v0.173.0 released (not vouched): CHANGELOG, REPORT, CONTEXT — the folder clean-up measured on three boxes, the rollback proven on Tester 1
gates / gates (push) Successful in 42s

Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
This commit is contained in:
admin committed 2026-11-10 16:16:42 +01:00
1 parent 776025c824
commit 5cebb766e3
3 files changed
+69 -20

No files matched your search

+51 -17
View File
@@ -1,23 +1,57 @@
# REPORT — the self-update folder cleans itself (R-1368 part 2) — committed, NOT released (2026-11-10, operator brief)
# REPORT — v0.173.0 released and delivered to the three test boxes; NOT vouched yet (2026-11-10, attended)
**Plain words first.** Each agent update left its downloaded file on the box for ever: 49 files and 740 MB on demo-hp.
The agent now removes the old files after an update that ended well. It keeps the file of the version that runs.
This is a commit only. No version, no build, no vouch. No box has it. It waits for the next agent release.
**Plain words first.** Each agent update left its downloaded file on the box for ever. Version 0.173.0 removes the old
files after an update that ended well. It is released and it runs on the three test boxes. The folders went from
49, 48 and 33 files to 1 file each. A rollback still works after the clean-up. The operator's vouch is still open.
The full report of the run is in `felhom.eu`: `REPORT-endings-kernels-fixes-2026-11-10.md`.
The full report of the run is in `felhom.eu`: `REPORT-release-space-fixes-2026-11-10.md`.
## One claim that was wrong
The CHANGELOG entry and the report of the morning said: a box cleans its folder at the first good update AFTER the
release. That was wrong. The commit runs in the new binary. So the box cleans the folder at the commit of the update
that brings the code. Measured on three boxes; the CHANGELOG entry is corrected.
## The release
| | |
|---|---|
| Change | `internal/selfupdate/prune.go` (new); the call in `MaybeCommit` (`commit.go`); one lock in `Execute` (`executor.go`) |
| Rule | After the wrapper's `commit` returned success: remove every staged binary but the running version's |
| Never removed | anything while `pending.json` exists or cannot be read; anything at a plain start or after a rollback; `pending.json`; a folder; a symlink; a name the code does not know |
| Rollback | Needs `pending.json` and `/usr/local/bin/felhom-agent.prev`. It reads no staged binary (`configs/felhom-selfupdate-guarded`, read first) |
| Tests | `TestPrune_*`, 6 tests. One runs the wrapper's real `commit` and `rollback` verbs from a sandboxed copy |
| Red seen | 3 times: no prune call; no pending refusal; no regular-file check |
| Gates | `python3 scripts/agent_gates.py --fast`: all OK |
| Not reached | `/usr/local/bin/felhom-agent.bak-*`: root-owned, made by a person. R-1368 part 4 stays open |
| Not proven | On a box. A box cleans its folder at the first good update AFTER the release that brings this code |
| Machines | None touched by this change |
| Version | 0.173.0, `scripts/release-agent.sh` from a clean clone of `776025c8` (tag `v0.173.0`) |
| Agent sha256 | `752b2727fb6043417ae5f342acbda52330521b21bb799fda90760b960feef7ea` |
| Bundle sha256 | `51ebc1f4c66d87918c87c035b4297c0e40e10ed31e3250c3837dfeaa0002e1c1` |
| CI | `776025c8`: run 2608, `gates #311`, success (read by commit on the public Actions page) |
| Delivery | one signed `agent_update` per box. No bundle job: `configs/` did not change since v0.172.1 |
| Vouch | **not given yet.** `pinned.day0_agent` stays 0.172.1 until the vouch is read back from the hub |
**Next.** A release session builds and vouches it. After two updates on a test box, read the folder: it must hold one
binary.
## The folder `/var/lib/felhom-agent/selfupdate`, before and after
| Box | Before (on 0.172.1) | After the commit of 0.173.0 | Commit (UTC) |
|---|---|---|---|
| demo-hp | 49 entries, 722,596 KB | 1 entry, 16,028 KB | 14:59:19, `removed=49` |
| demo-felhom | 48 entries, 706,968 KB | 1 entry, 16,028 KB | 14:51:50, `removed=48` |
| Tester 1 | 33 entries, 497,244 KB | 1 entry, 16,028 KB | 15:00:19, `removed=33` |
After each commit: `pending.json` absent, `/usr/local/bin/felhom-agent.prev` = 0.172.1, the unit active, no restart.
## The rollback proof (Tester 1)
1. A signed `agent_update` to 0.172.1 (the vouched version) was uploaded at 15:01:19 UTC. The box took it at its next
report, 15:14:31 UTC. The wrapper installed 0.172.1 and wrote the pending marker.
2. At 15:14:40 UTC, inside the 60-second dwell, one `systemctl kill -s SEGV` hit the agent's main process.
3. The unit's `OnFailure` ran the wrapper's `rollback`. Journal: `rolled back to previous binary and restarted
felhom-agent.service`. 0.173.0 started at 15:14:40 UTC.
4. Read at 15:15:56 UTC: agent 0.173.0, active, `pending.json` absent, no failed unit. The folder holds 2 files
(32,040 KB): the running version's and the failed 0.172.1. The failed file stays by design, until the next good
update.
This is the path the repo already has. No second version was released for the proof.
## What stays open
- The operator's vouch.
- `/usr/local/bin/felhom-agent.bak-*` (R-1368 part 4): demo-felhom holds 11 such files, demo-hp 1. The agent cannot
reach them.
- After a rollback, `/usr/local/bin/felhom-agent.prev` holds the same version as the running binary. That is how the
wrapper works; it is not new.
Evidence: felhom.eu `documentation/audits/release-space-fixes-2026-11-10/agent/`.