- Add support for switching ephemerality after message creation
- Move more methods from ChatBox and ChatRoom to utils/stanza.js
- Rename 'ephemeral' to 'is_ephemeral' since it's a boolean
- Move headless one-on-one chat functionality into converse-chat
- Split converse-headline into converse-headlines and converse-headlines-views
- Add api in `_converse.api.chatboxes` for creating chatboxes
- Add `_converse.api.controlbox.get` method
VCards were being created before `fetch` was completed, so once fetch
was done those VCards were unset from their collection.
Add a new event and promise `VCardsInitialized` that triggers after
successful fetching and wait for it before creating VCards.
Depend on latest backbone.browserStorage which has support for IndexedDB
via localforage.
Storage operations are now asynchronous and transactional.
Bugs fixed (mostly by waiting for operations to complete):
* Rooms are now fetched asynchronously, so wait before triggering `show`
or when closing.
* Make sure chat create/update transactions complete before firing events
* Make sure chats and messages have been fetched before creating new ones.
* When doing a `fetch` with `wait: false` on a collection and then
creating a model in that collection, then once the read
operation finishes (after creating the model), the collection is emptied again.
* Patch and wait when saving.
Otherwise we have previously set attributes overriding later ones.
* Make sure api.roomviews.close returns a promise
Test fixes:
* Chats are now asynchronously returned, so we need to use `await`
* Wait for the storage transaction to complete when creating and updating messages
* Wait for all chatboxes to close
Otherwise we get sessionStorage inconsistencies due to the async nature of localforage.
* Wait for room views to close in spec/chatroom.js
In the process, remove the `closeAllChatBoxes` override in
converse-controlbox by letting the `close` method decide whether it
should be closed or not.
This fix forces Converse.js to respect auto_join_rooms and allow auto join room to occur with server-side generated random room names as required for XEP 0142 - workgroup queues
Lately we've been fetching messages only after entering the MUC, so that
we already have occupants to attach to them (due to `fetchMembers` being
called before) and thereby avoid rerenders.
I've now moved message fetching into the `initialize` method and added
missing event handlers for attaching/removing the occupant from a
message as it comes online or goes offline.
We still avoid (some) rerenders because we fetch and wait for cached
occupants before fetching cached messages and we wait for `fetchMembers`
before triggering `enteredNewRoom` which causes MAM messages to be
fetched.
We want to fetch messages from cache only once. For one-on-one chats
this is the case because messages are fetched only in the `initialize`
method.
This used to be the case for MUCs as well, until we added the option
`clear_messages_on_reconnection`.
Now we only fetch messages after the user has entered the MUC.
in which you can set user affiliations and roles.
Also, let getAffiliationList return an Error instead of `null` if you're
not allowed to fetch a particular affiliation list.
Don't remove affiliation for occupants if we weren't allowed to
fetch that particular affiliation list.
Previously, if fetching the list failed, we would return an empty array,
which would imply that the affiliation list is empty and that all
users with the given affiliation should have their affiliations removed.
Instead we now return `null` and properly account for that when setting
affiliations on existing contacts based on the returned member lists.
We don't remove the existing chatboxes and upon refetching they
become stale and get removed (by Backbone) and then the UI for
them becomes unresponsive.
Don't remove cached presences in `afterTearDown` event.
We might reconnect again and resume the session, in which case we need
that data still.
Instead, we remove cached presences when `clearSession` fires, because
then we know we're not reconnecting.
When reconnecting, we don't fetch presences from cache, we still have
them (since we no longer remove them in `afterTearDown`).
When reconnecting and not resuming, we clear the presences from cache,
since we're starting a new session.
* Explicit forwards and backwards paging
* Include upper or lower bound when calling `RSM.prototype.next` or `RSM.prototype.previous`
* Bugfix: Don't override new RSM parameters (caused infinite recursion)
If we've fetched the roster before within this session, then we don't
have to do so again, regardless of whether versioning is supported or
not, otherwise, even if we have a cached roster, we still need to get
the latest one again.
by not logging in automatically when `auto_login` is `false` and we're
using anonymous, external or prebind authentication.
For `authentication` set to `login` the situation is much more
ambiguous, since we don't have a clear distinction between wether we're
restoring a previous session (`keepalive`) or whether we're
automatically setting up a new session (`auto_login`).
So currently if *either* `keepalive` or `auto_login` is `true` and
`authentication` is set to `login`, then Converse will try to log the
user in.
even if the `_converse.bosh_session` instance does not exist.
We do this to avoid trying to reconnect with stale tokens upon
reconnection. Especially relevant for anonymous connections, but could
also be applicable when switching between websocket and BOSH connections
when reconnecting.
Also renamed `startNewBOSHSession` to `startNewPreboundBOSHSession`
It's unlikely that `keepalive` needs to be set to `false`, but there are
some edge cases where it might be useful.
Here's a breakdown of the different configurations of `keepalive` and `auto_join`:
---
* auto_login: false, keepalive: true
This is the default config. Users need to log in manually, but then
Converse will try to reconnect them when the page reloads.
* auto_login: true, keepalive: true
Also very popular configuration. Used with `credentials_url` or `prebind_url`.
* auto_login = `true`, keepalive = `false`
Not sure what this could be useful for. No currently known application.
* auto_login = `false`, keepalive = `false`
Useful when you want some kind of specific action to trigger login, but
you don't want to maintain that state across page reloads.
Also relevant when using a websocket connection with anonymous login.
With websocket it's impossible to maintain anonymous sessions across page
reload because the the session gets removed server-side as soon as the
websocket connection drops and the server won't accept a full JID with
password for an anonymous connection anyway (which could happen upon
page reload when Converse tries to login in again if `keepalive` was `true`).
triggering connectionInitialized.
That way, when listening for `connectionInitialized`, we'll know when it
fires whether we've attached to a BOSH connection or not.
If `auto_login` is false, we shouldn't log in, so we only try to log in when
it is true. Otherwise we still try to restore the BOSH session (if applicable).
- Split between `login` auth and the rest (`external`, `anonymous`, `prebind`)
- It doesn't make sense to pass along credentials when not using `login` auth so avoid that flow.
- It's not necessary to pass the `reconnecting` flag to `attemptNonPreboundSession`
Instead optimize in other ways.
Also:
- Don't fade in when `animate` is false.
- `trimChats` now gets called in `afterShown`.
- Add now event `beforeShowingChatView`
Includes some refactoring:
- Don't send an `active` chat state notification when entering a MUC
I can't think of a good reason why this might be necessary or desired.
- Move `setChatState` form the view to the model
- Remove unused method `handleChatStateNotification`
- Don't store `role` and `affiliation` for the current user on the
ChatRoom object, but instead on the ChatRoomOccupant object representing
the user.
This change reverts the significant part of 1dfdb36d20
I can't say that I understand why libsignal throws `Error: Invalid signature`
when you try to build a session for your own (sending) device, but given that
messages can only be decrypted once, I guess it isn't really necessary
to encrypt for your own device, since you already have the plaintext.
In addition I've added some error handling so that we can recover
gracefully when session building fails for a subset of devices.
In some cases, it might be desirable to clear cached chat messages once you've
reconnected to the XMPP server.
For example, if you want to prevent the chat history from getting too long or
if you want to avoid gaps in the chat history (for example due to MAM not
returning all messages since the last cached message).
If you're using OMEMO, then you probably don't want to set this setting to
``true``. OMEMO messages can be decrypted only once, so if they then
subsequently get cleared, you won't get the plaintext back.
This is done to avoid unnecessary repaints and reflows (caused when a
message has already rendered and then an occupant is created and
attached to that message, cauring a re-render).
Related to #1266
and re-enter if necessary.
This solves the problem where we "clone" a tab (e.g. middle-click) and
then restore a MUC from cache which we haven't actually entered (given
that the new tab represents a new device and session).
Also... add `await` in a test to try and fix Travis flakiness
We then use this flag to determine whether we should use the values from
sessionStorage.
This appears to fix the problem I originally tried to fix in 607d798665.
When "cloning" a tab (e.g. via middle-click), the `active` flag will be
set and we'll create a new empty user session, otherwise it'll be false
and we can re-use the user session.
This reverts commit 607d798665.
Unfortunately this doesn't solve the issue because we have a
chicken-and-egg problem due to not knowing the resource beforehand.
So what happens after this change is that we never resume XEP-0198
sessions but instead always start a new one.
Otherwise we run into a bug where two tabs with Converse.js share the
same XEP-0198 SM-ID, causing both to go into a reconnection-loop as
the XMPP server switches XEP-0198 sessions between them.
This bug is due to a distinction in how sessionStorage behaves when you
open the existing site in a new tab (e.g. middle-click or
`target="_blank"), as opposed to creating a new tab
and then opening the site in that tab.
In the latter case, the newly created sessionStorage object is empty.
In the former, the contents of sessionStorage of the current page is
copied over to the new page!
Distinguish between getting an existing entity and
creating a new one.
When creating a new one, ensure that we don't fetch from the cache.
New API method for creating a disco entity.
Only fetch messages after we have the latest room features
Otherwise we run into race conditions where MAM messages are fetched
before we know whether (updated) the room supports MAM or not.
I'm not sure how much this is an issue outside of tests, where we might
run into race conditions arising to the fact that we don't always
respond to every IQ stanza
- Fixes#1575. Height not properly set on MUC invitation list
- Add new configuration setting: `muc_mention_autocomplete_min_chars`
- Bugfix. `min_chars` setting for MUC mentions wasn't having an effect
Unfortunately this doesn't work on Firefox and there's no proper
workaround. https://github.com/philipwalton/flexbugs/issues/108
Reverts:
Revert "Bugfix. Properly insert error messages and spinner"
This reverts commit 6a419cc145.
Revert "Use flexbox to keep the chat scrolled down"
This reverts commit dd91d3cc55.
By using `display: flex` and `flex-direction: column-reverse`, the chat
now automatically scrolls down when loaded, without requiring any
extra JavaScript.
We still need to scroll down with JavaScript when sending a message.
By using `column-reverse`, the messages container now works in reverse.
So the newest message is the first element in the container and the
oldest message is the last. This is the reverse of before.
Due to this, this change will likely break some plugins.
Up and down arrows only serve as message correction shortcuts if the
textarea is empty.
This means that:
- Up arrow doesn't start correction of previous message if we're currently writing a message
- Down arrow doesn't erase currently written message
This is based on user-feedback and is also how Slack handles these
shortcuts.
When reconnecting with a new transport, we call setUserJID
so that a new resource is generated, to avoid multiple
server-side sessions with the same resource.
We also call `_proto._doDisconnect` so that connection event handlers
for the old transport are removed.
https://xmpp.org/extensions/xep-0198.html#resumption
The <resume/> element MUST include a 'previd' attribute whose value
is the SM-ID of the former stream and MUST include an 'h' attribute
that identifies the sequence number of the last handled stanza sent
over the former stream from the server to the client.
- Set roster_fetched flag when we get an empty IQ[type="result"] stanza
- Set `restored` flag when resuming in converse-smacks
- Don't trigger change event when fetching xmppstatus values
- Removed two events, since we can just listen to `change` on `_converse.xmppstatus`.
- Remove the `keepalive` configuration setting. It is now always implicitly `true`.
- Remove the `expose_rid_and_sid` configuration setting.
- A `prebind_url` is now mandatory when setting `authentication` to `prebind`.
- It's no longer possible to pass in `rid` and `sid` values to `converse.initialize.
I'm using prettierx, which is a fork of prettier so that we can have
spaces before the brackets in a function declaration.
I don't like how restrictive prettier is concerning its configuration
and formatting.
- it messes up some indentation in specs/
- it removes a second newline when I specifically want to have two
- it doesn't allow spaces before brackets in function declarations (hence prettierx)
- it adds spaces for function parameters with default values
None of this is configurable.
So auto-formatting is not turned on by default, but it might be a useful
tool in some cases.
Fixes an issue where indentation was off in converse-core.js, causing
vim-sleuth to wrongly indent by three spaces only.
* Checkmark and edit modal button are now on the right
* Instead of showing checkmarks for MUC messages,
we now show them as light grey before MUC reflection
* Simplify message markup by requiring less `if` statements
- New plugin `converse-smacks`
- New config option `enable_smacks`
- Rename session cache id from `converse.bosh-session` to `converse.session`
- Refactor logout and login as consistently used api methods
- Refactor session cache to store per JID
Fixes#316
- Take locked_muc_domain into consideration in chatroom display name
- Use getDisplayName when rendering the heading name
- Give preference to bookmark name if it exists
and also first check whether we have an occupant with that nickname.
Since roles are session based, it doesn't make any sense to try to
modify the role for a user not in the room.
- new config option `singleton`.
- new plugin `converse-uniview`
- removed `converse-embedded`.
- various CSS changes, to properly render an embedded full app
- don't re-open cached and non-autojoined chats in singleton mode
The goal here is to extend the `embedded` `view_mode` so that the full app can
also be embedded, not just a single MUC or private chat.
To do this, we'll need to differentiate between multi and singleton chat apps.
* A singleton chat app contains only a single chat.
* A multi-chat app can contain zero or more chats
So we introduce a new config option, `singleton`, which when used with
`view_mode` set to `embedded` will determine whether a single chat or the full
app is embedded.
Similarly, in `overlayed`, `fullscreen` and `mobile` view modes, `singleton`
set to true will allow only one chat within the parameters of that view mode.
We're appropriating the word `singleton` and introducing the concepts of
`uniview` and `multiview` (see a785ca8) to cover what was
previously meant with `singleton`.
updates #1297
- Use native `getTime` method instead of moment's `valueOf`
- Use toISOString() instead of moment().format()
This makes the code timezone independent since we're now globally using UTC.
Messages (received directly or via MAM) are stored in a backbone
collection that's kept sorted by the message's "time" attribute.
Various parts of the codebase rely on this behavior.
For regular messages, the time attribute is generated by
`moment().format()`, which generates a timestamp in the browser's
local timezone, for example "...+02:00".
For MAM messages, we would previously copy the "delay.stamp" value
into the "time" attribute verbatim. However, that value is generated
by the server and is unlikely to be in the same timezone locale.
ISO timestamps expressed in different timezones cannot be sorted
lexicographically, breaking the premise of keeping the messages
collection sorted by time.
* Rooms Api
number of parameters in the documentation was made 2 from 3 as api which is used to open a room with default configuration has only 2 parameters.
That way we don't lose the user's input values, we avoid flashing and we
avoid unnecessary rendering.
In the process, fixed an annoying issue where Chrome auto-completes
what it thinks is the username into the "Language Tag" field of the MUC
config form.
Instead we tell Chrome that the MUC JID is the username, thereby also
letting it save the password to to that JID.
Looks like the `webkitTransitionEnd` event doesn't fire because `fade`
is already on the modal, so there's no transition, which means that in
bootstrap.native the event doesn't get fired to remove the overlay.
- Let box_id start with char for valid HTML.
- No need to use SHA1 for box id
- No need for the user_id attribute.
- Set nickname when we set the roster contact.
Also...
- _converse.api.contacts.get is now async
- _converse.api.chats.create is now async
This changes the API method's current behavior by not automatically
maximizing (in `overlayed` view mode) or bringing a background chat into
foreground (in `fullscreen` view mode). Instead `force` needs to be set
to `true` for that to happen.
Set reference to roster contact on the message and chatbox object and
listen for changes to the nickname.
Currently, because chat boxes are fetched and initialized before the
roster, messages and chats are repainted with the correct display name
only later, causing a "flash" effect.
Ideally we would only initialize the chat boxes after the roster
contacts have been fetched, but this is currently not easily possible
because we need the control box to render before everything else.
When no status message is in a presence stanza, the value gets set to
undefined, so we might as well make this the default value to avoid
unnecessarily triggering change events.
Instead of deleting devices that are not returned in the device list,
set an `active` flag, so that we don't remove the trust setting.
Set deactivated devices to `active` when a receive an OMEMO message from
it.
Also, set omemo_supported to true when we've succesfully decrypted a
message.
When set, the "Nickname" field won't be shown in the "Add groupchat"
modal and instead the nickname will be automatically set (based on other
config settings).
Until now, we've only been explicitly documenting the `_converse.api`
namespace and only considered the methods under it as forming the API
contract (which determines how we do semver releases).
It appears as if we've reached a point where trying to keep everything
under the `_converse.api` namespace no longer makes sense. Certain
methods are applicable to particular models and trying to shoehorn them
into the `_converse.api` namespace seems clunky and non-intuitive.
I've therefore decided to slightly refactor `sendMessage` to let it take
two simple parameters and to document it with JSDoc so that it's
presented as an API method, albeit only available on a chat model.
updates #1496
updates #1504
Eventually we need to add better support for receipts and markers of MAM
messages.
We'll need to do the following:
* First check whether the MAM page already contains the receipts or markers, to avoid duplication
* Only ever in catchup mode
* Only after full catchup to ensure that the receipt is not in a different page than the message
Currently, in order for Converse.js to recognize a pasted URL as an
image, it must be an URL whose protocol is `https:` (`https` in the
URI.js library's notation). This is sensible, but means that any
non-HTTPS image URL is not recognized as a valid URL (and thus will not
be rendered inline, even if `show_images_inline` is set to `true`).
It is important to always check for HTTPS URLs when in a secure context
(i.e., the initial page load was requested via HTTPS) in order to ensure
that non-secured content does not mix with secured content. However, the
inverse is not true: if the original page was loaded over HTTP, then
enforcing HTTPS for images adds arguably no meaningful protection while
also breaking the `show_images_inline` feature for the edge cases where
Converse.js is deployed without HTTPS and a user pastes an HTTP URL.
This patch changes the behavior of the `isImageURL` method such that the
requirement for the pasted URL's protocol to be `https:` is enforced
only when the `window.location.protocol` itself is also `https:`. By
doing this, we ensure that secure origins (i.e., when Converse.js is
loaded over HTTPS initially) are still secured and cannot have non-HTTPS
content introduced to the page via a pasted non-HTTPS URL, however it
also allows non-HTTPS origins to render both HTTP and HTTPS image URLs.
As per XEP-0004, the default "type" of data form fields is
"text-single", so a missing "type" attribute should not be treated
differently.
This fixes handling of CAPTCHAs offered by ejabberd.
