only be defined when
"switchToSiegeMode()" is called (done by default).
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
In the virtual participant mode Talkbuchet can be used to add
participants to a call and thus perform load test in clients. Virtual
participants require much less resources than real participants in the
system running Talkbuchet, so it makes possible to add a lot more
participants than what would be possible with real participants.
Note, however, that a single browser can run just a single virtual
participant. Therefore, virtual participants are recommended to be used
through the CLI (the support will be introduced in a following commit).
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Automatically joining the room after getting the "hello" message is
needed only for guest users. However, sieges as guest users are not
possible, as joining the room a second time from the same browser causes
the first guest to be unregistered. Due to this the code to
automatically join the room is removed.
The method to join the room, however, is kept, as it will be used by the
"Virtual participant" mode to be introduced in a later commit.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
This makes possible to identify which connection has a problem and
further investigate it if desired.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Starting tens or hundreds of connections may take a while, so the
progress is now printed to prevent the user from wondering if it is
working or not.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
This ensures that if media can not be started again for some reason (for
example, if neither audio nor video are now enabled) the old stream will
not be kept.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
When Talkbuchet CLI is executed through the Talkbuchet runner the
browser available in the Docker image is automatically found, thus
removing the need to explicitly set it.
If Talkbuchet CLI is executed by other means, or if a remote Selenium
server is used, the browser still needs to be explicitly set.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Firefox requires asynchronous functions to be executed using
"executeAsync", and logs to be got using the BiDi protocol.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
WebDriver BiDi protocol provides bidirectional communication between a
WebDriver client and a web browser. In the original WebDriver protocol
the client sends requests, but the browser is not able to send requests
to the client on its own. Therefore, if a log was written by the browser
the browser could not notify the client, and the client will not notice
the log until it explicitly performs a new request. With the BiDi
protocol the client can subscribe to events in the browser, so the
browser will send them as soon as they happen.
A new helper class is introduced to get browser logs using the BiDi
protocol. Objects of this helper class will start their own thread to
ensure that the logs can be received as soon as they happen, as the main
Python thread could be be blocked waiting while a script is being
executed in the browser and that script can cause log messages to be
printed.
Unfortunately, Chromium does not implement yet getting the logs using
the BiDi protocol, so they still need to be got in the old way. Firefox,
on the other hand, only allows getting the logs through BiDi, so the
BiDi helper introduced here will be required to add support for Firefox.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Executing "await someFunctionCall()" as synchronous code works as
expected in Chromium (the execution will not return until the awaited
function ended). However, it does not when explicit Promises are used
(it returns as soon as the main code finished, without waiting for the
Promise to be fulfilled).
WebDriver provides an explicit way to execute asynchronous functions and
signal when they finished to return, so this is used now for all
asynchronous code.
This will also make possible to add support for Firefox, as in Firefox
asynchronous functions executed as synchronous return immediately, so
they need to be explicitly executed as asynchronous.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
|
Talkbuchet is a JavaScript script meant to be loaded and run in a
browser. However, until now that required manually copying and pasting
the script in a browser console. This commit adds a Python script that
uses Selenium to launch a new browser, load Talkbuchet in it and control
it through the Python script.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
|
