audit: state the 42-commit distance to v0.5.0 (#93)
The Snapshot paragraph said five commits separated the read tree
(56ed2671) from the v0.5.0 tag (fcc5bf5a), and that they were the
release, CLI and docs housekeeping plus test and style commits that
change no criterion. `git rev-list --count 56ed2671..v0.5.0` is 42;
the range is 127 files and +8107/-658 and contains the M25/M26 blob
slices, the erased-call-answer behaviour and its reproduction tests,
the nested-replay instance-pool fixes, the microVM cgroup fix and the
connection/blob compatibility corpus.
State the true distance and composition, say that the rows are the
56ed2671 reading except the M26 and M24 rows revised for #89 and #94,
and make clear that the cited ci run is on the read tree, not the tag.
Add the counting command with its expected 42 to "How to reproduce"
under the v0.5.0 tag, and replace the ambiguous "snapshot's commit" /
"at this snapshot" uses with the read tree or the release.
No verdict or count changes. Verified the count, the shortstat, every
listed commit's membership in the range, the M24 perf result and the
M26 rebuild test at v0.5.0, and that run 43637751312's head_sha is
56ed2671.
3523c28
audit: give M24 criterion 1 one verdict (#94)
The M24 row said the steady-state memory condition was not settled
and that no result was recorded in docs/perf.md; the gaps list said
the same. But docs/perf.md has recorded both runs at ADR 0140's
default since e7f87e0f, before this page was revised: the 10-minute
run passed every condition and the 60-minute one passed all ten
checks, +9.4 % VmRSS over the hour with the heap leveling off.
Both pages now state the same partial verdict. The run was on the
test machine confined to two CPUs, standing in for the 2-vCPU test
VPS; the brief defines criterion 1 on the test VPS, which has not
run it, and no workflow runs scripts/acceptance-m24.sh load. The
memory measurement is recorded, not missing. No counts or other
rows change.
ee518bd
audit: record the M26 rebuild test and its remaining gap (#89)
The M26 row said no test erases a subject's blobs and then rebuilds
state.db from the log. `blob_erasure.rs` does: the test
`an_erased_subjects_blobs_stop_counting_against_the_holders_quota`
erases user:1's blobs, drops the node, deletes state.db, its WAL and SHM
files and snapshots/, starts again with that state rebuilt from the log,
and asserts the rebuilt index still marks the blob erased and uncharged
(`erased_for_charge`, and the charge excludes it).
That covers the rebuild at the index. It is not an end-to-end read: no
test asks a guest or the API for the erased blob after the rebuild, and
no test erases a subject whose blobs are rooted only by a fork. Keep the
verdict partial and name those gaps, so the row no longer sends a
maintainer to write a rebuild test that already exists. The changes
table's M26 cell and the second strongest known gap say the same.
3cd2a39
status: make cluster verification prospective (#95)
The known-gaps bullet said a release is verified on a three-node
cluster since v0.4.0, while the same page records the cluster
track (M16–M19) as design only with no cluster crate and no such
environment provisioned. No tagged release has a cluster to
verify, so state the policy as future work and name the absent
environment as the reason the M32 cluster targets are unmeasured.
def851e
status, compliance audit: v0.5.0 released
Project status opens the 0.6.0-dev line and lists v0.5.0 as the latest
release with its engine and contents. The audit's snapshot moves to the
tag (fcc5bf5a) and says what the commits between the audited tree and the
tag changed.
bcc7320
status: M24 criterion 1 passes at the default concurrency
e64a33d
status: blobs always on (M26 B20)
d798f71
status: M25 and M26 on main
f8af3cd
compliance audit: snapshot at 7e87e5ce, isolation proofs on main's commit
2686a1c
compliance audit: v0.5.0 snapshot of main (2026-11-14)
Maps every criterion again at main f725254a: M20, M21, M23 and M24 moved
since the v0.4.0 audit; the blob track is listed as work on branches; the
isolation proofs are cited with their hosted runs and environments.
1335d6f
status: M24 on main, M20 accepted, M20 and M21 proven
77b16dc
status: main has the isolation series, M23 and the audit fixes
a4ee804
status: v0.5.0 progress and the decisions of 2026-11-13
Record the state of M20-M26 on main and the isolation branch, the
maintainers' decisions (M21's deviation ratified, the M24-M26 designs
accepted, M20 experimental until its proofs run), the WASI 0.3 gate check,
and correct the M0 tag claim (#56) and the compliance audit's reproduction
commands (#46).
3f2236f
status: record the decided gaps and the manual workflows
CI and nightly start only from workflow_dispatch. Release archives
carry the license notices. The restore demo is not a destroyed VPS.
The 1 000 turns/s target is across actors. Untagged SQL writes are
refused. Fork crypto follows the grant's in-forks mode.
895e25a
Project status: v0.4.1 is the latest release
The patch release shipped with its advisory; main is the 0.5.0-dev line. The
M20/M21 rows name the isolation branch, its experimental/opt-in status and what
still has to be proved before the ADRs move: a real KVM boot and the Ubuntu
24.04/AppArmor (R1) pair.
f0dc448
Publish project documentation
6e82abc