A disaster-recovery snapshot carries the singleton `app_settings` row —
registration, the MFA requirement, the default locale and timezone, module
availability, the notification and AI configuration, the document size cap and
quota — and the restore wrote it back as a side effect of restoring one
account. An operator putting a single record back on Wednesday reconfigured the
whole host from Monday's file, silently, with nothing on screen saying so.
Rebuilding a host from a snapshot is a real case, so the write stays: it is now
`restoreInstanceSettings: true` on the request and a checkbox beside the typed
gate, off unless the operator ticks it. Restoring an account and reconfiguring
the installation are two decisions, so they are two answers.
The 422 confirmation sentence and the dialog copy drop the instance-wide claim;
the opt-in carries it instead, in all seven catalogues. Integration coverage
runs both arms: without the flag the settings row keeps its own values, with it
the file's values land.
Three consent-surface tests mocked the shared AI-consent bucket at 60 where it
is 20, and the analytics-read and MCP mocks named 60 against real caps of 120.
Nothing asserts the number today, so nothing was wrong on the wire — but the
mocked verdict now carries `limit` all the way to `X-RateLimit-Limit`, and the
first test that reads that header off a mock would pin the fiction.
The response-shape file's header pointed at a document that is not in the
repo; it says what it means directly instead. It also now records the one
thing the gate cannot see: `meta` is a loose object on purpose, so undeclared
keys inside it survive the parse and the strip comparison does not reach them.
The request-scoped capture kept the last verdict of any kind and fell back
to it when nothing had refused. That fallback fires in exactly one state —
no limiter refused this request — which is the definition of "this 429 came
from somewhere else". So a daily AI budget that is spent until midnight, an
hourly generation quota, and a provider-side 429 relayed onward all went out
carrying our own bucket's numbers: a remainder that says there is room left
and a Retry-After measured against an unrelated window. On the AI paths the
captured window can also expire while the provider call runs, so the dressed
delay could land at zero and tell the client to retry immediately into the
ceiling that had just refused it.
The cell now keeps refusals only. A request every bucket let through reports
nothing, and a 429 from another cause leaves undressed, which is correct:
there is no bucket to describe. The two-limiter rule is unchanged — the last
refusal wins, and a refusal still outranks a pass.
The unit case that pinned the fallback is turned round, and both directions
now run through the real wrapper: a limiter refusal gets the headers, a 429
raised without one gets none. The integration file gains the two shapes it
was missing — a route that consults two buckets must report the one that
refused, and a route whose 429 comes from the daily budget must carry no
rate-limit headers at all.
`GET /api/auth/me` publishes a boolean per module, and three different
situations collapse into the same `false`: the record has the module switched
off, the active grant's sections do not open it, or the operator switched it
off for the whole instance. A client holding the boolean has to guess between
them, so the empty state either says nothing or says something that may be
untrue -- offering a delegate a switch that is not theirs to reach, or telling
an account holder to turn on a module the operator removed.
`moduleAccess` carries the reason beside the booleans, which keep their exact
meaning: `enabled`, `disabled`, `not_granted`, `unavailable`, resolved
outside-in so the reason published is the one nothing further in can change.
`modules[key]` stays exactly `moduleAccess[key] === "enabled"` for every key,
and the two halves are built independently -- the boolean still runs through
`maskModulesToSections` -- so the test of that pairing is an assertion rather
than a restatement of a derivation. It is checked on the pure builder across a
matrix of grants, records and operator blobs, and again on the real route for a
scoped delegate, an unscoped one, and an operator-disabled module.
Both ends in one change, because a payload field with no reader is the failure
this repository keeps rediscovering. The reader is `ModuleDisabledNotice`, the
shared module-off empty state the empty-states audit asks for: it names the
reason in the reader's own language and offers the enable control only when the
record's own switch is what is off, since neither of the other two is fixable
from that page. The nutrients page adopts it and drops its bespoke block, which
had been handing every delegate a button that would refuse them.
Registered in the OpenAPI registry as `ModuleAccessMap` and carried on the
account payload; the native contract gains a field and loses nothing, since a
Bearer request carries no switch and never reads `not_granted`.
Response drift was structurally undetectable. `openapi:check` proves the
YAML matches the registry and the route-coverage guard proves the verb is
listed; both pass when the registered schema disagrees with what
`apiSuccess()` sends, and the coverage guard says so in its own doc
comment. That is how `/api/sync/changes` shipped five domains against a
three-domain contract and how the error envelope went years without
declaring the field the multi-issue 422 was built for.
Two bodies, the two that were broken: a full `/api/sync/changes` page
with live and soft-deleted rows in all five domains, its `cursorExpired`
short-circuit (a second shape under the same schema), and a 422 carrying
more than one issue.
`safeParse` alone would not have caught either finding, because Zod
strips undeclared keys instead of refusing them — a schema short of the
wire parses cleanly. So the parsed value is compared back to the raw
body; a strip means the contract is missing something the server sends.
Removing `cycleDays` / `cycles` from the sync schema, or `details` from
the envelope, fails all three cases.
`/api/measurements` and `/api/labs` sort on one non-unique column and
page with `skip`/`take`. When the sort key ties, Postgres does not
promise the same order for the query that fetches page N and the one
that fetches page N+1, so a client paging its own history can receive a
row twice and never receive another — silently, on both sides.
Ties are not hypothetical here. The partial unique index forbids two
live measurements of the same type at one instant, but the tie that
occurs in production is the cross-type one: a daily `stats:` rollup
posts steps, distance, flights and active energy at the identical
timestamp, and an unfiltered history list pages straight through the
block. A lab panel drawn in one sitting shares `takenAt` across every
analyte in it.
Appending `{ id: <same direction> }` to the `orderBy` makes the order
total. Same fix on the five siblings that page the same way — medication
intake history, illness day-logs, custom-metric entries, reminder
history, mental-health assessments — because the defect is the shape,
not the two routes it was found on. No wire change; `/api/sync/changes`
already ordered this way.
The fixtures seed ids in an order unrelated to insertion order, so the
assertion on order inside the tied block can only hold if the secondary
key is doing the work: with the tiebreaker removed the new file fails on
all three cases.
A refused client had nothing to back off against. `rateLimitHeaders()`
emitted `X-RateLimit-Remaining` and an ISO `X-RateLimit-Reset`, no
`Retry-After` and no cap, and only about 70 of the 199 route files that
consult a limiter attached even those two. The rest answered with a bare
prose envelope, including every endpoint an offline replay queue hits:
the four batch ingests, `/api/sync/changes` and `/api/auth/refresh`. A
first-run Apple Health import trips the 60-per-minute batch ceiling
reliably, and the only strategy left to the client was a blind backoff.
Threading headers through 129 handlers is 129 edits the next new route
forgets, so the verdict is carried instead: `checkRateLimit` records it
in a request-scoped cell and `apiHandler` dresses any 429 that leaves
without headers. A route that already built its own is left alone —
`Retry-After` is the presence test, and it comes from the same builder.
`Retry-After` is whole seconds rounding UP (RFC 9110 §10.2.3), floored
at zero. `X-RateLimit-Limit` is new and needs the cap, so the limiter
result carries it. `X-RateLimit-Reset` keeps its ISO form: it is the
wrong convention for the name, but it is on the wire today and changing
it is a decision of its own.
The cell lives in its own leaf module because 177 test files replace
`@/lib/rate-limit` wholesale with a mock factory, and `apiHandler` must
not depend on anything those factories stand in for.
Publishing the active record's module map and cycle flag on the account
payload was right for a whole-record grant and too generous for a narrowed
one. `accountAccess.active` is set for any live grant, level and scope
included, so a delegate scoped to `measurements` read whether the owner
tracks their cycle, their screeners, their illness episodes and their
supplement intake, and the cycle flag derives from the owner's recorded sex.
Both fields are now masked to the sections the active grant names: a closed
section answers `false`, which is the answer a module that is off already
gives and the answer the browser's route inventory has always acted on. An
unscoped grant names the whole record and is unaffected, so the own-record
and whole-record payloads are byte-identical to what they were.
The masking table lives beside the sharing vocabulary rather than inside it,
and runs one way only: no value in it can turn a module on, admit a read or
widen a grant.
`auditLog` serialises `details` with `JSON.stringify` into a text column,
so the assertion was reading a field off a string and finding nothing.
Refs #939
The native client replays its offline outbox with the same
Idempotency-Key on every attempt, and `POST /api/medications`, the
side-effect log and the inventory register ran the handler again on the
replay: after a lost success response the account was left with a
duplicate medication, a duplicate symptom entry or a second pen in the
supply count. The dedupe on `(externalSource, externalId)` only ever
covered a MIRRORED create; a manually entered medication carries
neither field and had nothing to collapse a retry onto.
All three wrap in `withIdempotency`, the composition the batch ingest
already ships, and the three operations publish the header so a client
generated against the contract can tell that a retry is safe. Zod
parsing and `requireRecordAuth` are untouched.
An integration test drives each route twice against real Postgres and
asserts what a status check cannot: one row in the table, and a replayed
201 carrying `X-Idempotent-Replay: true`.
The rollups and the cached status insights are recomputed after the
transaction, so a throw there left the rows repaired and every chart on
the pre-repair numbers — and no later run would notice, because the
audit row makes the account skip from then on.
The tail is wrapped now. The audit row is written with
`rollupsRefreshed: false` and only flipped once the recompute returned,
so a re-run can tell "repaired, rollups pending" from "done" and says so.
A run that could not finish the tail names the account, prints the
backfill to re-run — `scripts/backfill-rollups.ts --user <userId>` — and
exits non-zero.
Review finding M2.
The provenance criterion proves where a row came from, not which build
imported it. An account re-imported on the fixed build carries the same
`EXPORT_XML_SOURCE_MAX` stamp and the same `stats:` external id with the
right numbers and no audit row, so it is selected exactly like a broken
one — and its rest days, under 200 m because the phone stayed at home,
survive the plausibility check after another thousandfold multiply.
So one candidate row that would leave the range now refuses the whole
account: nothing written, no audit row, the offending rows and a printed
reason instead. The idempotency guard only ever held against the
script's own prior run; this holds against the case it could not see.
The runbook says the same in words — a re-imported account must not be
repaired, the re-import already did it — and says what the dry run is
for: it is where the decision is made, not a formality in front of one
already taken.
Integration coverage for the re-imported account: correct rows, an
`--apply` run, nothing changes and no audit row is left behind.
Review finding H1.
Turning a module off for a profile you look after wrote a column nothing
on screen was reading. GET /api/auth/me published the ACTOR's module map
while switched, and the navigation, the dashboard tiles and the insights
sections all gate on it — so a guardian who turned Cycle off for the child
kept seeing a Cycle destination inside the child's record, and a guardian
who tracks it saw one there whether or not the child did (#939).
The map and the resolved cycle flag now answer for the record the session
is inside. They are not display preferences: every surface they gate shows
the record's data, so the record is what decides. Locale, theme and layout
stay the actor's. The record is taken from the grant-validated active
entry rather than the session's own selector, so a revoked switch cannot
publish somebody's configuration, and with no switch the payload is what
it was.
Cycle also gets a row on the guardian's Modules surface, which it never
had. It is a delegated module, so its state is the record's own cycle
profile and modulePreferences.cycle is refused — which left a guardian no
way to turn it off for a record whose sex says it should be on. The family
names the real column instead of a blob entry the gate ignores.
Refs #939
A managed profile could be created and deleted and never changed. A name
typed as a placeholder was permanent, a household that moved kept the
timezone the record was created in, and there was no way to correct a date
of birth somebody finally had (#939).
PATCH /api/managed-profiles/{id} takes the fields creation takes, every
one optional and at least one required — an empty body is a refusal rather
than a save that audits a change nobody made. It carries the gate the rest
of the family carries: a cookie session with a fresh second factor, an
active Guardian grant on the record, and the same byte-identical 404 for
"no such record" and "not yours". A GET beside it answers the record as it
stands, because the account payload names a managed record and nothing
else, and a form filled from that would have had blank fields to send.
The new field is gender. Leaving it out of v1.37.0 meant every managed
record started with the column NULL, which is what the cycle module
derives its default from: a guardian creating a record for a child could
neither record the answer nor turn the module off from there. It reuses
the column and the three values every other surface writes, so there is
one word for one column rather than two.
Refs #939
The defect was invisible from the outside because every behavioural test in
the suite seeded a tagged reading. Two checks close that:
A structural guard that reads the dashboard surfaces and fails when a tile
derives presence from a list of enum buckets with no null / unknown arm and
no written reason, following one hop through a named constant. It carries
its own positive and negative controls, so a refactor that silently stops
matching anything fails here rather than going quiet.
An integration test that seeds twenty readings with a genuinely NULL context
against a real Postgres and reads the tile back through the same resolver
the page renders with: eligible, latest value intact, and a tagged reading
still filed under its own context.
Refs #943
The window-reset test burned its budget in a 50 ms window: four sequential
round-trips to a containerised Postgres had to finish inside a twentieth of a
second or the window closed under them, the counter reset, and the fourth call
came back allowed. Three failures in thirty days, two of them on `main` — run
33614804032 shows `expected true to be false` at rate-limit.test.ts:56 in a file
that took 1278 ms to run.
The short window bought nothing: the expiry this test is about is forced by hand
a few lines further down, by writing `reset_at` into the past. The burn now runs
in a 60 s window, so no round-trip can outlast it, and the test asserts the row's
count is 4 after the denial — a window that closed mid-burn would also have
produced three allowed calls, and the old assertion could not tell the two
apart.
Gates:
pnpm exec vitest run --config vitest.integration.config.mts \
tests/integration/rate-limit.test.ts
three consecutive runs, 2 passed / 2 passed / 2 passed.
Mutation: with the burn window back at 1 ms the run fails exactly as CI did —
"AssertionError: expected true to be false" on the denial assertion.
The unit fix only helps the next import. An account that imported an
archive before it carries a decade of days understated 1000x, and the
authoritative repair — re-importing the archive on the fixed build —
needs an archive the account may no longer have.
`scripts/repair-apple-health-distance.ts` is dry-run by default and
selects rows only by the stamp the archive fold itself writes: the
EXPORT_XML_SOURCE_MAX provenance plus the `stats:` external id minted
for that fold. Never by "the value looks small", so a row a native sync
has since overwritten is left alone. `--apply` multiplies one account
per transaction and writes the audit row in the same transaction, which
is what makes a second run a no-op. The archive's own unit is named by
the operator (`--unit=mi`) because the discarded attribute is the defect
itself, and a row whose repaired value would leave the plausible range
is reported rather than written.
Refs #944
The account-access integration contract compares whole entries, so the
two lists the payload now publishes belong in its expected shapes. The
managed-record leg posts the mood entry with the logged-at instant the
create schema requires.
The integration leg reads GET /api/auth/me as a guardian acting on a
managed profile and asserts both lists, lands a mood entry and an
allergy on the profile's record, is refused a vault write, and shows a
WRITE delegate writable sections without manageable ones together with
the refused mood entry the list withheld.
The browser leg opens the managed profile, adds a mood entry through
the control that used to be hidden and asserts the POST's status, and
finds the owner-only sentence and no upload control on the documents
page. The spec gets its own session jar and runs in the desktop
project only, like every other spec that moves the record selector.
"Sign out everywhere" revoked every other refresh token and nothing else.
A native login is two rows, and the Bearer check consults only the
paired access token, so the other phones kept working until that token
expired, up to a day on the default policy. The route also handed
destroyOtherSessions a session-kind credential for a Bearer caller, whose
session.id is the ApiToken row id; the exception matched nothing and the
calling phone revoked its own refresh token. destroyOtherSessions now
finds the paired access tokens through the refresh rows and revokes them
inside the same transaction, reporting the count, and the route names a
Bearer caller by its access-token hash. Programmatic tokens are untouched.
rotateRefreshToken marked the consumed row under usedAt: null alone, so a
revocation landing between the read and the write was overtaken and a
pair minted after the family ended stayed live. The guard now includes
revokedAt: null; on a lost race the minted pair is retired and the reason
distinguishes a revoked family from a double consume.
withIdempotency consulted the grant only at replay time, and only for its
existence; a narrow-scoped token was checked only on the delegated arm.
Both move into the cell key: a delegated cell carries the grant's id,
level and scope fingerprint, a narrow token's cells carry the token, and
a request with no live grant neither reads nor claims a cell. Cookie and
wildcard own-record cells are keyed byte-for-byte as before.
Tests: integration for the paired access tokens (cookie and Bearer
caller), a deterministic revoke-mid-rotation race against Postgres, unit
coverage for the new cell keys, and the structural guard moved to the
new invariant.
Rotation revoked the access token paired with the consumed refresh token
the instant the rotation committed. A native client with requests already
on the wire got 401 revoked on those, read that as "re-authenticate",
rotated again with the refresh token it had just spent, and tripped reuse
detection: the device's whole token family died because the client was
busy at the boundary.
Pull the paired token's expiry in to fifteen seconds instead. It is long
enough for an in-flight request on a slow mobile link to finish, including
a retry, and short enough that the token gains nothing meaningful over the
validity it already had. bearer.ts answers a passed expiry with reason
expired, which a client treats as "refresh", not as "sign in again".
LEAST(expires_at, $1) takes the minimum inside the single UPDATE, so the
write can only ever shorten: a token already inside its last seconds keeps
its own expiry, and there is no read-then-write window. The bound value is
an ISO-8601 UTC string cast in SQL so the comparison does not shift with
the host's timezone.
Reuse detection is untouched. A replayed refresh token, or one presented
under a spoofed device id, still revokes the family and its access tokens
on the spot, and so does the logout path.
Nothing calls it any more: the weekly pass and the off-host job both
walk the payload through the streaming writer. Keeping an exported
function that builds the entire record as one string invites the next
caller to reintroduce the memory profile this branch removes.
The off-host job was said to benefit from the streaming work the weekly
in-database backup got, and it did not. Only the payload builder was
shared: `runOffhostBackup` called `buildFullBackupJson` for the whole
document, `gzipSync` over that whole string, a whole-buffer
`createCipheriv` pass over the result, and handed the finished buffer to
a single `PutObject`. Four full copies of the record alive at once, on
the same arithmetic the weekly pass had already been fixed for.
Configuring an S3 target on a long-lived record therefore restarted the
container. On an account of 445 000 measurements the JSON alone is
242 MB, and against a 1 GB container's 524 MB heap the run died of heap
exhaustion seventeen seconds in — taking every signed-in session with
it, because the job shares the app process.
The write path is a stream end to end now: `streamFullBackupJson`
produces the document a page at a time, gzip and an incremental
AES-256-GCM writer consume it as it arrives, and `@aws-sdk/lib-storage`
puts it up in two 8 MB parts. What the process holds is fixed by the
pipeline's shape rather than by the record. Measured on that same
445 000-measurement account under `--max-old-space-size=450`, with the
heap pre-loaded to what a warm server holds: the old path dies with
`Ineffective mark-compacts near heap limit`, the new one finishes with a
peak 79 MB above its baseline and writes a 9.1 MB object.
The stored envelope gains version 3, which moves the GCM tag from in
front of the ciphertext to the end. That is the only change the stream
required — the tag exists only once the last block is in, so a leading
one means the whole object must exist before its first byte can be sent.
Nothing about the authentication moves: the tag still covers every
ciphertext byte and is verified before a byte of plaintext comes back.
Versions 1 and 2 still read, so every object already in a bucket
restores through the same reader with no flag to tell them apart.
The incremental writer is the crypto module's, split into its byte-level
half so the base64 form the in-database blob uses and the raw form an
object needs are one implementation under two framings. The off-host key
handling is unchanged: a separate `BACKUP_ENCRYPTION_KEY`, 64 hex or
32-byte base64, passed in rather than looked up.
A run that uploaded nothing for anybody is a failed job now, not
`ok: true` with an empty bucket behind it. Wrong credentials, a missing
bucket and an unreachable endpoint all fail every account rather than
one, and the target's own sentence rides out as the cause. A run where
some account got a copy still succeeds; retrying the whole cohort over
one object would re-upload everybody's.
The one ceiling left is structural rather than a memory bound, and it is
counted rather than discovered: a multipart upload carries 10 000 parts,
so an object past 80 GB is refused for that account with a clear message
instead of failing halfway through with an SDK error about part numbers.
`tests/integration/offhost-backup-streaming-memory.test.ts` pins the
budget from both sides — what the streaming uploader holds, and that the
materialising path does not fit the same budget on the same fixture in
the same process, so the number cannot pass by measuring nothing.
The per-user date order reached the entry form and nothing else. A
reporter switched away from MM/DD/YYYY, the date picker followed, and
the dashboard chart axes and the measurements list kept the old order.
`makeFormatters` took `timeFormat` and `dateFormat` as optional
parameters defaulting to "AUTO", so six of the seven constructions in
the tree silently rendered the locale default and nothing anywhere
said the setting had been dropped. Drop both defaults. The compiler
then names every caller: the chart bucket labels, the legacy
`@/lib/format` helpers the measurements list renders through, the
medication compliance and efficacy charts, the doctor-report PDF and
the clinician share view.
Thread the preference:
- `makeBucketLabelFormatters` takes the order; the health and mood
charts read `useDateFormatPreference()`. The UTC pin decides which
day the label names, the preference decides how it is spelled, and
the two are separate questions.
- `format.ts` reads `readStoredDateFormat()` beside the timezone and
hour-cycle mirrors it already read.
- The PDF takes `dateFormat` as a required render option, alongside
`timeFormat`, which was optional for the same reason.
- The clinician share view spells the OWNER's dates the owner's way.
It also picked up the owner's hour cycle, which it had been ignoring
while the PDF route beside it already honoured it. Both resolve
through the new `resolveUserFormatPreferences`, the counterpart to
`resolveUserTimezone`.
Textual months stay on the UI locale. The preference pins numeric
field order by rendering through de-DE / en-US / en-CA, and a textual
month through those would change the month name's language.
Add `date-order-explicit-formatter-guard`: a raw numeric date renderer
must take its locale from the preference or carry a written reason,
and a formatter construction may not hard-code "AUTO" for the order.
Its allowlist holds thirteen calendar-day key builders, each with the
reason it has no order to get wrong. The sibling timezone guard passes
every one of them, because naming a calendar says nothing about
spelling a date.
Refs #922
The mint's rate limit bounds a burst and does nothing about accumulation:
a stuck automation retrying politely inside its 10-per-minute allowance
still leaves several hundred live credentials behind in an afternoon,
each one to be revoked by hand.
Count live rows before the create and refuse at ten.
Live, not minted-ever, so revoking frees a slot and an expired token
stops occupying one — the ceiling bounds what currently exists rather
than what the account has ever done. Scoped to measurements:write for a
reason worth stating: signing in mints a wildcard access token per
session and those accumulate on their own, so counting every row would
let ordinary browser use exhaust the budget and then refuse a mint for a
reason invisible to the person hitting it.
409 rather than 429. This is a conflict with the account's current
state, not a rate: it does not clear by waiting, it clears by revoking,
and a 429 would tell the caller the opposite of what is true.
Placed after validation, so a malformed request hears about the
malformation rather than the ceiling, and before the create, so hitting
the limit costs no credential.
Ten is set to catch a loop rather than to ration — a household running
Home Assistant, a scale bridge and a couple of scripts sits at three or
four, and it is not a security boundary since the holder can revoke and
mint freely.
Tests pin the boundary, the query shape and the ordering, and were
mutation-checked: turning >= into > fails the ceiling case, and dropping
the scope filter fails the query-shape case. The integration leg seeds
a wildcard and an expired row as decoys, then revokes one live token and
asserts exactly one slot opens — the leg that fails if the count is of
mints rather than of tokens alive.
The writer added in v1.38.6 compared the process's live heap usage against
80 % of V8's heap limit and aborted above it. That reading is not the
backup's footprint; it is the whole process's, garbage included. A
long-running app sits at 400 MB of largely collectable heap, so on a 1 GB
container (a 419 MB budget) the weekly pass aborted the first flush of every
account: four out of four in seven seconds, the smallest of them a demo
record whose entire stored copy is 1.2 MB, each told its record was too large
and its operator to buy memory.
Drop the gauge. What the writer holds is bounded by its shape and needs no
reading to prove: the three unbounded tables are pulled a page at a time,
every other section is released as soon as its JSON is in the sink, and at
most one flush of text is pending. The one copy that does grow with the
record is the stored blob, so the bound moves there — counted in the
ciphertext bytes actually produced, against a fifth of this process's heap
limit, which is what two copies of the answer plus a live set have to fit
inside. The message names both numbers.
A pass that wrote nothing for anybody now fails its job instead of reporting
`ok: true` with `backed: 0`, so the failing-queue panel and the backups page
read the same thing an operator would. A pass where some accounts were
written keeps passing with the failures as counts.
Measured under --max-old-space-size=450 from a heap deliberately dirtied
first: 445 000 measurements peak at 28 MB held by the backup, a one-row
record at under 1 MB.
Two integration assertions carried the import path's own copy of the
mmol/L conversion, so unifying the two conversions left them asserting
numbers the code no longer produces. They now multiply by the exported
factor, which is what the unit tests already do.
The reason these were missed before the merge is worth writing down:
`pnpm test` does not run tests/integration, and neither the change that
unified the factor nor the gate on the integrated branch ran the suite
that covers the import path. A literal here re-pins whichever copy of a
conversion the caller happened to use when the test was written, so the
assertion is derived rather than pinned.
`DataBackup.data` holds every weekly disaster-recovery snapshot and every
uploaded pack, AES-256-GCM at rest like everything else, under a column
called `data`. The rotation registry was keyed on the `*Encrypted` naming
convention, so the script never read a backup, reported zero rows
remaining, and the runbook read that zero as permission to retire the
previous key. Following the documented procedure destroyed every stored
backup — the copy an operator reaches for precisely when nothing else
works.
Register it, and register the idempotent-replay response cache the same
sweep turned up. The backup column walks in bounded batches (one row is a
whole compressed account) and re-seals the ciphertext without reading the
plaintext, so both stored envelopes come back byte-identical and so will
any later one. The cache is marked disposable: a row it cannot read is
dropped rather than counted as an error, because keeping it only
guarantees the next replay gets a body it cannot parse.
Running the script against a seeded database rather than reading it turned
up two more. It died on the mood day-context note, whose table is keyed on
the entry rather than on an `id`, and abandoned every column after it —
the backups included. And the document index's verbatim text rotated
without ever being named in the summary.
Make the naming assumption stop carrying weight. A new guard traces the
encryption helpers through their wrappers, walks every Prisma write
payload, and fails when a column receives ciphertext without being
registered. It asserts an anchored, non-zero match count so a matcher that
stops matching fails instead of passing. A second check pins each model's
primary key against the schema.
The script now ends by naming the registered columns it did NOT walk and
exiting non-zero when any is missing: a zero that means "nothing left" and
a zero that means "never looked" must not print the same. The runbook's
retire-the-key step was rewritten around that signal, and tells an
operator who already rotated on an older release how to get their backups
back.
The compression that landed last release was real and was measured in the
wrong place. On a development machine with a multi-gigabyte heap the pass
finished; the container the job runs in is capped at 1 GB, which gives V8
a 524 MB limit, and there the same pass died of memory about fifteen
seconds in. Because the job shares the application process, that was not
one failed backup — it was a restart, and every signed-in session on the
host went with it.
Reproduced before touching anything, under `--max-old-space-size=450`
against a seeded account of 445 000 measurements, 30 000 mood entries and
60 000 intake events: `FATAL ERROR: Reached heap limit` at 34 s. The same
fixture doubled to 890 000 measurements now completes inside a 296 MB
limit, holding 16 MB of live heap where the old writer held 168 MB.
The document is written incrementally now. `buildFullBackupPayload` takes
a `deferBulk` mode that declares the three unbounded tables instead of
reading them, and `streamFullBackupJson` walks the payload in key order,
pulls those tables through page by page, and releases every other section
as soon as its JSON reaches the sink. `packBackupBlobStreaming` feeds that
into gzip and an AES-256-GCM stream, so nothing between the database rows
and the stored value ever exists whole. Both writers go through the same
per-row serialisers, and the round-trip suite pins their output byte for
byte rather than trusting that they agree.
The intake and mood reads are ordered totally now. A keyset cursor needs
it, and a partial order also let two exports of one account differ over a
tie, which every multi-dose day has.
The streamed ciphertext is a new shape at rest, because GCM only produces
its tag once the last block is in and a leading tag is what makes the old
layout unstreamable. It carries a marker no older writer can produce, so
the two families are disjoint by construction rather than by sniffing.
All three stored shapes — plain, compressed, streamed — restore, proved
both ways.
The writer also watches its own live heap and stops at 80 % of V8's
limit. An account too large for its host is now one failed backup,
counted in the run's meta, instead of a restart for everybody on it.
The stored column is the ceiling that is left: `data_backups.data` is a
single `text` value, so the finished artifact has to exist as one string,
and that is the only part of the job still growing with the record. The
operator runbook says so, with the number at which it starts to matter.
The unit suite reported all 1937 files green and then failed the job with
`EnvironmentTeardownError: Closing rpc while "onUserConsoleLog" was
pending`, naming a different scoring test each time.
Vitest replaces globalThis.console with a sink that forwards every write
to the main thread as an onUserConsoleLog RPC call. When a worker
finishes a test file it awaits the calls in flight and then rejects
whatever was started after that point, so a console write from work that
outlived its test lands in the gap and fails the run. The blame goes to
the file being torn down rather than to the test that started the work,
which is why the named file never logged anything itself.
Three paths detach a promise and log on the handle: fireAndForget, the
stale-while-revalidate cache refresh, and the offline query persister.
They now register the handled handle in a set that both suite setups
drain in afterEach, so no test leaves work in flight. The registry is
inert outside test: nothing there awaits it, and holding a reference to
an unsettled promise would retain it for the life of a worker. No
production behaviour changes, so there is no changelog entry.
Evidence. A focused harness that starts failing background tasks and
returns hit the teardown error 10 runs out of 10 before the change and
0 out of 10 after. Instrumenting the real suite showed two detached
console writes per run escaping their test, both from the rollup tail;
after the change, none, and thirteen writes per run now complete inside
the test that started them.
The guard asserts the behaviour, not a count: a registered task must
keep settleBackgroundTasks from resolving ahead of it, and the set must
drain back to empty. It goes red on unregistered work, on a disabled
registry, on an unwrapped call site, on a sweep that matches nothing,
and on a setup that drops the hook.
The export was its only user, and the export now reads through a helper that
marks an undecryptable note instead of blanking it. Nothing else wanted the
silent null: the list DTO never carries the note at all, only `hasNote`, and
the single-resource GET has always used the throwing decrypt so a key gap
surfaces there.
Deleted rather than kept behind a dead-code exemption. A fail-soft twin
sitting unused next to a strict one is the exact shape this branch has been
clearing out — the next person writing an export reaches for whatever is
nearest, and the whole point was that the nearest thing should not silently
swallow a row. It had no test of its own, so none was removed; the three
comments that pointed at it are rewritten to say what is true now.
The weekly pass can stop and leave no trace where anyone looks. The
handler's own error never runs when the job is killed rather than thrown;
`recordError()` increments an unnamed counter; and the admin status page's
failing-queue panel looks back 72 hours, which for a weekly queue means a
permanent failure is on screen three days in seven. The backups page
itself showed an ordinary row with an ordinary timestamp, because a copy
made six weeks ago reads exactly like Sunday's until somebody subtracts
the dates.
The listing now carries two facts that fail at different times: the age of
the newest scheduled copy, which never leaves any retention window, and
the terminal state of the last scheduled run, which is the only place the
reason is written down. `readLastQueueRun` takes the newest run rather
than a window, so a weekly queue's outcome stays readable until the next
one replaces it, and it unwraps the value pg-boss stores a thrown error
under, so the page says "job timed out" instead of raw JSON. The
failing-queue reader gets the same unwrap.
Manual uploads are excluded from the cadence judgement by the shape of the
input, so an ad-hoc upload cannot make a dead cron look alive, and a fresh
instance is not called overdue for having no copy yet.
The listing also stops reading every blob into the web process to measure
it; `octet_length` answers that in the database, which on an instance with
large records was its own way of running out of memory.
The weekly pass built the whole account in memory, serialised it whole,
encrypted the string whole, and handed the result to the database whole.
That is four full copies of the record at the same moment — the object
graph, the JSON, the base64 ciphertext at 1.33x the JSON, and the copy the
driver makes of that parameter. Measured against a seeded account of
445 000 readings, the JSON alone is 242 MB and the run dies of heap
exhaustion; under a modest heap it never reaches the insert. The pass was
not slow, it could not finish.
A health record's JSON is extremely repetitive, so gzip takes that 242 MB
to 14 MB and every copy after it shrinks with it. The same run completes
in about twenty seconds with a peak heap of 434 MB and a stored row of
25 MB instead of 322 MB — which also matters because the copy lives inside
the database it would be needed to restore. Both backup jobs now hand the
serialised form back from a frame of its own, so the payload graph is
released before the compress step allocates anything.
The marker sits inside the ciphertext, so the stored column keeps the one
shape `decrypt()` already understands and rows written before this still
restore. The off-host object gets the same treatment through the version
byte its envelope always carried; version 1 objects still read back, which
both the restore drill and the restore script depend on.
The obvious saving — dropping soft-deleted readings, four rows in five on
a real record — is not taken, and the round-trip test now seeds a
tombstone and asserts it makes the journey. A tombstone inside the
retention window is what tells a synced client the row is gone and what
makes a re-sent sample a no-op; a restore without them brings back every
measurement the account ever deleted on the next device sync.
A lab note or a biomarker context written under a key this instance no
longer holds came out of a portable export as `null` — byte-identical to a
column that was never written. The restore importer then wrote nothing, so
the text was gone and nothing said it had ever been there. The whole loss
fit inside a fail-soft `catch` written for a list view, where returning null
is the right answer.
Both now export a short marker naming what happened, and the marker survives
a restore. The Coach transcript already did this with a placeholder sentence
and the doctor report with a rendered flag; the marker is now one constant
the transcript shares, and its comment says why an empty string will not do.
The unit test that required `note: null` here pinned the loss. It keeps the
half that matters — one bad row must never abort the export — and asserts
the marker instead. The round trip is proven end to end against a real
restore.
`AppSettings.telegramGlobal`, `.ntfyGlobal` and `.webPushGlobal` were read
in two places, and neither of them was the one that matters: the setup-card
visibility on Settings -> Integrations and an echo in the preferences
payload. The dispatcher never looked at them. An operator who switched a
channel off instance-wide lost the setup card and kept every existing
delivery on that channel, indefinitely.
Gate each of the three in the dispatcher, in the shape
`isApiGloballyEnabled()` already uses for the twelve API routes. The read
sits above the per-user channel row and below the per-event preference, it
happens at most once per dispatch, and only when a switchable channel is
actually in the cascade. A skip writes a `globally_disabled` row to the
push-attempt ledger so the admin diagnostics name the cause, and does not
count as an attempt, so `dispatched` keeps telling the truth. The account's
own channel row is left enabled: the switch is the operator's, and flipping
it back has to restore delivery without every user re-enabling anything.
APNs, the generic webhook and email carry no instance-wide switch and are
not gated here; they are turned off at their source instead.
The unit test drives the real dispatcher over a stubbed settings row; the
integration test writes the actual singleton and reads the ledger back out
of Postgres, which is the layer that proves the column the admin route
writes is the column the dispatcher reads.
Neither read is a bug a behavioural test can see: both shapes return the
right rows, and only the plan differs. So the guards split by what each
can actually answer.
The unit guard asserts on the SQL the readers build, with comments
stripped so it matches the query text and never the prose describing it:
no DISTINCT ON on this path, a LATERAL driven from an unnested type list
with a LIMIT on the per-type branch, a bound type predicate on the rollup
read sitting alongside the user and granularity filters, and the route
still calling the shared readers rather than growing a second copy. Four
of its five cases fail against the pre-change SQL.
The integration companion runs the shipped readers and the pre-change SQL
side by side over one fixture on the real Postgres and requires identical
rows. The fixture carries the cases the rewrite could have broken: a type
whose only rows are tombstoned, a type with no rows, a type whose newest
row is a tombstone over a surviving one, and coarser rollup granularities
the DAY read must ignore. Dropping the soft-delete filter or shortening
the type list makes both cases fail.
Plan assertions are deliberately not in the integration file: a
testcontainer fixture is too small for the planner to choose the access
paths a real account produces, so an EXPLAIN there would pin nothing.
The scope-fence test asserts that a refused section leaves a
sharing.access.denied audit row naming the reason. That row is written
without being awaited, on purpose, so the audit never delays the 403 the
caller is waiting on, and it lands some time after the response returns.
The test bridged that gap with a fixed 150 ms sleep, which a slow runner
outran: the read found an empty table and the assertion reported no
refusal, which reads exactly like the section having reached the wire,
the one thing this file exists to catch.
The positive case now polls for the row and fails with a message naming
the caller when none arrives within five seconds. The absence case keeps
a fixed wait, longer than before, because an absence claim is only worth
something if it waited. Proven by dropping the audit write behind an env
flag: eleven cases fail with the new message, and pass again with the
write restored.
Marc-André Bombeck