Users have seen raw MISSING_MATRIX_RTC_TRANSPORT. Inventory every failure surface: EC widget error codes (transport missing, MatrixRTC join failure, LiveKit connect/token errors), the voice-guard refusals (room full — already "Channel Full (N/N)", audio-only policy, screenshare disallowed), clock skew (#158), permission denied (mic/camera), and the "call not supported" path. One sentence each, with the fix when there is one ("Allow microphone access in your browser's site settings"), in the existing call-status area / join button state. Raw codes stay in a tooltip and Crypto Diagnostics. Unit-test the mapping.
Users have seen raw `MISSING_MATRIX_RTC_TRANSPORT`. Inventory every failure surface: EC widget error codes (transport missing, MatrixRTC join failure, LiveKit connect/token errors), the voice-guard refusals (room full — already "Channel Full (N/N)", audio-only policy, screenshare disallowed), clock skew (#158), permission denied (mic/camera), and the "call not supported" path. One sentence each, with the fix when there is one ("Allow microphone access in your browser's site settings"), in the existing call-status area / join button state. Raw codes stay in a tooltip and Crypto Diagnostics. Unit-test the mapping.
jared
added this to the Features 2026-Q4 milestone 2026-09-17 23:06:34 -04:00
Inventory of every call-failure surface as of today (host = cinny, EC = the fork):
Surface
What the user sees
Raw code visible?
No LiveKit transport advertised (MISSING_MATRIX_RTC_TRANSPORT)
host prescreen: "Your homeserver does not support calling." (livekitSupport check, before EC even loads) — so the raw code only ever reaches users of other clients; EC's own page for it says "Call is not supported / The server is not configured to work with Lotus Call. Please contact your server admin (Domain…, Error Code…)"
code in parentheses (EC upstream style)
No WebRTC
host: "WebRTC is not supported in this browser."
no
Mic permission denied
host prescreen: "Microphone access is blocked. Enable it in your browser settings to join."
EC's own error page — every one already has a plain title + sentence (error.* in the fork's locale), with the code in small print
code in parentheses
Voice-limit guard refused the token (403 "This voice channel is full." / permission text) — the race where a second Lotus user joins as it fills, and every other client
was EC's generic page "Something went wrong (OPEN_ID_ERROR)" — the guard's sentence was dropped
fixed: fork 0.25.0-lotus.5 (d9ac9a0f) throws SFUTokenRefusedError → "Can't join this call" + the guard's sentence verbatim; unit-tested
So after lotus.5 there is no failure a user can hit that ends in a bare code; the codes that remain are the parenthesised support hints on EC's own pages, which is the existing upstream design (and where "Crypto Diagnostics" would otherwise have to be). Closing — reopen with a screenshot if anyone meets a raw code.
Inventory of every call-failure surface as of today (host = cinny, EC = the fork):
| Surface | What the user sees | Raw code visible? |
|---|---|---|
| No LiveKit transport advertised (`MISSING_MATRIX_RTC_TRANSPORT`) | host prescreen: "Your homeserver does not support calling." (`livekitSupport` check, before EC even loads) — so the raw code only ever reaches users of *other* clients; EC's own page for it says "Call is not supported / The server is not configured to work with Lotus Call. Please contact your server admin (Domain…, Error Code…)" | code in parentheses (EC upstream style) |
| No WebRTC | host: "WebRTC is not supported in this browser." | no |
| Mic permission denied | host prescreen: "Microphone access is blocked. Enable it in your browser settings to join." | no |
| Voice channel full (host-side) | host: "Channel full (N/N)" + Join hidden; sidebar second-click → toast | no |
| No power to join | host: NoPermissionMessage | no |
| Already in another call | host: "Already in another call — End the current call to join!" | no |
| EC failed to load / timed out (25 s watchdog) | host: "The call failed to load. Check your connection and try again." + Back | no |
| Admin turned off screenshare/camera mid-call | host toast (#223) | no |
| Clock skew | host banner + call bar line (#158) | no |
| LiveKit connect failure / timeout / capacity (503) / room creation restricted / connection lost / E2EE unsupported / membership manager | EC's own error page — every one already has a plain title + sentence (`error.*` in the fork's locale), with the code in small print | code in parentheses |
| **Voice-limit guard refused the token (403 "This voice channel is full." / permission text)** — the race where a second Lotus user joins as it fills, and every other client | **was** EC's generic page "Something went wrong (OPEN_ID_ERROR)" — the guard's sentence was dropped | **fixed**: fork `0.25.0-lotus.5` (`d9ac9a0f`) throws `SFUTokenRefusedError` → "Can't join this call" + the guard's sentence verbatim; unit-tested |
So after `lotus.5` there is no failure a user can hit that ends in a bare code; the codes that remain are the parenthesised support hints on EC's own pages, which is the existing upstream design (and where "Crypto Diagnostics" would otherwise have to be). Closing — reopen with a screenshot if anyone meets a raw code.
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.
Users have seen raw
MISSING_MATRIX_RTC_TRANSPORT. Inventory every failure surface: EC widget error codes (transport missing, MatrixRTC join failure, LiveKit connect/token errors), the voice-guard refusals (room full — already "Channel Full (N/N)", audio-only policy, screenshare disallowed), clock skew (#158), permission denied (mic/camera), and the "call not supported" path. One sentence each, with the fix when there is one ("Allow microphone access in your browser's site settings"), in the existing call-status area / join button state. Raw codes stay in a tooltip and Crypto Diagnostics. Unit-test the mapping.Inventory of every call-failure surface as of today (host = cinny, EC = the fork):
MISSING_MATRIX_RTC_TRANSPORT)livekitSupportcheck, before EC even loads) — so the raw code only ever reaches users of other clients; EC's own page for it says "Call is not supported / The server is not configured to work with Lotus Call. Please contact your server admin (Domain…, Error Code…)"error.*in the fork's locale), with the code in small print0.25.0-lotus.5(d9ac9a0f) throwsSFUTokenRefusedError→ "Can't join this call" + the guard's sentence verbatim; unit-testedSo after
lotus.5there is no failure a user can hit that ends in a bare code; the codes that remain are the parenthesised support hints on EC's own pages, which is the existing upstream design (and where "Crypto Diagnostics" would otherwise have to be). Closing — reopen with a screenshot if anyone meets a raw code.