Release 1.0.21

[lollox80]

Release 1.0.21: 11 PRs merged since v1.0.20 (#150-#152, #154-#160, #162; #154 has Changelog: none). #161 is not in this release.

• manifest.json: 1.0.20 → 1.0.21.
• CHANGELOG.md: the 1.0.21 section composed with tools/release_notes.py from the PRs' Changelog: lines. No PR has a Before you update: line. The three Dependabot PRs (ci: bump actions/checkout from 4 to 7 #150-chore: bump the dev-tools group with 7 updates #152) have no Changelog: line and are filed under Other changes from their titles. The fix(log_redact): mask plant data in the logs #157 line was reworded in its PR text to match docs(troubleshooting): say which plant data the logs mask #162 (a log needs far less cleaning, not none).

Tested on Windows 11, Python 3.13: CHECK VERDE: release/1.0.21 (d828963) | ruff pulito | 2236 passed, 14 skipped | copertura 97.15% (soglia 95%)

Field test on the intercom: done after this merge and before the release is published (ADR-19). The merge commit is deployed on the development HA, an outside view is opened on the 40507 (local UDP), and the debug log is checked for the plant's real IP, SIP id, GID, MAC and domain: none may appear. The train has no SIP, media, door or transport change; the test is about the log masking of #157/#159 on a real plant.

Changelog: none

🤖 Generated with Claude Code

[Apeiv]

Tested on a 40515 on the cloud (TLS + SRTP), main at 1ba3f21.

Check	Result
Test suite (full / media / browser)	pass: 2161 / 15 / 86
Loads, registered, clean log, card served, /av passive	pass
Outside view, 4 calls, 17 to 21 s each	pass: video OK, SRTP failures 0, audio both ways
Photo during the call	first JPEG 0.8 s after the grabber started, the street, clean

Your own test covers a 40507 on local UDP, so I spent most of the time on what only the cloud setup shows: the masking. Debug log on for all four calls, then I searched the raw text for every real value of this plant. None showed up in clear:

Plant data	Masked as
Our SIP id	id…NN#xxxx in From/To/Contact
SIP domain (with the Tab's MAC in it)	domain#xxxx
Tab's LAN IP	192.x.x.N in Route/Via
Home public IP	masked in received=, Contact, Route
Device UUID	uuid#xxxx
Authorization, SRTP keys	***

The MAC on its own, GID, IMEI and MyName didn't appear during calls.

One gap on 40515 cloud plants: the phone ids here are 5 digits (6xxxx), the same length as a panel extension. Ours is masked because it is registered, but another phone's id stays readable as a URI user, since _URI_USER only takes ids longer than _PANEL_EXT_MAX. redact_plant() on <sip:6xxxx@…> returns it unmasked. It didn't leak today, but it would in an INVITE or MESSAGE from another phone. Same thing one place over: _mask_field leaves a 5-digit SIP_ID readable, so a phonebook row for another phone would show it too. Masking 5-digit ids that aren't the panel's own extension would close both. Happy to open a PR.

Not plant data: the Vimar relay address shows in full in the media lines (First SRTP, First video RTP, RTCP) and in the SDP. It's Vimar's server, fine to leave.

Still to do by hand: a ring answered from the card on iPhone, Echo via Scrypted, the door message. I'll add them here.