- Don't check the protocol in `checkFileTypes`, it should be doing one
thing only, and that is check whether the URL ends with a particular
file extension.
- Raise an error when a URI object can't be created from the passed in URL
Adds new function `isAllowedProtocolForMedia` which checks whether the
URL points to a file on the file system (`file:`), is in a Chrome
extension or uses HTTPs.
Use that in `shouldRenderMediaFromURL` to filter out URLs that shouldn't
be rendered.
Re-add utility methods to the `u` object so that 3rd party plugins can
use them.
Render the form based on `api.settings` instead of its own model.
When the login form is submitted, save the JID, password and connection
URL to `api.settings`.
Set the `service` on the Strophe connection object just before
connecting for the first time, otherwise a user supplied URL (via the
login form) is never used.
New API setting: show_connection_url_input
Add test case for incoming OMEMO message corrections.
The correction was being ignored because the parsed `msgid` of an
incoming correction was set to the `msgid` of the message being
replaced.
For error messages we still use the `message` attribute, since error
messages generally don't have a body, and if one does, it likely refers
to the `body` of a rejected message that the error refers to.
We're still setting both `body` and `message` attributes, but usage of
`message` for a normal `chat`, `groupchat` or `headline` stanza should
be considered deprecated.
Add the ability to send OMEMO corrections.
Refactor how OMEMO messages are sent to avoid having to override
`sendMessage` and thereby also allowing corrections of OMEMO messages to
be sent out.
Add two new hooks.
- getOutgoingMessageAttributes
- createMessageStanza
to be fetched from the server before triggering OMEMOInitialized.
For some contacts, the IQ to fetch the device list never receives a
response. IQ stanzas take 20 seconds to timeout, which means that all
OMEMO operations are blocked for 20 seconds (because everything waits
for `OMEMOInitialized`).
Create a new API method `api.omemo.devicelists.get` and use that to
fetch and `await` for any devicelist. That way we lazily wait for
devicelists to be fetched from the server and can continue with other
OMEMO operations unrelated to users who's clients don't respond to
devicelist queries.
- Clear timer when a messages changes from epehemeral to non-ephemeral
- Set MUC occupant on `groupchat` message when `type` changes to `groupchat` (from `error`)
- Set roster contact on `chat` message when `type` changes to `chat` (from `error`)
Thanks @afriedmanGlacier
Promises only get recreated upon logout, not reconnection.
However OMEMO was getting re-initialized on reconnection and
`_converse.omemo_store` was also deleted.
This caused a race condition where an incoming MAM message would cause
`parseEncryptedMessage` that then throws an AttributeError because
`_converse.omemo_store` is undefined because OMEMO isn't yet
initialized.
Waiting for the `OMEMOInitialized` promise doesn't help because it's
still the old (already resolved) promise from before the reconnection.
When calling `getDeviceList`, wait for the devices to be fetched
Otherwise a race condition might occur, whereby a new device gets
created in the collection, and then removed again as the collection is
replaced with the values fetched from the browser-storage cache.
Also created `converse-omemo-fingerprints` component to asynchronously
render fingerprints in the user details modal. Was done as part of this
commit because due to `getDeviceList` being async, the relevant test for
the modal were also failing
before parsing message stanza for encryption parameters.
Otherwise we might not know what our own device-id/sid is, and therefore
can't decrypt the incoming message.
Fixes#2733
As mentioned in the XEP, don't show error messages for OMEMO messages
that can't be decrypted because they were already decrypted before or
because they weren't encrypted for this device.
This let's us populate the `from_real_jid` attribute for messages in
cases where the user's nickname has changed.
Only save the occupant-id if the MUC supports it
Store all advertised features on the `chatbox.features` model.
This allows us to look up a feature without using the async
`disco.supports` API.
Updates #2241
- Unregister nickname before sending unavailable presence
- Send unavailable presence before destroying state
- Call `leave` after unregistering, otherwise the disco entry for the MUC gets removed in `leave` while it's still necessary to unregister
- Wrap `this.leave` in try/except in `onConnectionStatusChanged` handler
- Add new MUC connection status, `CLOSING` to avoid `this.leave()` being called when `hidden` gets set to `true` while the MUC is in the process of being closed.
I noticed that a marker was sent for an error message (created due to an
unencryptable OMEMO messages), and this caused the user to be kicked
from the MUC.
This allows for more flexibility in configuring which media URLs will
automatically render and which media URLs may be manually rendered by
the user (via the message actions dropdown).
For example, suppose you want to automatically render all media URLs
from https://xmpp.org, but still allow other media (which won't render
by default) to be rendered manually by the user (by clicking the "Show
URL previews" message dropdown action).
In this case, you set `render_media` to `['xmpp.org']` and
`allowed_image_domains` to `null` or `undefined`.
Or if you want to automatically render images from xmpp.org, and
restrict the domains users might manually click to render, you can add
those extra domains to `allowed_image_domains`.
It's now necessary to add a `converse-root` element in the DOM where you
want Converse to render (previously it was any element with the id
`#conversejs`).
Also, turned `converse-chats` element into a Lit element and re-render
`converse-root` and `converse-chats` when the `view-mode` or `singleton`
settings change. This is a step towards being able to change the view
mode on the fly and have the entire chat re-render appropriately.
Fixes#2647
even if the global configuration is to disallow it.
* When parsing, include all media URLs, not just the ones from allowed domains.
That makes it possible to change allowed domains on-the-fly,
while still allowing media in individual messages to be shown manually
(via the message actions dropdown).
* Merge `embed_audio`, `embed_video` and `show_images_inline` into `render_media`
* Create new config settings for allowable domains for images, video and audio
* Check the URL domain against a whitelist for the message actions dropdown
Add caps element to the MUC join presence, so that the MUC MEP node can
know whether we're interested in receiving MEP messages.
Create info messages for any `conference-info` tags that contain `activity` tags.
Check for both `headline` and `normal` MEP messages (even though the XEP
only show `headline` examples), since `normal` messages can be archived
in MAM, but `headline` ones not.
Update the XEP-0372 reference-parsing code to take the `anchor`
attribute into consideration, specifically to check which text element
the reference applies to.
Add support for rendering XEP-0372 mentions in "info" messages and for
triggering HTML5 Desktop notifications for such mentions.
Background:
-----------
XEP-0316 describes a way for a MUC to send out PEP-like messages to MUC
participants. This feature can be used to describe custom activity happening
in the MUC.
The issue was that batched writing was incorrectly also being applied
for sessionStorage stores, so when the `active` flag in
`_converse.session` was being set to `false` on window unload, the
change wasn't persisted before the window was unloaded.
This caused a new session to be created upon reload, thereby losing the
SMACKS data.
We already solved this for persistent stores by flushing them, and
doing so for the session stores would also work, but session stores
don't have to be batched in the first place, so we turn that off.
It's the only way I could get a failing test (due to the Jasmine upgrade AFAIK) to pass.
I don't understand why this happens, given that the promise has a
`catch` clause in `retractOwnMessage`, but for some reason the promise
rejection gets caught by Jasmine, causing the test to fail.
The disco entities collection gets recreated and repopulated every time
the connection is re-established or after the user logs in again after
having logged out.
The old disco entities weren't being removed, due to an erroneous
`shouldClearCache` call (not applicable to sessionStorage data).
In addition we need a handler to fire and clear the entities cache when
Converse is going to restart.
This prevents it from being persisted across page loads and makes more
sense logically.
Also move markScrolled to utils and MUC unread messages indicator to bottom panel.
If set to a positive integer, the chat history will be kept to that size
by removing older messages.
This happens as new messages come in (as long as the chat isn't scrolled up)
and when the user scrolls down.
Also add the `pruning_behavior` setting
More work on making bottom panel sub-components declarative
- Handle auto-completion in the converse-muc-message-form element
- Make message limit indicator a component
- Rename template
This causes an empty bookmarks list in the controlbox, because the
collection gets cleared (and then asynchronousy repopulated) while the
`bookmarksInitialized` promise stays `true` (resulting in a race
condition).
Alternative fix could be to clear the bookmarks and recreate the promise
before reconnecting but doesn't seem necessary to do all that work.
This required triggering the `converse-loaded` event in the entry.js
file, which means it won't be triggered for `@converse/headless` when
used in isolation.
Not ideal, but probably ok because consumers of `@converse/headless`
should probably import it into their own project in any case.
Had to make various other changes due to incompatibilities.
- Use the new `lit` package instead of `lit-html` or `lit-element`
- Drop `haunted` since it breaks the rules by specifying `type: module`
but then doesn't import with file extensions
- Use Sass Dart instead of node-sass (fixes#2445)
- Upgrade Karma
- Update to the latest @converse/skeletor (which uses mergebounce to batch writes)
- Flush storage before logging out
- Flush when reloading the tab
- Create initStorage method
when creating occupant from affiliation change stanza.
This means that when a presence for that occupant arrives, a duplicate
occupant is created (presumably because the original can't be find due
to the wrong nick being set).
- Move top-level functinos to utils files
- Get rid of unused/obsolete overrides
- Only import components there where they're used
- Bump version in package.json
