* Change stage view and use newly reducedUImainToolbarButtons config to show different custom buttons as main toolbar buttons for when web is in reduced UI.
This partially reverts commit 02787b1 to avoid looping between request subtitles and start recording which results not closing the start recording dialog.
Addresses multiple issues when enabling/disabling PiP dynamically:
1. External API: Replace dead config-overwrite event handler with
interception in executeCommand. This properly manages the intersection
observer and PiP state when pip config changes via overwriteConfig.
2. PiPVideoElement: Fix ref access pattern - access videoRef.current
inside useEffects instead of capturing at render time. The captured
value was null on first render, causing blur/focus listeners to never
be set up when component mounted into a stable conference.
3. useCanvasAvatar: Return streamRef object instead of refs.current.stream
so consumers can access .current inside their effects. The stream is
created in an effect and wasn't available at render time.
4. Add on-mount focus check with loadedmetadata wait to handle PiP enable
while app is in background, ensuring video source is ready before
attempting to enter PiP mode.
Implements Picture-in-Picture functionality for the Electron wrapper to maintain video engagement when users are not actively focused on the conference window. This feature addresses the need to keep users visually connected to the conference even when multitasking.
Key features:
- Automatic PiP mode activation and deactivation based on user interaction
- Displays large video participant's stream or renders their avatar on canvas when video unavailable
- Provides audio/video mute controls via MediaSession API directly in PiP window
- Adds API events (_pip-requested) for Electron wrapper integration
Implementation includes new pip feature module with Redux architecture, canvas-based avatar rendering with custom backgrounds support, and integration with existing mute/unmute logic. Depends on jitsi-meet-electron-sdk#479 for proper user gesture handling in Electron.
* Change toolbar background color from IFrame API #16468 fixed
* fix(toolbar #16468): implement toolbar background color via configOverwrite for web and mobile
* keep toolbarConfig defaults commented in config.js
* add trailing comma to commented toolbarConfig.backgroundColor
* fix: resolve linting errors
* feat(api): add toolbarVisibilityChanged event to IFrame API
* fix lint
* fix(recording): allow samesite iframe embeds to work with local recording
Skip capture handle validation when inside an iframe to ensure local
recording works. This only applies if the iframe is served from the
same domain.
* fix(recording): add missing line breaks for better readability in LocalRecordingManager
* test: Add iframe API expectation.
* test: Add expectations for recording and live streaming.
* test: Remove iframe references from jaas/.
* test: Add a transcription expectation.
Make sure we execute before prosody cleans it up from the list of room. If we try to look it up after that we will not find it. If we also add at 0 we cannot guarantee the order of hook execution.
* Change toolbar background color from IFrame API #16468 fixed
* fix(toolbar #16468): implement toolbar background color via configOverwrite for web and mobile
* keep toolbarConfig defaults commented in config.js
* add trailing comma to commented toolbarConfig.backgroundColor
* fix: resolve linting errors
Fixes#16468
* fix(tests): Avoids clicking UI buttons to avoid being blocked by notification.
In AV moderation tests sometimes clicking mute/unmute buttons is blocked by askedToUnmute notification.
* squash: fix waiting for button.
* squash: adds some docs.
In the `process_set_affiliation` function, an undefined `session` variable was used when revoking moderator privileges. This prevented the `jitsi_meet_context_features` from being cleared for the occupant.
Introduces a comprehensive disableChat config option that disables the entire chat feature including button visibility, notifications, sounds, private messages, and keyboard shortcuts. When disabled, the chat tab is hidden from the chat panel while allowing other tabs (polls, files, CC) to remain accessible.
Adds dedicated buttons for polls and file sharing in the toolbar overflow menu, following the pattern of the CC button. Both buttons open the chat panel with their respective tab selected when clicked.
When a user joins a very large call with SS, sometime SS is not auto-pinned to stage. This may happen when lot of participant joins are processed at the same time and therefore the state for remoteScreenShares may not get updated in time. Added extra logging to help debug if this issue reproduces.
- Fix missing Chinese translations in main-zh-CN.json and main-zh-TW.json
- Fix language selection not persisting for hyphenated locales (zh-CN, zh-TW, es-US, fr-CA, pt-BR)
- Update normalizeCurrentLanguage to check exact match before normalization
Moves the querySelector/querySelectorAll polyfill implementation from the local codebase to the @jitsi/js-utils package (v2.6.1) to reduce code duplication and make the polyfill available to other projects. This centralizes maintenance and ensures consistent behavior across the Jitsi ecosystem.
* fix(large-video)pin prev speaker on stage when local user is dominant speaker.
Also fix the case where the dominant speaker sometimes doesn't show up in the filmstrip when SS is pinned to stage.
* fix(test) Add more tests for activeSpeaker
* fix(test) Do not check for large-video if there is no remote video track.
* fix(filmstrip) Get updated display name for calc speaker order
To avoid trying to click too quickly while animations are still rendered. Avoids: Can't call elementClick on element with selector "//input[@name="lockKey"]" because element wasn't found
I suspect some intermittent test failures are caused by not waiting for
the codec change to complete. Might be exacerbated by
ensureThreeParticipants only waiting for 1 remote stream, which means
it the "ensureTwo(); ensureThree()" call may return before p2 sees p3.
Integrates file sharing into the chat interface so uploaded files appear as messages in the chat timeline alongside text
messages.
Changes:
- Created FileMessage component for inline file display in chat
- Extracted FileItem component for reusable file UI across chat and file sharing tab
- Show "A file was deleted" placeholder instead of removing message when file deleted
- Hide message menu (3-dot) when no actions are available for file messages
- Add button backgrounds in chat context to hide text on hover
- Fix timing: local participant only sees file message after upload completes (progress: 100%)
Technical implementation:
- Added fileMetadata field to IMessage interface
- Added isDeleted flag to IFileMetadata for soft-delete state
- Middleware dispatches addMessage when files uploaded (ADD_FILE action)
- Middleware uses editMessage when files deleted to preserve chat history
- Minimal state retention (only isDeleted flag) for deleted files
This provides a unified messaging experience where file sharing is part of the conversation flow.
This PR addresses a sporadic issue where cp would fail with a "directory not found" error during file operations. Replaced cp with ditto, which handles directory copying more reliably on macOS and resolves the random failures observed.
Large video was being updated through scheduleLargeVideoUpdate even when
the large video container was hidden via CSS. This occurred in multiple
layout modes: tile view, stage filmstrip (with 2+ participants), and
etherpad editing. These updates caused expensive operations including
setting video streams, managing track listeners, updating avatars, and
running show/hide animations - all wasted CPU cycles since the container
wasn't visible.
The fix introduces a centralized shouldHideLargeVideo() function that checks
all cases where the large video container is hidden. This function is used in
selectParticipantInLargeVideo() to guard to not update the participant id.
A state listener monitors transitions from hidden to visible states and ensures
the large video participant id is properly updated when the container becomes
visible again and set to undefined when large video is hidden.
This improves performance by eliminating unnecessary video element manipulation
and handler execution across all layout modes where large video is not displayed.
* fix(prosody): Avoid using stale room instances.
In very rare cases a participant can request a room and jicofo join there, but the participant don't show up (waiting for host) so jicofo leaves and in the mean time if someone tries to use the room instance just before and after the room is being destroyed, strange things can occur like web connected and joined to a stale room where nothing is received exchanged compared to the live meeting room.
* squash: Revert meeting-id one, will fix it in the problem place where there is an async call.
* squash: Change to a simple check.
* Add a sample "expectations" config.
* feat: Add configurable expectation for dial in.
* Add JaaS unauthenticatedJoins expectation.
* test: Move grantModerator to moderation/, add expectation.
* test: Move kick test to moderation/, fix p2p enabled case.
* test: Add a test case for non-moderator kick.
* ref: Move the jaas util out of specs/.
* ref: Extract a more generic joinMuc utility.
* ref: Rename joinMuc to joinJaasMuc.
* ref: Move tileView.spec.ts out of 2way, use joinMuc.
* ref: Enforce that "name" is p1, p2, p3, p4 using types.
* fix: Fix mute test filename.
* ref: Split the chat test into jaas and iframe tests.
* test: Add webhook verification to jaas visitor tests.
* ref: Remove the iframe/visitors test (ported to jaas).
* ref: Move the transcriptions test to jaas.
* ref: Make getEndpointId work from outside the iframe.
* ref: Remove TestProperties.useIFrameApi. Use the flag in IParticipantOptions instead.
* ref: Do not set a special tenant when the iFrame API is used, leave it to tests to determine.
* ref: Remove the jaas-specific tests from iframe/participantsPresnce (will be re-added under jaas/ later).
* ref: Move the dial in/out tests to jaas/.
* Add tests for jaas join/leave webhooks (port back from iframe/participantsPresence).
* config: Fallback to IFRAME_TENANT and JWT_* for jaas configuration.
* ref: Simplify boolean expression.
* ref: Remove the skipFirstModerator option (unused).
* ref: Do not override token if specified.
* fix: Do not generate token for alone/invite test.
* ref: Extract more dial-in utilities.
* test: Verify Invite UI in jaas.
* Do not generate token for dial in (case covered in jaas/).
* ref: Remove preferGenerateToken (unused).
* ref: Move mute utils in their own helper.
* fix: Fix setting the jaas enabled flag.
* Do not run alone/invite for jaas (temp fix).
* fix: Switch back to meeting window.
* Do not run alone/dialInAudio on jaas.
* Disable the SIP jibri test (broken).
When opening the connection to a new visitor node we send all the messages in current history to populate that new history and newly joined visitors will see the messages from the main room.
* feat(lang): Complete French and Canadian French translations
- Added 91 missing French translation keys to main-fr.json
- Updated Canadian French (main-fr-CA.json) with complete translation coverage
- Applied authentic Canadian French terminology (réunion vs conférence, É.-U. vs États-Unis)
- Removed 44 legacy keys from Canadian French to match English/French structure
- All files now have identical key coverage (1,469 keys each)
- Maintains regional linguistic preferences while ensuring 100% feature coverage
Fixes text truncation issue in participants pane footer context menu
items (the three-dot menu). Menu items now wrap naturally to multiple
lines instead of being truncated mid-word, improving readability for
languages with longer text strings like French.
The fix uses standard CSS properties (whiteSpace, wordBreak,
overflowWrap) without browser-specific prefixes. It is specific to
the footer context menu in the participants pane and does not affect
other context menus.
* feat(conference): Process unauthenticated access disabled error.
Shows notification with a button to login.
* squash: Fix texts.
* feat(visitors): Propagate and use allowUnauthenticatedAccess.
* squash: Avoids always sending a value, even when not set.
* squash: Rename error.
* squash: Fix comments.
* squash: Move check before log.
* feat(i18n): Complete Bulgarian translation with missing keys
- Added 587+ missing Bulgarian translations for all untranslated keys
- Achieved 100% translation coverage (1,469/1,469 keys)
- Maintained consistency with existing Bulgarian terminology and tone
- Removed 34 orphaned keys that weren't present in English version
- Applied proper 4-space indentation formatting
- Sorted all keys alphabetically to match project standards
- Created comprehensive professional Bulgarian localization
---------
Co-authored-by: Дамян Минков <damencho@jitsi.org>
Co-authored-by: bgrozev <boris@jitsi.org>
* squash: Renames module.
* squash: Loads polls component.
* squash: Attach needed logic when components/hosts load.
* squash: Moves to use component.
* squash: Uses json-message format with types.
* squash: Checks for polls support.
* squash: Fixes comments and moves validate polls to backend.
* squash: Fix debian build.
* fix(polls): Fixes polls in breakout rooms.
* squash: Further simplify types.
Separate type that needs to go into ljm and those used only for the UI part.
Simplify answer/voter type to be unified across operations which simplifies and its logic.
* squash: Change voters structure to be {id, name}.
* squash: Update react/features/conference/functions.any.ts
Co-authored-by: Saúl Ibarra Corretgé <saghul@jitsi.org>
* squash: Drops roomJid from messages. Uses the connection information as breakout does.
---------
Co-authored-by: Saúl Ibarra Corretgé <saghul@jitsi.org>
* feat(lobby) integrate login in lobby + configs
* fixed toolboxContainer styles, used HangupButton
* make hangup button visible by default
* use hangup button
* feat(prejoin): fixed indent, import extension
* squash: Restore back wait for owner dialog.
* squash: Drops not used state and functions.
---------
Co-authored-by: Calin-Teodor <calin.chitu@8x8.com>
Co-authored-by: damencho <damencho@jitsi.org>
* fix: Fix the tenant used for webhook proxy.
* squash: Linting, skip test if WH proxy is required but not configured.
* ref: Change visitorsLive to use the JaaS utils.
* ref: Move visitorsLive to the specs/jaas.
* squash: Fix import paths.
* fix: Use the iframe configured tenant for iframe tests.
Try to fix the error we see:
Error: waitUntil condition failed with the following reason: Command script.callFunction with id 116 (with the following parameter: {"functionDeclaration":"function anonymous(\n) {\nreturn (/* __wdio script__ */()=>typeof APP!==\"undefined\"&&APP.conference?.isJoined()/* __wdio script end__ */).apply(this, arguments);\n}","awaitPromise":true,"arguments":[],"target":{"context":"10352FFE685AC1D0503E1ECA3BFD33B2"}}) timed out
Seems like we do not wait for all checks to happen and start joining again in the middle of switching/checking.
* ref: Inline enterTileView.
* ref: Refactor tileView, remove tileView.LastN.
The "last n" cases are not related to tile view and are covered in lastN.spec.ts.
* ref: Remove redundant "skipInMeetingChecks: true".
skipInMeetingChecks is only used in ensureTwoParticipants, ensureThreeParticipants and ensureFourParticipants.
* ref: Move recording test to jaas/, more refactoring.
* ref: Rename and document switchToAPI() and switchInPage().
* ref: Move the tileView into 2way (temp).
The display name is used in messages when messages are coming from visitors or from the history. The display name is used only when the participant is not available in the meeting to get its name.
It is not available due to cross-origin or not able to start setCaptureHandleConfig in iframe.
error 1: Failed to execute 'showSaveFilePicker' on 'Window': Cross origin sub frames aren't allowed to show a file picker.
error 2: Failed to execute 'setCaptureHandleConfig' on 'MediaDevices': Can only be called from the top-level document.
* Once we started targeting SDK 35 on a device running Android 15 or higher, by default, we display edge-to-edge.
We can handle overlaps by using insets.
In cases like waiting-for-host lobby, jicofo can leave the room and rejoin later, without the room being destroyed. We need to make sure the metadata will reach jicofo on second attempt.
* fix: Fix example file.
* fix: Fix using webhook proxy with iframe and jaas tests.
* fix: Fixes detecting max users notification.
* ref: Clear _joinParticipant to not depend on participant names.
* ref: Use joinParticipant in jaas tests.
* ref: Drops JAAS_DOMAIN as BASE_URL is used.
* fix: Drops ctx from function parameters.
* ref: Drops not needed context members.
* ref: Drops extra function.
* ref: Participant.joinConference to use roomName from options.
* doc: Updates docs.
* fix: Adds roomName to joinOptions.
Make it possible to override the generated value.
Replaces hard-coded pixel values with relative rem units across UI components to improve typography responsiveness and maintainability.
Co-authored-by: Hristo Terezov <hristo@jitsi.org>
* ref: Don't use global context for local state.
* ref: Don't use global context to store the pin.
* feat: Add a test for setting passcode via settings provisioning.
* Use local state.
* Remove "data" from context.
* ref: Rename a function.
* test: Fail quick when join muc fails, assert specific errors (e.g. "token expired").
* test: Add tests for joining a JaaS MUC with different token options.
* ref: Refactor token generation and usage
* ref: Reduce usage of global context
* test: Add a maxOccupants jaas test.
Before the chat message context menu was appearing on the left if the private chat message was disabled. The fix makes the context menu appear on the left only for messages from the local partcipant which are the only messages rendered to the right (therefore the context menu have to appear on the left side). For all other messages the context menu should appear on the right side because the message is positioned on the left side.
The participant pane lost its scrolling capability when commit 2305ae85a removed the overflowY: 'auto' property from the container styles. This prevented users from scrolling through long lists of participants, breakout rooms, or visitors when the content exceeded the available height.
Additionally, context menus were being clipped on the left side due to the overflow constraints. This became apparent after the av-moderation feature added longer menu items like "Stop screen-sharing for everyone else".
Fix:
- Restore overflowY: 'auto' to enable vertical scrolling
- Add maxWidth constraint (285px) to context menus to prevent horizontal clipping
- Allow menu text to wrap to multiple lines instead of being cut off
- Add TODO comment for future portal-based implementation
This temporary solution provides both functional scrolling and fully readable context menus until a proper architectural change can be implemented to portal context menus outside the scrollable container.
- Move namespace declaration to correct location in build.gradle
- Remove deprecated package attribute from AndroidManifest.xml
- Update README with gradle plugin version requirement
- Fix Android namespace configuration for React Native SDK
These changes resolve installation and build errors when integrating
the Jitsi Meet React Native SDK into new projects.
Fixes: SDK installation failures on Android with newer Gradle versions
When a moderator joins or someone is granted moderation we update the whitelist for any media type for which moderation is enabled. The updated whitelist is sent to all the moderators including the newly joined or granted one.
Visitors with empty or undefined names now show the configured
defaultRemoteDisplayName or 'Fellow Jitster' as fallback, matching
the behavior of regular remote participants.
We were comparing if the number of waiting participants have changed with the wrong property from the state - the number of visitors. The result was that we won't update the state when the new waiting value matches the number of visitors already in the state. Most of the times this will be 0 and we would never go to 0.
Remove filtering on the receive side, because:
1. It's not applied to visitors, and should be for the "all" case
2. We don't want to strip stats-id from stanzas sent to jicofo
* JIT-14750 Do not show names to visitors
* apply review
* change name and email too
* fix: Fix filtering initial presence to vnodes.
* Also strip stats-id and identity.user.name.
* Move filtering logic to a util, filter all identity in main room
---------
Co-authored-by: Boris Grozev <boris@jitsi.org>
Adds .web suffixes to all web specific files to prevent beeing included in the native build. Before this it seems those files were included in the build but by some chance nothing was failing.
When we open a custom scheme URL before the window load event has been fired it seems that GUM prompt is not displayed after this due to Chrome bug. See more details here https://issues.chromium.org/issues/41398687.
The result in Jitsi Meet is the following:
If the user is joining a call for first time and haven't granted A/V permissions and lands on the deeplinking page we try to open the desktop app via redirect to a custom scheme URL. If the user chooses cancel and "Launch in web" we go to the prejoin screen and proceed with the initial GUM. At this point any GUM call won't display the permission prompt due to the browser bug and will go on forever making it impossible for the user to unmute camera or microphone.
* This is a huge update, mostly because we updated Gradle on the Android side, which includes a more strict bundle process for third party modules. On iOS, even though new architecture is disabled, we had to be explicit about it because of this react native update and because some updated dependencies have it enabled by default and are using turbo modules which are not available, YET, in our project.
When there is empty array cjson produces array: {} while prosody's json impl checks is it array and produces the correct value (array: []). Prosody impl is a little bit slower, but this is not a hot path and those are not huge json strings.
* Reusing existing translation string for virtual background error notification, instead of current hardcoded english value.
* Update VirtualBackgroundPreview.tsx
* Attempt to call a nil value (global 'get_visitors_room_metadata')
* make the linter happy
* more trailing whitespace + cleanup
* apply review
* use default false
Skip updating the transcribing state when the 'audio-recording-enabled'
property is not provided.
This fixes a race when a transcriber is already in the room, we'll see
it before properties are updated (sometimes) and without checking for
undefined we'd flip the local value to false.
For all intents and purposes 8x8 apps are integrating the SDK as a 3rd
party.
Yes, we are a 1st party of sorts, but that's ok because 8x8.vc allows
embedding.
We have observed terrible framerates due to the decoder getting
constantly restarted. Surprisingly, this only happens when using the
JVB, it does not happen when using P2P with Chromium as the other
endpoint.
We shall look into that from the bridge side.
At any rate, I'm disabling the HW video decoder for AV1, which means we
will be using dav1d, the software video decoder.
We do not need to keep granted permissions in separate field. We can always check the granted user-id or whether the current participant has a token (session.auth_token).
* feat(visitors): Allow participants pane button.
* feat(visitors): Do not count the local participant when in visitor mode.
* feat(visitors): Use same buttons on web and native.
* feat(visitors): Always show the visitors count.
It was shown only for the main participants.
* feat(visitors): Skips showing local in participants pane when visitor.
* feat(jwt): Delay loading of avatar.
Only load avatar when jwt has been validated server-side (after connected). The drawback is not loading the avatar on pre-join page only the first time.
* squash: fix comments.
* feat(metadata): Sends a list of main participants to jicofo.
Drops adding metadata in the form, jicofo, jigasi and client are reading both and now we send initial message before the self-presence.
* squash: Fix error and make a shallow copy of metadata before sending to jicofo.
We have observed some failed recordings which are lacking the EBML
header. The only way in which that seems plausible is if the
dataavailable event is received while processing the stop event. This is
allegedly not possible, but it's the only plausible explanation, so
let's defend against that.
Extend the timeslice back to 5s too.
* chore(deps): replace moment.js with day.js for locale handling and
pin day.js version to 1.11.13 in package-lock.json
---------
Co-authored-by: Calin-Teodor <calin.chitu@8x8.com>
Rework sync so uploading multiple files at once or several moderators
uploading files simultaneously doesn't break synchronization.
The current room metadata plugin operates on <key,value> pairs and we
were using a generic "files" key and using a nested object as our value.
Since with every operation the entire object is replaced it's easy to
get out of sync because one needs to be sure to have the full state
before overwriting it.
This is not realistic.
We'll look into making the metadata plugin more flexible in order to
support add / delete operations also on nested objects, but for the time
being the following will suffice:
Use a key prefix, so each file has en entry in the room metadata, like
so: "files.<the file ID> -> file metadata". This means that when a file
is deleted we just empty the metadata. The metadata plugin doesn't
currently support removing existing keys.
After a lot of back and forth, WebM seems to be the only option we
really have. In terms of containers and codecs, here is the rundown:
- WebM, any codec: the resulting file is not seekable
- MKV, any codec: the resulting file is not seekable
- MP4, vp9 + opus: video artifacts and audio clipping, file is seekable
- MP4, av1 + AAC: all good, but not supported on Linux :-/
MP4 looked very promising but there is no combination that leads to
something that works reliably everywhere, oh well. In addition, MP4
files can be opened with QuickTime on macOS, but not with the codec
combination we'd use, so that is somewhat a disadvantage.
So, we are back to where we started: WebM with VP8 and opus. But we need
to fix the duration in a potentially long file... the trick is to _only_
fix the duration. We can do that by inserting the right segment in the
metadata section. Something we cannot do without reading the whole file
is create cue points, but players like VLC seem to work well without
them.
Setting the UA string in Electron doesn't propagate the change to the
iframe where the meeting is loaded (🤦).
Thus make it more resilient by trying different things:
- A freshly introduced "iAmSpot" config option, similar to Jibri
- The app ID is present in the UA string, so we can test for that
- As a last-ditch effort, check if the display name is the default
"Meeting Room"
* Revert "feat(base/ui/native): Convert rem to px (#15934)"
This reverts commit 057dc0e4d2.
* Revert "fix(StageParticipantNameLabel): size"
This reverts commit a01f4468a0.
* Revert "fix(subtitles): position part1"
This reverts commit 6c6ed8d7a8.
* Revert "fix(ITypographyType): wrong type of fontSize and lineHeight props"
This reverts commit bffcc9092b.
* revert(Tokens): font sizes and line heights back to px from rem
Turns out there are many places that does not expect rem. Temporary reverting this change from commit 6fa94b0bb4. We should bring it back along with proper handling of rem everywhere.
In the 1st incarnation of local recordings we used to use VP8 as the
video encoder. Upon switching to MP4 that combiantion is not supported
for some reason, so I used VP9 instead.
Some anecdotal evidence suggests VP9 is behqaving more erratically, with
rendering errors and fixes.
Turns out Chrome also supports the Matroska container! And VP8 inside it
at that! The bonus we get from using it is that QuickTime on macOS won't
try to open it, thus avoiding some confusion with MP4 files, which it
recognizes, but cannot open due to the video codec.
Fixes an issue where StageParticipantNameLabel is smaller. This is caused because the font size and line height props are calculated to an invalid (NaN) value after we started using rem instead of px for lineHeight and fontSize in theme.
Reference: #15917
Fixes an issue where subtitles are displayed in the middle of the screen. This is caused because the bottom prop is calculated to an invalid (NaN) value after we started using rem instead of px for theme.spacing.
Reference: https://github.com/jitsi/jitsi-meet/pull/15934
Fixes an issue where subtitles are displayed in the middle of the screen. This is caused because the bottom prop is calculated to an invalid (NaN) value after we started using rem instead of px for lineHeight in theme.
Reference: https://github.com/jitsi/jitsi-meet/pull/15917
In a previous comit about accessibility we changed the fint size and line height to use rem (expressed as string) instead of numbers for px but the types for the interface were not updated.
Currently the clientWidth is not representing the window width but it is representing the available video space width since we are subtracting the width of the participants pane and chat area.
When turned on, the consent dialog won't be displayed for the users who
are already in the meeting, it will only be displayed to those who join
after the recording was started.
Flush the file after the 'stop' event is emitted, which happens _after_
the last 'dataavailable' has been emitted, and thus when the
MediaRecorder is really done.
In addition, lower the time slice as added precaution against crashes.
It's OK if we don't have any local audio track, we'll add it to the
mixer later.
The original bug / limitation that prompted the previous code no longer
applies since we always have a MediaStream (with audio tracks) which
we are recording.
Capture the tab audio, which will include all participants and sound
effects, YouTube videos, anything playing in the tab.
This requires the `suppressLocalAudioPlayback` constraint since
otherwise the shared tab won't keep playing audio.
Local audio still needs to be injected seprarately, since it's not
played back to the local user.
A given recording should only trigger a single consent request.
The mechanism to notify about recording status updates may fire multiple
times since it's tied to XMPP presence and may send updates such as when
the live stream view URL is set.
Rather than trying to handle all possible corner cases to make sure we
only show the consent dialog once, keep track of the recording session
IDs for which we _have_ asked for consent and skip the dialog in case we
have done it already.
Use the `showSaveFilePicker` File System Access API to pre-select the
file for download and stream the contents there. The browser uses a
temporary file as the buffer, thus not requiring us to buffer the
contents in memory.
Also change the container to MP4, since we have no way to fix the
seeking problem since we don't have the file in memory. Good news is
that it's supported since Chrome 126 and we can feature detect it!
Finally, add a helper `isSupprted` method which feature-detects
everything we need to make this work.
Ref: https://developer.mozilla.org/en-US/docs/Web/API/Window/showSaveFilePicker
Ref: https://groups.google.com/a/chromium.org/g/blink-dev/c/p1OMVj1FrMI/m/6FdLk7rZAQAJ
We either expose those events in LJM or live with strings since they
match standard WebRTC states, but depending on the package just for 3
events is just not right.
If the desktop picker window is closed before we load the sources, a JS error is thrown. From there the app goes into a broken state where when the screen sharing button is pressed nothing happens. Explanation:
When the error from the _onCloseModal handler is thrown we don't reach the line to call the onSourceChoose callback. The result is that we never call the callback received by setDisplayMediaRequestHandler. It seems that when this happens on subsequent gDM calls electron won't call the setDisplayMediaRequestHandler and therefore we don't display the desktop picker.
* feat(tests): Simplifies display names and participant create.
Moves token creation only when needed.
* squash: Skip webhook check of user id for guest participants.
* squash: Waits for kick reason dialog.
* squash: Simplifies by matching participant name and display name.
* squash: Drop displayname field.
* style(general) Replaced font-size fixed units with rem
* style(general) Replaced font-size fixed units with rem in the tokens
* style(general) Replaced line-height fixed units with rem
When any of the backend is used 'anonymous', 'jitsi-anonymous', 'internal_hashed', 'internal_plain', 'cyrus' and a participant becomes a moderator, because of external module or because set from jicofo we send to client with the self-presence about becoming moderator a default set of permissions which can be controlled via prosody config.
If using 'token' authentication the above applies only if there is a token and the token does not contain context.features.
- **Abstract Components** - Base classes for cross-platform components
- **Platform-Specific Components** - Separate implementations in `web/` and `native/` directories
- **Hook-based patterns** - Modern React patterns for component logic
### Testing Framework
- WebDriverIO for end-to-end testing
- Test files are located in `tests/specs/` and use page objects in `tests/pageobjects/`.
- Environment configuration via `.env` files
- Support for Chrome, Firefox, and grid testing
## Development Guidelines
### Adding New Features
1. Create feature directory under `react/features/[feature-name]/`
2. Follow the standard file structure (actionTypes, actions, reducer, etc.)
3. Register reducers and middleware using the registry pattern
4. Define TypeScript interfaces for state and props
5. Use platform-specific files for web/native differences
6. Add feature-specific logger for debugging
### Working with Existing Features
1. Check for existing `.any.ts`, `.web.ts`, `.native.ts` variants
2. Follow established action-reducer-middleware patterns
3. Use existing base utilities rather than creating new ones
4. Leverage abstract components for cross-platform logic
5. Maintain type safety across the entire state tree
### Testing
The project uses WebDriver (WebdriverIO) for end-to-end testing. Test files are located in `tests/specs/` and use page objects in `tests/pageobjects/`.
### Build System
- **Webpack** - Main build system for web bundles
- **Makefile** - Coordinates build process and asset deployment
- **Metro** - React Native bundler (configured in `metro.config.js`)
### Platform-Specific Notes
- Web builds exclude files matching `**/native/*`, `**/*.native.ts`, etc.
- Native builds exclude files matching `**/web/*`, `**/*.web.ts`, etc.
- Use `moduleSuffixes` in TypeScript config to handle platform-specific imports
- Check `tsconfig.web.json` and `tsconfig.native.json` for platform-specific exclusions
## Environment and Setup Requirements
### System Requirements
- **Node.js and npm** are required
- Development server runs at https://localhost:8080/
- Certificate errors in development are expected (self-signed certificates)
### Development Workflow
- Development server proxies to configurable target (default: https://alpha.jitsi.net)
- Hot module replacement enabled for development
- Bundle analysis available via `ANALYZE_BUNDLE=true` environment variable
- Circular dependency detection via `DETECT_CIRCULAR_DEPS=true`
## Code Quality Requirements
- All code must pass `npm run lint:ci` and `npm run tsc:ci` with 0 warnings before committing
- TypeScript strict mode enabled - avoid `any` type
- ESLint config extends `@jitsi/eslint-config`
- Prefer TypeScript for new features, convert existing JavaScript when possible
## Code Style and Standards
### Conventional Commits Format
Follow [Conventional Commits](https://www.conventionalcommits.org) with **mandatory scopes**:
```
feat(feature-name): description
fix(feature-name): description
docs(section): description
```
Available types: build, chore, ci, docs, feat, fix, perf, refactor, revert, style, test
### Feature Layout Structure
When adding new features:
```
react/features/sample/
├── actionTypes.ts
├── actions.ts
├── components/
│ ├── AnotherComponent.tsx
│ └── OneComponent.tsx
├── middleware.ts
└── reducer.ts
```
### TypeScript Requirements
- All new features must be written in TypeScript
- Convert JavaScript to TypeScript when modifying existing code
- Import middleware in `react/features/app/middlewares.{any,native,web}.js`
- Import reducers in appropriate registry files
- Avoid `index` files
### Bundle Size Management
- Bundle size limits are enforced to prevent bloat
- For increases, analyze first: `npx webpack -p --analyze-bundle`
- Open analyzer: `npx webpack-bundle-analyzer build/app-stats.json`
- Justify any dependency additions that increase bundle size
## Testing and Quality Assurance
### Tests
- End-to-end tests are defined in the tests/
- Tests run automatically for project member PRs via Jenkins
- Tests cover peer-to-peer, invites, iOS, Android, and web platforms
- Beta testing available at https://beta.meet.jit.si/
### Manual Testing Checklist
- Test with 2 participants (P2P mode)
- Test with 3+ participants (JVB mode)
- Verify audio/video in both modes
- Test mobile apps if changes affect mobile
- Check that TLS certificate chain is complete for mobile app compatibility
## Common Issues and Debugging
### P2P vs JVB Problems
- **Works with 2 participants, fails with 3+**: JVB/firewall issue, check UDP 10000
- **Works on web, fails on mobile apps**: TLS certificate chain issue, need fullchain.pem
- Use the tests from tests/ directory to verify functionality across platforms
### Development Server Issues
- Certificate warnings are normal for development (self-signed)
- Use different backend with WEBPACK_DEV_SERVER_PROXY_TARGET environment variable
- Check firewall settings if local development fails
### Configuration and Customization
- Extensive configuration options documented in handbook
- See `config.js` for client-side options
- Options marked 🚫 are not overwritable through `configOverwrite`
- Reference [Configuration Guide](https://jitsi.github.io/handbook/docs/dev-guide/dev-guide-configuration) for details
## Architecture Deep Dive
### Core Application Files
- **`./conference.js`** - Foundation for user-conference interactions (connection, joining, muting)
- **`./modules/external-api`** - External API for iframe integration and events
- **`./lang/`** - Translations in `main-[language].json` files
if ! grep -q -- '--"polls";' $PROSODY_HOST_CONFIG ;then
sed -i "s/\"polls\";/--\"polls\";/g" $PROSODY_HOST_CONFIG
PROSODY_CONFIG_PRESENT="false"
fi
# Old versions of jitsi-meet-prosody come with the extra plugin path commented out (https://github.com/jitsi/jitsi-meet/commit/e11d4d3101e5228bf956a69a9e8da73d0aee7949)
# Make sure it is uncommented, as it contains required modules.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.