mirror of
https://github.com/meshtastic/firmware.git
synced 2026-09-16 08:30:04 -04:00
ea7d4aa4101f2b90568c2a5d30449e8900a83abb
3
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
ee15508494 |
time: arm the remaining 0-means-unset stamps through the helpers (#11830)
* time: add skipZero/safeMillis/timerEndsAtMillis helpers
skipZero() steps a millis value past 0, since stored stamps and deadlines
conventionally use 0 for "unset" and the one tick per ~49.7-day wrap that
lands on 0 would otherwise read as never-set.
safeMillis() covers a bare stamp; timerEndsAtMillis(delayMs) covers a
deadline, where the sum is what has to dodge 0 - a non-zero read plus a
delay lands there once per wrap - so it is not safeMillis() + delayMs.
* time: replace hand-rolled zero-dodging with the UptimeClock helpers
PacketHistory rxTimeMsec, EncryptedStorage s_lastFailMillis (stamps), and
SGM41562 lastRefreshMs_ / NextHopRouter learnedAtMsec (ternary stamps) each
hand-rolled skipZero() in place; swap in safeMillis()/skipZero() directly.
HapticFeedback pulseOffAt/delayedPulseAt and GPS fixHoldEnds hand-rolled the
deadline form - millis() + delay, then remap a 0 result to 1 - swap in
timerEndsAtMillis(delay).
No behavior change; each site keeps the value it already computed.
* time: guard the remaining 0-means-unset deadline/stamp writes
rebootAtMsec, shutdownAtMsec, and NotificationRenderer::alertBannerUntil are
all read back with a bare == 0 / != 0 check for 'not scheduled', but every
write site computed millis() + delay (or a bare millis() stamp) with no
guard against landing exactly on 0 - the same wrap hazard skipZero() exists
for, just never applied here.
Route every rebootAtMsec/shutdownAtMsec/alertBannerUntil write through
timerEndsAtMillis()/safeMillis(); RadioLibInterface's reboot-on-stuck-tx
sums an already-captured stamp rather than "now", so it goes through
skipZero() directly instead.
No behavior change outside the ~1-in-2^32 wrap window each site was
already exposed to.
* time: guard three more 0-means-unset deadline writes
ntp_renew (ethClient.cpp), suppressTouchTapUntilMs (Events.cpp), and tx_after
(RadioLibInterface.cpp) all read back 0 as a real state - forced NTP renewal,
no suppress window active, no TX delay armed, respectively - but each arm
site wrote a bare millis()/getMillis() + delay with no guard against the sum
landing exactly on 0.
Route each through Time::timerEndsAtMillis(). No behavior change outside the
wrap window each site was already exposed to.
Refresh the Throttle.h TODO list to note ntp_renew is converted too.
* motion: guard the calibration deadline and use Throttle::deadlinePassed
endCalibrationAt's arm site wrote millis() + calibrateFor with no guard
against landing on 0, the same value finishCalibrationIfExpired()/
drawFrameCalibration() treat as "not calibrating". Route it through
Time::timerEndsAtMillis().
Also swap finishCalibrationIfExpired()'s hand-rolled (int32_t)(now - deadline)
< 0 for Throttle::deadlinePassed(): same wrap-safe comparison the codebase
already provides, without the signed-cast pattern Throttle.h documents as
implementation-defined past INT32_MAX, and it drops the file's last direct
millis() call in favor of the Time:: wrapper the rest of it already uses.
* time: fix Throttle::execute()'s own zero-dodging
Both places execute() writes *lastExecutionMs - the first-ever-run branch
and the regular update - used bare Time::getMillis() with no guard against
landing on 0, which is the exact sentinel this function reads back as
"never run" one line above. A hit there makes the next call re-fire
immediately instead of respecting minumumIntervalMs.
Capture now via Time::safeMillis() once; every use downstream (the elapsed
comparison, the stored value) is then safe by construction instead of
needing the guard reapplied at each write.
* revert some safeMillis cases where overflow is a bad thing
* test(uptime): pin skipZero/safeMillis/timerEndsAtMillis at the wrap boundary
Covers the zero case, an ordinary nonzero value, and a sum that lands
exactly on 0 from a nonzero start - the case timerEndsAtMillis() exists
for, and the one the prior suite had no direct coverage of.
* time: restore the route-health write normalization and put it on one clock
noteRouteLearned()/noteRouteSuccess() lost their `now ? now : 1` normalization,
leaving learnedAtMsec able to store 0 - which getOrAllocRouteHealth() reads as an
ever-growing age, making the slot the first eviction candidate and permanently
stale. Normalize at the write, where the block comment already says it happens,
so every caller is covered rather than just today's two.
Both callers, the two isRouteStale() sites and doRetransmissions() now read
Time::getMillis(), so the stamp and every comparison against it share a clock.
doRetransmissions() goes back to getMillis(): its `now` feeds only comparisons,
never a 0-sentinel field, so skipping zero there only cost accuracy.
* time: read the haptic, InkHUD and calibration deadlines on the write's clock
These three deadlines were converted to Time::timerEndsAtMillis() on the write
side while their reads stayed on millis(), so each spanned two clocks and would
fire immediately or never under an injected test clock. Convert the reads to
match: HapticFeedback::scheduleNext()/runOnce(), the InkHUD tap-suppression
window, and the calibration countdown's read-back of screen->getEndCalibration().
MotionSensor's sampledAtMs is left alone - its write and read are both millis()
and consistent already.
* time: correct the sentinel notes to match what the code actually does
The Throttle.h enumeration claimed the remaining timerEndsAtMillis() callers
"already dodge the sentinel", which reads as a completeness claim the same branch
contradicts: RadioLibInterface's tx_after and activeReceiveStart are both 0=unarmed
and both still arm from bare millis(). Name them instead, so the deadline-type
conversion has the real list. The ntp_renew entry now separates a deliberate 0
("due now", forced at link-up) from a computed one, which is what changed there.
The three TODO(elapsed-stamp) blocks ran four and five lines against the repo's
one-or-two rule, and two of them argued their case wrongly. Throttle.cpp implied
safeMillis() simply doesn't help; in fact neither store is safe on the wrap tick -
the 1 underflows a same-instant read, the 0 re-takes the never-run branch - which
is the symmetry worth recording. PacketHistory.cpp called its dodge "reflecting
the previous pattern" when it is load-bearing: rxTimeMsec 0 means "empty slot"
(PacketHistory.h:21) and insert() drops a record stamped 0 outright, so without it
a packet arriving on the wrap tick is never stored and loses its dedup.
Also picks up trunk fmt's trailing-whitespace fix in Throttle.cpp and the comment
realignment in SGM41562.cpp that this branch's added comment knocked out.
* test(nexthop): pin the route-health stamp against the 0 sentinel
The uptime suite covers skipZero/safeMillis/timerEndsAtMillis themselves, but
nothing covered a call site, so the branch deleted noteRouteLearned()'s
normalization and stayed green. None of the existing route-health tests pass 0 as
`now` - they use 1000, learnAt, or millis() - (TTL + 5000) - which is exactly the
gap the regression went through.
Both new tests fail with "Expected 0 to be not equal to 0" when the skipZero() is
backed out of NextHopRouter, and pass with it. noteRouteSuccess() only refreshes
an existing record, so its twin learns a route first to reach the write.
Also drops a self-referential assertion in the uptime suite: comparing
getMillis() against safeMillis() passes even if safeMillis() does no dodge at
all, so it now asserts the literal.
* discard safemillis for skipzero (better semantics and therefore maintainability) and make consistent use of getmillis where it is called (to permit testing)
* more wrapzero safety
* STM gets some too
* time: stop the next 0-means-unset deadline being armed from raw millis()
The fields this branch armed through Time::timerEndsAtMillis() / Time::skipZero()
are the kind that get added by copy-paste: `rebootAtMsec = millis() + N` appears
at twenty-odd sites across six files, and the next module to defer a reboot will
be written from one of them. Nothing catches the mistake afterwards - the sum
lands on 0 for one tick per ~49.7-day wrap, so a test run, a soak and a bench
session all pass while a pending reboot, shutdown, DFU jump or banner expiry is
silently dropped.
Two guards, at the two places it can go wrong.
The helpers themselves: skipZero() is constexpr, so its contract is now pinned by
static_assert in the header rather than only by test_uptime_clock. The asserts are
chosen against the two plausible rewrites - `ms | 1` perturbs every even value and
`ms + 1` turns the last tick of the wrap into the 0 the function exists to avoid.
Both compile, and both pass a test that only checks skipZero(0); each trips a
distinct assert here, naming the failure mode.
The call sites: bin/lint-unset-sentinel-millis.sh flags a sentinel field in src/
assigned from a raw millis()/getMillis() read, and names the helper to use. It is
name-driven because the 0 contract is declared in src/main.h and enforced in six
other files, so no single-file scan can infer it; every one of the thirteen fields
was checked to actually test against 0 before being listed. nagCycleCutoff and
LinuxJoystick's nextRepeatX/nextRepeatY are deliberately absent - their unset state
is a separate bool - and the nine remaining `millis() + x` sites in src/ are locals
that never store 0 for anything to misread.
Blocking, unlike its note-level neighbours: there is no run-time enforcer to pair
with, and the tree has zero violations today, so gating costs nothing. Scoped to
src/ so test_uptime_clock can keep building raw wrap values on purpose.
bin/test-lint-unset-sentinel-millis.sh pins the scanner against 23 fixtures -
reads, disarms, shadowing locals, comments, string literals and the already-fixed
forms all have to stay quiet.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* time: guard the four 0-means-unset stamps this branch had missed
Sweeping src/ for the `if (stamp && <deadline check>)` idiom - the shape that
makes 0 mean "unset" - turned up four stamps still armed from a raw clock read,
so the new lint rule would have had to either ignore them or go red on checkout.
Each is the same one-tick-per-wrap hole the rest of the branch closes:
* TrackballInterruptBase lastInterruptTime, armed in all four ISR handlers and
explicitly disarmed to 0 at the threshold reset. getMillis() is the ISR-safe
read by construction - it compiles to millis() outside PIO_UNIT_TESTING - and
skipZero() is pure, so neither adds anything to interrupt context.
* NeighborInfoModule lastSentReply, read as `if (lastSentReply && ...)` before
the 3-minute reply throttle. Needed the UptimeClock.h include.
* PositionModule lastSentReply, same throttle; already on the injectable clock
but still missing the guard.
* NodeDB lastSort, whose own read spells the sentinel out as `lastSort == 0 ||`.
On the wrap tick each would read as never-stamped: a trackball debounce window
lost, a neighbour or position reply sent inside the throttle it was meant to
respect, one extra NodeDB sort. Cheap individually, which is why they were missed.
All four are now listed in bin/lint-unset-sentinel-millis.sh, so the rule covers
every field in the tree that actually tests against 0 rather than a subset, and
the header records the eight stamps left off for the opposite reason - their unset
state is a separate flag (isNagging, busyTx, heldX/heldY, formatted_this_boot,
heartbeat, gotwind, haveSample, lastIaqValid), so 0 is a value they may legally
hold. The rule is silent across src/ on this tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* lint: let a site opt out of the sentinel rule, with its reason on the record
The rule is blocking, so it needs an escape hatch for the site where 0 genuinely
is a legal timestamp - and the hatch should cost something, or it becomes the
first thing anyone reaches for. `unset-sentinel-ok: <reason>` in a comment on the
write, or on a comment line above it, suppresses that one statement:
// unset-sentinel-ok: busyTx carries the armed state, so 0 is a legal stamp here
lastTxStart = Time::getMillis();
The reason is mandatory. A bare `unset-sentinel-ok`, or a colon with nothing
after it, is reported instead of honoured - with a message saying so - so the
only way to silence a site is to write down why it is safe. trunk-ignore still
works, but this states the justification at the write and also applies when the
script runs outside trunk.
The marker is read from comment text collected during the same character-level
pass that strips comments and literals, not by re-scanning the raw line. That is
what keeps it out of reach of data: LOG_DEBUG("unset-sentinel-ok: ...") mutes
nothing, because a string literal is not a comment. It is also consumed by the
statement it was written for, so it cannot leak onto the next write - while still
carrying across any number of intervening comment lines to the statement below,
which is where a real justification wants to be written.
Twelve fixtures added for the new behaviour: both comment styles, block and
multi-line block comments, the bare form, the marker-in-a-string cases, and three
leak cases. 35 total, all green, under bash 3.2 as well.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* lint: watch the separate-flag stamps too, with their exemption stated at the write
The nine stamps whose armed state lives in a companion boolean were previously
just absent from the rule's list, which meant the reasoning for leaving them out
existed only as prose in a shell script. They are now listed and individually
opted out at the write, naming the flag that actually carries the armed state:
// unset-sentinel-ok: haveSample carries the armed state, so 0 is a legal stamp
lastSampleMs = Time::getMillis();
The point is what happens later. If someone rewrites `if (haveSample && ...)` as
`if (lastSampleMs && ...)`, the field has silently acquired the 0 contract; with
the opt-out sitting at the write, the claim to re-examine is in front of whoever
makes that edit instead of buried in bin/.
Every exemption was checked against its real read sites before being written, and
three candidates did not survive that check. They stay off the list, because
listing one would mean stamping an opt-out over a claim that does not hold:
* nagCycleCutoff. handleInputEvent reads `if (nagCycleCutoff != UINT32_MAX)`
without consulting isNagging, so at that read the field is its own armed flag
with UINT32_MAX as the sentinel - and the arm at ExternalNotificationModule
.cpp:521 can land exactly there. skipZero() cannot help: it lifts 0 to 1 and
leaves UINT32_MAX alone, which UptimeClock.h's own static_assert pins. There
is also a live boot-state bug behind this - the in-class initializer is 1
while isNagging starts false - and fixing the read is a behaviour change that
belongs in its own PR.
* TouchScreenBase::_start. Overloaded as an event stamp AND a `+ 30000`
suppression deadline compared by signed subtraction, so a near-zero value
reads as "long ago" rather than "armed 30s out" and LONG_PRESS re-fires.
skipZero() does not fix this one either: 1 reads as long-ago exactly as 0
does. It needs the stamp and the deadline held separately.
* StoreForwardModule::retry_delay. No reads at all today, so nothing misbehaves
yet; exempting it now would pre-approve the raw arm for whoever implements the
retry its own comment promises.
The rule is silent across src/ on this tree, and the header records all three
rejections so the next person does not have to re-derive them. The self-test's
negative fixture no longer uses nagCycleCutoff as its example of a safely
unlisted field - that would have encoded the opposite of what the header says.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* fix(time,lint): guard the recomputed tx_after, and judge one write at a time
Two review findings, both real.
setTransmitDelay() recomputes p->tx_after from a clamp of three candidates, and
that recomputation was still raw. Two lines above it, `if (p->tx_after)` is the read
that takes 0 as "no delay wanted", so a clamp landing on 0 drops the CSMA backoff
and the packet goes out immediately instead of after its computed delay. The first
arm site in this function was already guarded; this one was missed because the
value is not a plain `now + delay` and so does not fit timerEndsAtMillis() - it
takes skipZero() instead.
The narrowing order matters here and is spelled out at the site: add_delay is
unsigned long, 64-bit on the portduino host, so the clamp can exceed UINT32_MAX
there. skipZero() on the wide value would pass 0x100000000 through as non-zero and
the store to this uint32_t field would then truncate it back to the 0 being
avoided, so the cast comes first.
The lint rule judged each write by the wrong text. rhs was taken from the write to
the end of the accumulated statement, so a neighbour on the same line decided the
verdict - and it was wrong in both directions:
rebootAtMsec = millis() + 5; shutdownAtMsec = Time::timerEndsAtMillis(10);
the later helper call suppressed a genuine raw arm
rebootAtMsec = otherDeadline; shutdownAtMsec = millis();
the later millis() reported a safe copy
rhs is now cut at its own semicolon. Six fixtures cover it, including both cases
above, two raw writes on one line, two helper writes on one line, and a statement
split across lines, which must still see its whole right-hand side. 41 fixtures
total, green under bash 3.2.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* time: arm the remaining 0-means-unset stamps through the helpers
The follow-up sweep to the sixteen fields the previous commits covered. A field
here uses 0 to mean "unset" - some read spells `if (f)`, `f != 0`, `f == 0 ||` or
`f > 0`, or a site disarms it with `f = 0` - but it was armed from a raw clock read,
so once per ~49.7-day wrap it stores the value its own readers treat as never-set.
34 fields, 58 arm sites.
The rule could not have found most of them first. It treated `field = <variable>`
as inheriting whatever that variable did, which made the commonest shape in the tree
invisible: one `now = millis()` at the top of a runOnce(), then several
`xStartTime = now` below it. Listing those names would have bought no protection at
all, so the scanner now tracks a local assigned from a clock and treats a write from
it as the raw arm it is. One hop, one function, name-based, and it forgets a local
reassigned from anything else; taint is dropped at each function boundary. Twelve
fixtures pin it, including the negative cases - no leak across functions, `now` does
not match `nowMs`, and neither `==` nor `+=` records anything.
That pass immediately found a site the previous commits missed: setTransmitDelay()
recomputes p->tx_after from a tainted `now`, two lines under the `if (p->tx_after)`
read that takes 0 as "no delay wanted".
Three of the fields are worth naming because the consequence is not cosmetic:
* UpDownInterruptBase press/up/downStartTime - xDetected is only cleared INSIDE
the block guarded by `xDetected && xStartTime > 0`, so a stored 0 makes both the
entry and the exit condition unreachable and that button is dead for the rest of
the boot, not for one tick.
* PhoneAPI lastContactMsec - ServerAPI reads `lastContactMsec > 0` before the TCP
idle close, and the field stays 0 until the next inbound packet, so a client that
never speaks again leaks the socket for the life of the connection.
* EInkDisplay lastDrawMsec - `if (lastDrawMsec)` gates every plain display() call
on a keyframe having been shown, so a stored 0 stops the screen updating until
something calls Screen::forceDisplay() again.
TransmitHistory needed more than its arm sites. getLastSentToMeshMillis() returns 0
to mean "module has never sent", and besides the two stores, both reconstruction
helpers end in `millis() - msAgo`, which can produce a 0 of their own. All three
computed returns are guarded; the deliberate `return 0;` sentinels are untouched.
Judged and deliberately not changed:
* nRF54L15 connect_time_ms is armed from k_uptime_get_32(), not millis(). It is
guarded with skipZero() but keeps its own clock - swapping in Time::getMillis()
would have it compared against a k_uptime now at the watchdog read. The rule now
recognises that clock too, so listing the field is not an empty gesture.
* RotaryEncoderInterruptBase pressStartTime shares a name with the UpDown field and
has a different contract: no read here tests the stamp against 0, pressDetected
is the only armed flag. Opted out at the write. Its lastPressLongEventTime
sibling IS a `== 0` latch and is fixed.
* PositionModule line 38 copies a value the enclosing `if (restored != 0)` has
already proven non-zero. Opted out.
* pmMeasureStarted, adminKeyFallbackRefillMs, the two autosave stamps and
scrollStartDelay are lazy initialisations whose wrap behaviour costs at most one
interval and drops nothing. Left alone, and not listed.
47 lint fixtures green, the rule silent across src/, full native suite 1421/1421.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* time: dodge the wrap at the clock read, not only at the store
Review on #11830 made a point that was right and that this branch had wrong.
Applying skipZero() at the STORE while a reader measures elapsed time against a
raw clock splits the two sides apart for one tick per ~49.7-day wrap: the stamp
becomes 1 while `now` is still 0, so `now - stamp` is UINT32_MAX and a brand new
stamp reads as about 49.7 days old. Every elapsed-since guard then fires when it
must not. Concretely, UpDownInterruptBase computed `now - pressStartTime` and
emitted a long press for a fresh press, and TraceRouteModule read
`now - lastTraceRouteTime < cooldownMs` as false and bypassed its cooldown.
So the dodge moves to the read. Time::stampMillis() is getMillis() with the one 0
tick called 1; a site that both stores a stamp and measures against stamps reads
the clock once through it and stores that value directly. Nine files, and the 1 ms
skew is the same one skipZero() already documents.
Where the clock arrives as a PARAMETER the store keeps its own skipZero() as well,
because the function cannot assume the caller dodged anything. Removing that was a
real regression and test_nexthop_routing caught it: noteRouteLearned() and
noteRouteSuccess() are called with a literal 0 by
test_health_learn_never_stores_zero_sentinel and
test_health_success_never_stores_zero_sentinel, which assert the store normalises
it - 0 is the empty-slot marker getOrAllocRouteHealth() evicts on. The two guards
compose without shifting twice, since skipZero() of a non-zero value is itself.
trySmartBroadcast() and directResponseAllowed() have the same parameter shape and
keep their store-side guard for the same reason. Only stores fed by a stampMillis()
local in the same function are bare.
EInkParallelDisplay was missed the first time: the third class in the family, still
storing skipZero(getMillis()) while rate-limiting against a raw millis() local.
Normalised like its siblings.
The lint rule gained three false positives with the class-scope tracking, all of
them shapes that are not class bodies at all:
template <class T> void f(T x) { uint32_t lastSort = millis(); }
class Foo { void tick() { uint32_t lastSort = millis(); } };
void g(struct Bar *b) { uint32_t lastSort = millis(); }
Two causes. pending_class matched class/struct anywhere on the line, so a template
parameter list and an elaborated type in a parameter list both marked the following
FUNCTION body as class scope; it is anchored to the start of the line now. And
update_scope() runs at the end of a line, so a body opened earlier on the same line
had not been counted when the statement was judged; is_declaration() now also
counts unmatched braces earlier in the statement. The rule is blocking and
`template <class T>` is ordinary C++, so these would have reddened files nobody
touched.
note_taint() also never received the per-write `;` cut the judging path was given
earlier in review, so on a line holding two statements it learned taint from the
neighbour. Same cut applied.
Five tests in test_uptime_clock pin the contract, including one that asserts the
old store-only shape really does produce UINT32_MAX, and one that pins the 1 ms
skew at 399 rather than 400 so nobody "corrects" it back into a raw read. Lint
fixtures 53 -> 65. Full native suite 1426/1426, rule silent across src/.
Known residual, deliberately not changed: a store that dodges zero while its reader
measures through a Throttle:: helper still splits for that one tick, because those
helpers read the clock internally and raw. About eight sites tree-wide, including
PositionModule trySmartBroadcast and the lastContactMsec TCP idle check. Closing it
means making Throttle read through the dodge, which was proposed on #11692 and
declined there pending a caller audit, so fixing one site here would only make the
tree inconsistent. The direction is also the same one the un-dodged code already
took: a fresh stamp reads as old, and the guards involved were already passing on a
0 stamp.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* lint: a dodged value is safe to copy, not to do arithmetic on
Two more review findings on the rule, both real, both false negatives.
Arithmetic on an already-dodged value was excused. stampMillis() guarantees only
its own result, so `now + 5000` can carry a non-zero stamp straight back onto the
sentinel - 0xFFFFEC78 + 5000 is exactly 0. That sum is precisely what
Time::timerEndsAtMillis() exists to dodge, and the rule was waving it through
because a helper name appeared somewhere in the expression. Worse, a fixture
asserted that behaviour was correct, so the self-test was pinning the hole open.
A local holding a dodged value is now tracked separately from a tainted one: it may
be stored or copied straight through, but + or - applied at the OUTERMOST level is
reported and the message points at timerEndsAtMillis(). Depth-aware, so the operator
inside Time::skipZero(getMillis() - msAgo) is still fine, and so is the
`(d == 0) ? 0 : timerEndsAtMillis(d)` arming form, which has no top-level operator
at all. The wrong fixture is replaced by four: store-through, copy one more hop,
arithmetic on a dodged local, and arithmetic on a direct helper call.
A class body that opens and closes on one line was never recognised. The header
check rejected it because the line ends in a semicolon, which a one-liner body
always does, and even once armed the class brace counted as a function body and
excused the member. Both halves fixed: the header arms on the brace rather than on
the absence of a semicolon, and when the body opened on the statement being judged,
one unmatched brace is class scope while two is a method body inside it. Getting
that wrong first broke every multi-line class, because setting the per-statement
flag without also arming pending_class meant update_scope() never registered the
body - the three existing class fixtures caught it.
72 fixtures, green under bash 3.2, shellcheck clean, rule silent across src/.
No src/ or test/ file changes, so the native suite is untouched by this commit.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* style(time): trim comments to the house limit
---------
Co-authored-by: Tom <116762865+Nestpebble@users.noreply.github.com>
Co-authored-by: nomdetom <nomdetom@protonmail.com>
Co-authored-by: Tom <116762865+NomDeTom@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
||
|
|
80cfa52665 |
Add zero guards on time calculations where they were missing (#11692)
* time: add skipZero/safeMillis/timerEndsAtMillis helpers
skipZero() steps a millis value past 0, since stored stamps and deadlines
conventionally use 0 for "unset" and the one tick per ~49.7-day wrap that
lands on 0 would otherwise read as never-set.
safeMillis() covers a bare stamp; timerEndsAtMillis(delayMs) covers a
deadline, where the sum is what has to dodge 0 - a non-zero read plus a
delay lands there once per wrap - so it is not safeMillis() + delayMs.
* time: replace hand-rolled zero-dodging with the UptimeClock helpers
PacketHistory rxTimeMsec, EncryptedStorage s_lastFailMillis (stamps), and
SGM41562 lastRefreshMs_ / NextHopRouter learnedAtMsec (ternary stamps) each
hand-rolled skipZero() in place; swap in safeMillis()/skipZero() directly.
HapticFeedback pulseOffAt/delayedPulseAt and GPS fixHoldEnds hand-rolled the
deadline form - millis() + delay, then remap a 0 result to 1 - swap in
timerEndsAtMillis(delay).
No behavior change; each site keeps the value it already computed.
* time: guard the remaining 0-means-unset deadline/stamp writes
rebootAtMsec, shutdownAtMsec, and NotificationRenderer::alertBannerUntil are
all read back with a bare == 0 / != 0 check for 'not scheduled', but every
write site computed millis() + delay (or a bare millis() stamp) with no
guard against landing exactly on 0 - the same wrap hazard skipZero() exists
for, just never applied here.
Route every rebootAtMsec/shutdownAtMsec/alertBannerUntil write through
timerEndsAtMillis()/safeMillis(); RadioLibInterface's reboot-on-stuck-tx
sums an already-captured stamp rather than "now", so it goes through
skipZero() directly instead.
No behavior change outside the ~1-in-2^32 wrap window each site was
already exposed to.
* time: guard three more 0-means-unset deadline writes
ntp_renew (ethClient.cpp), suppressTouchTapUntilMs (Events.cpp), and tx_after
(RadioLibInterface.cpp) all read back 0 as a real state - forced NTP renewal,
no suppress window active, no TX delay armed, respectively - but each arm
site wrote a bare millis()/getMillis() + delay with no guard against the sum
landing exactly on 0.
Route each through Time::timerEndsAtMillis(). No behavior change outside the
wrap window each site was already exposed to.
Refresh the Throttle.h TODO list to note ntp_renew is converted too.
* motion: guard the calibration deadline and use Throttle::deadlinePassed
endCalibrationAt's arm site wrote millis() + calibrateFor with no guard
against landing on 0, the same value finishCalibrationIfExpired()/
drawFrameCalibration() treat as "not calibrating". Route it through
Time::timerEndsAtMillis().
Also swap finishCalibrationIfExpired()'s hand-rolled (int32_t)(now - deadline)
< 0 for Throttle::deadlinePassed(): same wrap-safe comparison the codebase
already provides, without the signed-cast pattern Throttle.h documents as
implementation-defined past INT32_MAX, and it drops the file's last direct
millis() call in favor of the Time:: wrapper the rest of it already uses.
* time: fix Throttle::execute()'s own zero-dodging
Both places execute() writes *lastExecutionMs - the first-ever-run branch
and the regular update - used bare Time::getMillis() with no guard against
landing on 0, which is the exact sentinel this function reads back as
"never run" one line above. A hit there makes the next call re-fire
immediately instead of respecting minumumIntervalMs.
Capture now via Time::safeMillis() once; every use downstream (the elapsed
comparison, the stored value) is then safe by construction instead of
needing the guard reapplied at each write.
* revert some safeMillis cases where overflow is a bad thing
* test(uptime): pin skipZero/safeMillis/timerEndsAtMillis at the wrap boundary
Covers the zero case, an ordinary nonzero value, and a sum that lands
exactly on 0 from a nonzero start - the case timerEndsAtMillis() exists
for, and the one the prior suite had no direct coverage of.
* time: restore the route-health write normalization and put it on one clock
noteRouteLearned()/noteRouteSuccess() lost their `now ? now : 1` normalization,
leaving learnedAtMsec able to store 0 - which getOrAllocRouteHealth() reads as an
ever-growing age, making the slot the first eviction candidate and permanently
stale. Normalize at the write, where the block comment already says it happens,
so every caller is covered rather than just today's two.
Both callers, the two isRouteStale() sites and doRetransmissions() now read
Time::getMillis(), so the stamp and every comparison against it share a clock.
doRetransmissions() goes back to getMillis(): its `now` feeds only comparisons,
never a 0-sentinel field, so skipping zero there only cost accuracy.
* time: read the haptic, InkHUD and calibration deadlines on the write's clock
These three deadlines were converted to Time::timerEndsAtMillis() on the write
side while their reads stayed on millis(), so each spanned two clocks and would
fire immediately or never under an injected test clock. Convert the reads to
match: HapticFeedback::scheduleNext()/runOnce(), the InkHUD tap-suppression
window, and the calibration countdown's read-back of screen->getEndCalibration().
MotionSensor's sampledAtMs is left alone - its write and read are both millis()
and consistent already.
* time: correct the sentinel notes to match what the code actually does
The Throttle.h enumeration claimed the remaining timerEndsAtMillis() callers
"already dodge the sentinel", which reads as a completeness claim the same branch
contradicts: RadioLibInterface's tx_after and activeReceiveStart are both 0=unarmed
and both still arm from bare millis(). Name them instead, so the deadline-type
conversion has the real list. The ntp_renew entry now separates a deliberate 0
("due now", forced at link-up) from a computed one, which is what changed there.
The three TODO(elapsed-stamp) blocks ran four and five lines against the repo's
one-or-two rule, and two of them argued their case wrongly. Throttle.cpp implied
safeMillis() simply doesn't help; in fact neither store is safe on the wrap tick -
the 1 underflows a same-instant read, the 0 re-takes the never-run branch - which
is the symmetry worth recording. PacketHistory.cpp called its dodge "reflecting
the previous pattern" when it is load-bearing: rxTimeMsec 0 means "empty slot"
(PacketHistory.h:21) and insert() drops a record stamped 0 outright, so without it
a packet arriving on the wrap tick is never stored and loses its dedup.
Also picks up trunk fmt's trailing-whitespace fix in Throttle.cpp and the comment
realignment in SGM41562.cpp that this branch's added comment knocked out.
* test(nexthop): pin the route-health stamp against the 0 sentinel
The uptime suite covers skipZero/safeMillis/timerEndsAtMillis themselves, but
nothing covered a call site, so the branch deleted noteRouteLearned()'s
normalization and stayed green. None of the existing route-health tests pass 0 as
`now` - they use 1000, learnAt, or millis() - (TTL + 5000) - which is exactly the
gap the regression went through.
Both new tests fail with "Expected 0 to be not equal to 0" when the skipZero() is
backed out of NextHopRouter, and pass with it. noteRouteSuccess() only refreshes
an existing record, so its twin learns a route first to reach the write.
Also drops a self-referential assertion in the uptime suite: comparing
getMillis() against safeMillis() passes even if safeMillis() does no dodge at
all, so it now asserts the literal.
* discard safemillis for skipzero (better semantics and therefore maintainability) and make consistent use of getmillis where it is called (to permit testing)
* more wrapzero safety
* STM gets some too
* time: stop the next 0-means-unset deadline being armed from raw millis()
The fields this branch armed through Time::timerEndsAtMillis() / Time::skipZero()
are the kind that get added by copy-paste: `rebootAtMsec = millis() + N` appears
at twenty-odd sites across six files, and the next module to defer a reboot will
be written from one of them. Nothing catches the mistake afterwards - the sum
lands on 0 for one tick per ~49.7-day wrap, so a test run, a soak and a bench
session all pass while a pending reboot, shutdown, DFU jump or banner expiry is
silently dropped.
Two guards, at the two places it can go wrong.
The helpers themselves: skipZero() is constexpr, so its contract is now pinned by
static_assert in the header rather than only by test_uptime_clock. The asserts are
chosen against the two plausible rewrites - `ms | 1` perturbs every even value and
`ms + 1` turns the last tick of the wrap into the 0 the function exists to avoid.
Both compile, and both pass a test that only checks skipZero(0); each trips a
distinct assert here, naming the failure mode.
The call sites: bin/lint-unset-sentinel-millis.sh flags a sentinel field in src/
assigned from a raw millis()/getMillis() read, and names the helper to use. It is
name-driven because the 0 contract is declared in src/main.h and enforced in six
other files, so no single-file scan can infer it; every one of the thirteen fields
was checked to actually test against 0 before being listed. nagCycleCutoff and
LinuxJoystick's nextRepeatX/nextRepeatY are deliberately absent - their unset state
is a separate bool - and the nine remaining `millis() + x` sites in src/ are locals
that never store 0 for anything to misread.
Blocking, unlike its note-level neighbours: there is no run-time enforcer to pair
with, and the tree has zero violations today, so gating costs nothing. Scoped to
src/ so test_uptime_clock can keep building raw wrap values on purpose.
bin/test-lint-unset-sentinel-millis.sh pins the scanner against 23 fixtures -
reads, disarms, shadowing locals, comments, string literals and the already-fixed
forms all have to stay quiet.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* time: guard the four 0-means-unset stamps this branch had missed
Sweeping src/ for the `if (stamp && <deadline check>)` idiom - the shape that
makes 0 mean "unset" - turned up four stamps still armed from a raw clock read,
so the new lint rule would have had to either ignore them or go red on checkout.
Each is the same one-tick-per-wrap hole the rest of the branch closes:
* TrackballInterruptBase lastInterruptTime, armed in all four ISR handlers and
explicitly disarmed to 0 at the threshold reset. getMillis() is the ISR-safe
read by construction - it compiles to millis() outside PIO_UNIT_TESTING - and
skipZero() is pure, so neither adds anything to interrupt context.
* NeighborInfoModule lastSentReply, read as `if (lastSentReply && ...)` before
the 3-minute reply throttle. Needed the UptimeClock.h include.
* PositionModule lastSentReply, same throttle; already on the injectable clock
but still missing the guard.
* NodeDB lastSort, whose own read spells the sentinel out as `lastSort == 0 ||`.
On the wrap tick each would read as never-stamped: a trackball debounce window
lost, a neighbour or position reply sent inside the throttle it was meant to
respect, one extra NodeDB sort. Cheap individually, which is why they were missed.
All four are now listed in bin/lint-unset-sentinel-millis.sh, so the rule covers
every field in the tree that actually tests against 0 rather than a subset, and
the header records the eight stamps left off for the opposite reason - their unset
state is a separate flag (isNagging, busyTx, heldX/heldY, formatted_this_boot,
heartbeat, gotwind, haveSample, lastIaqValid), so 0 is a value they may legally
hold. The rule is silent across src/ on this tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* lint: let a site opt out of the sentinel rule, with its reason on the record
The rule is blocking, so it needs an escape hatch for the site where 0 genuinely
is a legal timestamp - and the hatch should cost something, or it becomes the
first thing anyone reaches for. `unset-sentinel-ok: <reason>` in a comment on the
write, or on a comment line above it, suppresses that one statement:
// unset-sentinel-ok: busyTx carries the armed state, so 0 is a legal stamp here
lastTxStart = Time::getMillis();
The reason is mandatory. A bare `unset-sentinel-ok`, or a colon with nothing
after it, is reported instead of honoured - with a message saying so - so the
only way to silence a site is to write down why it is safe. trunk-ignore still
works, but this states the justification at the write and also applies when the
script runs outside trunk.
The marker is read from comment text collected during the same character-level
pass that strips comments and literals, not by re-scanning the raw line. That is
what keeps it out of reach of data: LOG_DEBUG("unset-sentinel-ok: ...") mutes
nothing, because a string literal is not a comment. It is also consumed by the
statement it was written for, so it cannot leak onto the next write - while still
carrying across any number of intervening comment lines to the statement below,
which is where a real justification wants to be written.
Twelve fixtures added for the new behaviour: both comment styles, block and
multi-line block comments, the bare form, the marker-in-a-string cases, and three
leak cases. 35 total, all green, under bash 3.2 as well.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* lint: watch the separate-flag stamps too, with their exemption stated at the write
The nine stamps whose armed state lives in a companion boolean were previously
just absent from the rule's list, which meant the reasoning for leaving them out
existed only as prose in a shell script. They are now listed and individually
opted out at the write, naming the flag that actually carries the armed state:
// unset-sentinel-ok: haveSample carries the armed state, so 0 is a legal stamp
lastSampleMs = Time::getMillis();
The point is what happens later. If someone rewrites `if (haveSample && ...)` as
`if (lastSampleMs && ...)`, the field has silently acquired the 0 contract; with
the opt-out sitting at the write, the claim to re-examine is in front of whoever
makes that edit instead of buried in bin/.
Every exemption was checked against its real read sites before being written, and
three candidates did not survive that check. They stay off the list, because
listing one would mean stamping an opt-out over a claim that does not hold:
* nagCycleCutoff. handleInputEvent reads `if (nagCycleCutoff != UINT32_MAX)`
without consulting isNagging, so at that read the field is its own armed flag
with UINT32_MAX as the sentinel - and the arm at ExternalNotificationModule
.cpp:521 can land exactly there. skipZero() cannot help: it lifts 0 to 1 and
leaves UINT32_MAX alone, which UptimeClock.h's own static_assert pins. There
is also a live boot-state bug behind this - the in-class initializer is 1
while isNagging starts false - and fixing the read is a behaviour change that
belongs in its own PR.
* TouchScreenBase::_start. Overloaded as an event stamp AND a `+ 30000`
suppression deadline compared by signed subtraction, so a near-zero value
reads as "long ago" rather than "armed 30s out" and LONG_PRESS re-fires.
skipZero() does not fix this one either: 1 reads as long-ago exactly as 0
does. It needs the stamp and the deadline held separately.
* StoreForwardModule::retry_delay. No reads at all today, so nothing misbehaves
yet; exempting it now would pre-approve the raw arm for whoever implements the
retry its own comment promises.
The rule is silent across src/ on this tree, and the header records all three
rejections so the next person does not have to re-derive them. The self-test's
negative fixture no longer uses nagCycleCutoff as its example of a safely
unlisted field - that would have encoded the opposite of what the header says.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* fix(time,lint): guard the recomputed tx_after, and judge one write at a time
Two review findings, both real.
setTransmitDelay() recomputes p->tx_after from a clamp of three candidates, and
that recomputation was still raw. Two lines above it, `if (p->tx_after)` is the read
that takes 0 as "no delay wanted", so a clamp landing on 0 drops the CSMA backoff
and the packet goes out immediately instead of after its computed delay. The first
arm site in this function was already guarded; this one was missed because the
value is not a plain `now + delay` and so does not fit timerEndsAtMillis() - it
takes skipZero() instead.
The narrowing order matters here and is spelled out at the site: add_delay is
unsigned long, 64-bit on the portduino host, so the clamp can exceed UINT32_MAX
there. skipZero() on the wide value would pass 0x100000000 through as non-zero and
the store to this uint32_t field would then truncate it back to the 0 being
avoided, so the cast comes first.
The lint rule judged each write by the wrong text. rhs was taken from the write to
the end of the accumulated statement, so a neighbour on the same line decided the
verdict - and it was wrong in both directions:
rebootAtMsec = millis() + 5; shutdownAtMsec = Time::timerEndsAtMillis(10);
the later helper call suppressed a genuine raw arm
rebootAtMsec = otherDeadline; shutdownAtMsec = millis();
the later millis() reported a safe copy
rhs is now cut at its own semicolon. Six fixtures cover it, including both cases
above, two raw writes on one line, two helper writes on one line, and a statement
split across lines, which must still see its whole right-hand side. 41 fixtures
total, green under bash 3.2.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---------
Co-authored-by: Tom <116762865+Nestpebble@users.noreply.github.com>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
||
|
|
fdb644e0b7 |
Fix millis() rollover in deadline, interval, and timestamp handling (#11291)
* Add native test coverage for the UptimeClock monotonic seam
src/UptimeClock.{h,cpp} shipped without a dedicated test suite. Port the six
tests from the monotonic-time branch (test/test_time), retargeted to the
renamed header.
The wrap test crosses 0xFFFFFFFF via advanceTestMillis() rather than a second
setTestMillis(): setTestMillis() sets clockSourceChanged, which makes
getMillis64() rebase its accumulator and swallow the wrap.
* NextHopRouter: fix 49.7-day millis() rollover in retransmission timing
Resolves the "FIXME, handle 51 day rolloever here!!!" in
NextHopRouter::doRetransmissions() by switching the retransmission-due
comparison from plain unsigned <= to a signed-difference cast.
The previous p.nextTxMsec <= now comparison silently breaks across the
~49.7 day millis() wraparound: pending retransmissions either stall
for the remainder of the wrap window, or all fire simultaneously at
the rollover boundary. Long-running router/infrastructure nodes do hit
this in practice.
The replacement (int32_t)(p.nextTxMsec - now) <= 0 is the standard
Arduino/embedded idiom for rollover-safe deadline checks and behaves
identically to the original for any non-wrap timing.
* Address Copilot review: use unsigned half-range for rollover-safe retransmit check
Review feedback from @Copilot on PR #10227: casting a uint32_t
subtraction to int32_t is implementation-defined in C++ when the
unsigned value exceeds INT32_MAX (even though it works on typical
two's-complement targets).
Switch to the fully well-defined unsigned half-range form:
nextTxMsec is in the past-or-equal iff (now - nextTxMsec) has not
wrapped past 2^31 ms. Future offsets < 2^31 ms wrap into the top
half and read as 'not yet'.
Same semantics as the signed-cast version on every two's-complement
platform we care about, but portable to any conforming C++ impl.
* Use monotonic time for airtime windows
* Document monotonic airtime windows
* Fix test_packet_signing sentinel that #10227's rollover fix inverts
test_C3_invalid_repeated_packet_cannot_ack_or_change_retry_state parked a
pending packet at nextTxMsec = UINT32_MAX to mean "never retransmit", then
asserted that a rejected repeated packet leaves the retry state untouched.
NextHopRouter::doRetransmissions() now tests whether a retransmit is due with
an unsigned half-range compare, (uint32_t)(now - nextTxMsec) < 0x80000000u,
so that retransmission timing survives the ~49.7 day millis() wrap. Under it
now - 0xFFFFFFFF == now + 1, a small positive delta, so UINT32_MAX reads as
~1ms in the past: the retransmit fires and rewrites nextTxMsec, and the test
failed with "Expected 4294967295 Was 6247".
Use a representable future time instead. Production is unaffected either way -
nextTxMsec is only ever written as millis() + d, and UINT32_MAX came from the
test harness alone - so the sentinel is what needs to go, not the comparison.
Special-casing UINT32_MAX in the retransmit path would keep a value that reads
as "expired" under any wrap-correct compare.
The value is held in a local because millis() advances across
runPipelineIngress(), so recomputing it at the assertion would compare against
a different number.
Reported upstream on meshtastic/firmware#10227, whose branch predates this test.
* Make Throttle time-injectable and add hasElapsed()
Throttle backs ~94 call sites, which makes it the highest-leverage place in
the tree to put the clock seam: reading Time::getMillis() instead of millis()
in its three call sites turns all of them into time-injectable code at once,
without touching any of them. The 32-bit millis() wrap is not otherwise
reachable from a native test.
The read is behaviour-preserving - Time::getMillis() returns millis() unless a
test injects a clock - and the full native suite passes with it live.
Also add hasElapsed(), the complement of isWithinTimespanMs(), because 51 of
the 94 call sites are spelled !isWithinTimespanMs and read poorly. Its
boundary is inclusive (>=) since isWithinTimespanMs uses <; both are
documented. It deliberately does not treat lastExecutionMs == 0 as "never
run": call sites pair that test with the interval check themselves, and
absorbing a sentinel into the one helper every module depends on is exactly
the value-overloading hazard being removed elsewhere.
Migrating the existing !isWithinTimespanMs sites is cosmetic and deliberately
left out of this commit.
test/test_throttle/ covers window semantics, both boundaries, the complement
identity, execute()'s first-run and throttled paths, and - the point of the
exercise - a window opened before the wrap closing correctly after it,
including at the 24h interval that is the longest in the tree.
* Stop disarmed deadline sentinels reaching the comparison
Two deadline variables encoded "inactive" as a magic value that only reads as
"never" because the comparison against it is a naive millis() compare. Under
any rollover-correct comparison both invert to "expired ~49 days ago", so they
have to be untangled before those comparisons can be fixed.
Power::reboot() set rebootAtMsec = -1 on platforms with no reboot
implementation, intending "never fire". Every reader already treats 0 as the
disarm value - powerCommandsCheck() tests `if (rebootAtMsec && ...)`, and
AdminModule writes 0 to cancel - so -1 was both wrong and unnecessary. Use 0.
Left as UINT32_MAX it would reboot-loop the moment the comparison is corrected.
ExternalNotificationModule's nag window compared against nagCycleCutoff, which
holds UINT32_MAX once stopped and 1 at boot. isNagging is the real armed flag,
so test it first and short-circuit: a disarmed cutoff can no longer reach the
arithmetic, while an idle module still takes the same sleep path that the
boot-time value of 1 was relying on.
Note this fixes the sentinel only. The comparison itself is still a naive
`nagCycleCutoff < millis()` and remains on the list to convert.
* Fix millis() rollover in every deadline and interval comparison
Roughly 20 sites compared against millis() directly - `millis() > deadline`,
`deadline < millis()`, `last + interval < millis()`. All of them break for
about 24 days after the 32-bit millis() wrap: depending on which side of the
wrap each value sits, the action either stalls for weeks or fires immediately
and repeatedly. The longest affected interval is the 12 hour NTP renewal, a
~50x margin against the wrap, so none of these needed the range - only the
correct comparison.
Add Throttle::deadlinePassed(deadlineMs) for sites that store an absolute
deadline they cannot re-express as "interval since an event". It uses the same
unsigned half-range test as NextHopRouter::doRetransmissions() rather than
introducing a competing signed-cast idiom, and unlike the signed cast it is
defined for every input. Sites that do store an event use the existing
isWithinTimespanMs / hasElapsed. Nothing gained new state.
Because both helpers read Time::getMillis(), every converted site is now
reachable from a native test that drives the clock across the wrap; the
comparison itself is covered directly in test/test_throttle/.
Sentinel handling is the reason this could not be a mechanical rewrite. The
disarm convention is not uniform: 0 means "inactive" for rebootAtMsec,
shutdownAtMsec, alertBannerUntil, fixHoldEnds, suppressUntilMs and
touchResumeBlockUntilMs; 0 means "due now" for ntp_renew, which is forced to 0
at link-up; UINT32_MAX means "inactive" for nagCycleCutoff; and
alertBannerUntil == 0 in isOverlayBannerShowing() means "show indefinitely".
Every inactive marker is arithmetically far in the past, so a correct
comparison fires on it - each site tests its sentinel before the arithmetic,
and keeps the meaning it had.
Two sites carried a second bug found on the way:
BME680Sensor tested (stateUpdateCounter * STATE_SAVE_PERIOD) < millis(). With
a 6 hour period and a uint16_t counter that product overflows uint32_t after
about 198 saves, independently of the millis() wrap. It now measures the
interval since the last save.
EInkDynamicDisplay had `if (previousRunMs > millis()) return;` as a millis()
overflow guard, which skipped rate limiting entirely for the whole post-wrap
period - the bug it meant to prevent. Every check below it already goes
through Throttle, so the guard is removed rather than fixed.
MotionSensor's calibration countdown is converted to a signed delta rather
than deadlinePassed, because it needs the remaining magnitude and not a
boolean; that matches the already-correct check in the same file.
* Remove getMillis64() and use Throttle for the NodeInfo reply window
getMillis64() had exactly one caller and no callers in tests. It also carried
obligations that made it the wrong shape for this firmware: a wrap accumulator
in mutable statics, which is not ISR-safe, and which must be polled at least
once every ~49.7 days or it silently misses a wrap and returns a time ~49 days
short.
Its one caller only wanted to know whether a 12 hour suppression window had
elapsed - which Throttle answers correctly across the wrap without any
accumulator. NodeInfoModule now stores Time::getMillis() in lastNodeInfoSeen
and tests the window with Throttle::isWithinTimespanMs, so the map holds
milliseconds rather than seconds derived from a 64-bit read.
USERPREFS_NODEINFO_REPLY_SUPPRESS_SECS is user-overridable and now feeds a
multiply by 1000, so a static_assert rejects any value too large to express in
milliseconds instead of letting it wrap.
clockSourceChanged goes too. It existed solely to rebase getMillis64()'s
accumulator when a test swapped clock sources, and it made the wrap untestable
through the injection API: setTestMillis() set the flag, so a wrap crossed by
two setTestMillis() calls was swallowed. With the accumulator gone the flag has
nothing to rebase, and the injection API is a plain settable clock.
The three getMillis64 tests are dropped as they no longer describe anything.
One test replaces them, pinning that advanceTestMillis() wraps past
0xFFFFFFFF rather than saturating, since the Throttle wrap tests rely on it.
Also fix eviction in pruneLastNodeInfoCache(): it picked the entry with the
smallest stored stamp, which is the wrong victim once some stamps sit on the
far side of the wrap. It now evicts the largest elapsed time.
* Add CI guard and docs rule against naive millis() comparisons
Fixing the existing sites does not stop the next one being added. The
millis-deadline-check job rejects millis() placed directly next to a comparison
operator, in either order, anywhere in src/. It lives in test_native.yml
alongside suite-count-check, which sets the precedent for a repo-hygiene guard
that CI enforces and bin/run-tests.sh does not.
The correct idioms all subtract before comparing, so none of them match the
pattern. Line comments are stripped first, so documentation is free to name the
broken form - as the guard's own comment and the coding conventions both do.
Writing the check before finishing the sweep turned out to be worth it: it
found roughly 14 sites that a by-hand audit of deadline variables had missed,
including two extra nagCycleCutoff compares, both boot-screen timeouts, and a
6 hour sensor save interval that was also overflowing a uint32_t multiply.
.github/millis-deadline-allowlist.txt covers the cases that are genuinely not
deadline tests. Both current entries are uptime thresholds - "has the device
been up N ms" - with no stored deadline and no event to measure from: a 30s
button holdoff against phantom shutdown from floating pins, and a 10s window
for the OEM boot logo. Each re-crosses its threshold once per wrap, which is
harmless for boot-holdoff logic and not worth new state to avoid. Entries are
keyed on file plus exact source text, without line numbers, so an edit above an
entry does not silently invalidate it.
Locally the guard reports 19 matches before the sweep and 2 after, both
allowlisted.
The Throttle bullet in the coding conventions is rewritten from "prefer
Throttle for rate limiting" to "never compare against millis() directly", lists
all four helpers with when to use which, names the CI guard, and documents the
sentinel hazard with the rebootAtMsec = -1 case that would have become a reboot
loop. Mirrored into AGENTS.md; CLAUDE.md gets a pointer row.
* Trim rollover comments to what the code needs
The comments added with the millis() rollover fixes carried too much of the
investigation that produced them: how many sites were found, which document
recorded them, what the old code used to do. That belongs in the commit history,
not in the source, and some of it was already stale - Power::reboot() still
described the check it disarms as "a naive millis() > deadline" when that
comparison had been fixed in the same series.
What stays is the non-obvious part at each site: which sentinel value the
variable overloads and what it means there, since that differs between call
sites and is what a correct comparison gets wrong. 0 means "not scheduled" for
rebootAtMsec, "renew now" for ntp_renew, and "show indefinitely" in
isOverlayBannerShowing().
Exposition is kept where it earns its place: the Throttle helpers, the uptime
clock's note on why there is no 64-bit variant, and the tests. The Throttle
docs lose only the site count and the "longest interval in the firmware"
statistic, both of which would age badly; the range trade-off between the two
forms is what a caller actually needs.
Comments only - no code changed, verified by diff.
* possible fixes
* Address review feedback on the rollover fixes
- BME680Sensor: checkpoint lastStateSaveMs after a successful write instead of
at the interval test. The first save (IAQ accuracy >= 2) left it at 0, timing
the next save from boot, and stamping before the write deferred the retry a
full period when the write failed. Reads Time::getMillis(), the same clock
Throttle compares against.
- Throttle: add deadlinePassedAt(now, deadline) for loops that snapshot the
clock once and test many deadlines; deadlinePassed() now delegates to it.
NextHopRouter::doRetransmissions() uses it, replacing the inline half-range
compare adopted from #10227 (nightjoker7) - same arithmetic, credited at the
call site - and takes its snapshot from Time::getMillis() so setNextTx()
deadlines and the due test cannot diverge under an injected test clock.
- test_native.yml: set -euo pipefail in the millis-deadline guard, matching the
sibling suite-count job. Without -e a partially failed scan could report "no
violations" from truncated output.
- test_packet_signing: build the not-due deadline from Time::getMillis() rather
than millis(), so the test and the router read one clock.
- test_throttle: cover deadlinePassedAt(), and correct a wrapped-value comment
(0xFFFFFF00 + 400 is 0x00000090, not 0x00000094).
Two review comments were declined: the AirTime mutex (every airTime-> caller
runs in the single cooperative loop, WebServerThread included) and the
MotionSensor 0-sentinel countdown (the calibration frame is only installed
while a window is open).
clod helped out here
* Correct the described failure window of a naive millis() compare
The comments and agent docs said a bare `millis() > deadline` "breaks for ~24
days after the wrap". That figure belongs to the fix, not the bug: it is the
half-range limit of deadlinePassed(), which reads deadlines more than 2^31 ms
ahead as already passed, and the range over which a UINT32_MAX sentinel reads
as passed.
The naive compare's actual failure is an inversion lasting only while the
deadline sits on the far side of the wrap, so it is bounded by the interval:
the action fires immediately and loses its wait, or blocks for about the wait
it should have performed - days for the nRF52 flash-corruption backoff,
one skipped cycle for a seconds-long retransmit timer.
Comments and docs only; the ~24.8 day statements that correctly describe
deadlinePassed()'s own range are left as they were.
clod helped out here
* Restore a monotonic uptime clock and consolidate the wrap counters
Time::getMillisMonotonic() is the getMillis64() shape - a 32-bit wrap
counter carried across reads - promoted to the shared timebase, with
Time::getUptimeSecs() as the derived whole-seconds view. This deliberately
reverses the earlier removal of getMillis64(), and the distinction matters:
removal was right for a lazily-read accumulator with one rare caller, where
a 49.7-day gap between reads silently swallowed a wrap. Here every read is
the poll and AirTime::runOnce() guarantees one per second; the missed-wrap
contract is pinned by a test rather than left as a footnote.
Three private wrap counters collapse into it:
- AirTime::syncNow() takes its seconds from Time::getUptimeSecs() and drops
its lastSyncMsec checkpoint; window rotation is unchanged.
- DeviceTelemetryModule loses refreshUptime()/uptimeWrapCount/uptimeLastMs;
uptime_seconds comes from Time::getUptimeSecs(), which also removes the
0.296s-per-wrap truncation of (0xFFFFFFFF / 1000) * wraps. Its two
interval checks move to Throttle::hasElapsed().
- HostMetricsModule's copies of those members were never read (its uptime
comes from /proc/uptime) - deleted.
Not ISR-safe (unguarded mutable carry): ISRs keep using getMillis(), which
stays a pure read. Audited: no interrupt-context file reads getTime(),
getValidTime(), or the new accessors.
test/native-suite-count 44 -> 45: the bump for test_uptime_clock was lost
in a branch history rewrite, leaving every later value off by one -
run-tests.sh reports AMBER and CI's suite-count-check fails on the current
push until this correction.
* Anchor the wall clock in monotonic milliseconds
getTime() computed elapsed-since-time-set as a 32-bit millis() delta, so a
node that took time once and stayed up past 49.7 days reported a wall clock
one full cycle in the past - and last_heard, rx_time, message and position
stamps all inherited it. The anchor is now the 64-bit monotonic count
(timeStartMsec -> timeStartMs64) and the elapsed term is computed in 64-bit,
so the wall clock is exact at any uptime.
All six anchor writers follow: the five hardware-RTC read branches and
perhapsSetRTC(), which keeps a truncated 32-bit copy of the same instant for
its Throttle-checked rate-limit stamps. The test seams anchor the same way.
Two native regression tests drive getTime() across the wrap through the
Time seam - one anchored before the wrap and read after it, one anchored
after a counted wrap - with the test epoch derived from BUILD_EPOCH so the
plausibility window cannot rot as the build date advances.
* Stamp the rx_time placeholder in monotonic uptime seconds
computeRxTimeStamp() stamped Time::getMillis() when the clock was untrusted,
and reconcilePendingRxTimes() back-calculated with a 32-bit millis() delta -
correct within one wrap, but a placeholder older than 49.7 days aliased to a
small elapsed value and reconciled to a plausible-but-wrong recent epoch:
the exact failure has_rx_time exists to prevent, reachable by an ordinary
unattended router whose phone connects two months in.
The placeholder is now Time::getUptimeSecs(). Both stamps come off the
monotonic counter, so the elapsed term is exact at any age and the aliasing
window is gone outright rather than widened. If elapsed somehow exceeds the
epoch itself, the packet stays un-dated (absent, never wrong) instead of
clamping to a pre-1970 value. Defence in depth: a placeholder that leaks
needs ~50 years of uptime to cross MIN_PLAUSIBLE_EPOCH, where milliseconds
took 18.3 days.
The stream-API reconciliation tests keep their scenarios with the placeholder
unit switched, and ScopedTimeFixture resets the monotonic carry so uptime
seconds are deterministic per case.
* Date nodes heard before the clock arrives, without polluting last_heard
A node first heard while the wall clock was untrusted got no last_heard at
all, and nothing backfilled it once time arrived - the phone showed "Last
heard: unknown" for a node it had just announced. The arrival instant now
waits in a RAM-only sidecar (NodeNum -> uptime seconds, 32 slots,
reuse-oldest - the RouteHealth shape) and is converted to a real epoch on
the clock-becoming-trusted transition, beside the existing rx_time
reconciliation. last_heard itself never holds anything but a real epoch or
0: it persists to flash and the warm tier, where an uptime-relative value
would be meaningless after reboot.
The sidecar's write sites are updateFrom()'s no-trusted-clock path (the
rx_time placeholder already carries the arrival instant, so this is a store,
not a second clock read) and addFromContact's anti-eviction stamps, which
previously wrote a bare getTime() - boot-relative seconds on a clockless
node, the exact value lastHeardIsWallClock() exists to catch. Eviction
ranking honours the stamps: heard-this-boot outranks every stored epoch,
ordered among themselves, so a stamped contact is not the first victim.
PhoneAPI re-reads last_heard at nodeinfo send time: a record prefetched
before the clock became trusted can carry 0 while the store has since been
backfilled, and re-reading at the pop makes handshake ordering (time-set vs
node-list download) irrelevant. Backfill never moves last_heard backwards
and skips the pathological elapsed-exceeds-epoch case. A node evicted to
the warm tier before time arrives is still absorbed with last_heard 0 -
same as before, bounded to the untrusted window.
* Update the agent docs for the monotonic timebase
The conventions bullet asserted there is deliberately no 64-bit millis; the
monotonic uptime clock restored for timestamps changes that contract. State
the split explicitly: Throttle for deadlines and intervals (no carry state),
Time::getMillisMonotonic()/getUptimeSecs() for timestamps, polled by
construction and not ISR-safe.
* Publish the monotonic wrap carry from a single writer
getMillisMonotonic() was a read-modify-write on two unguarded statics, and it
is reached off the main loop: the nRF52 Bluefruit task via
onFromRadioAuthorize() -> PhoneAPI::getFromRadio -> getValidTime(), and the
portduino civetweb workers via the same path. Two readers interleaving inside
the wrap window could each increment the carry, putting every uptime and
wall-clock reading 2^32 ms ahead for the rest of the boot - a permanent ~49.7
day jump in rx_time, last_heard and ClientNotification.time.
Readers no longer write. serviceMonotonic() publishes a snapshot behind a
seqlock and is the only writer; a reader adds its own unsigned elapsed time to
that snapshot, which is exact across the wrap, so it never inspects the
boundary and cannot miscount it. The main loop publishes every iteration, so
the once-per-49.7-days obligation now has the whole window of margin instead of
resting on an instruction-wide race.
AirTime was the guaranteed poller and is now a pure reader, so the two airtime
wrap tests step the clock the way loop() does. The test clock itself is atomic
so a suite can drive it from one thread while others read.
* Re-arm the GPS ephemeris hold when none is in force
The rollover sweep guarded the hold re-arm with `fixHoldEnds != 0 &&`, which
reads like the sentinel rule but inverts this site. The comparison it replaced,
`(fixHoldEnds + GPS_THREAD_INTERVAL) < millis()`, was always true when nothing
was armed - that was the point, since 0 means "not holding" and so is a reason
to arm. With the guard, a publish that cleared the hold without sleeping (the
`shouldPublish && !tooLong && !holdExpired` path, which does not call down())
left hasValidLocation set and prev_fixQual non-zero, so no disjunct held:
nothing re-armed, nothing published, and the receiver stayed powered at the
200ms poll until searchedTooLong() fired.
State the question positively instead. fixHoldInForce() is the only place the
sentinel is interpreted, and both of runOnce()'s decisions derive from it - the
asymmetry is now visible rather than implied, since arming does not require a
prior hold but expiring does. Its `!= 0` test is not redundant with the
arithmetic: deadlinePassed() is an unsigned half-range test, so past 2^31 ms of
uptime the sentinel reads as a deadline ~24.9 days in the future.
Kept beside its caller rather than in a header; the native test build compiles
GPS.cpp, so the suite declares the prototypes.
Also converts the getACK() wait to isWithinTimespanMs(start, interval): it has
both the start instant and the interval in hand, which gives the full 49.7-day
range instead of 24.8 days ahead, and takes its anchor from Time::getMillis()
so the wait is injectable.
* Date the NodeInfo reply window in uptime seconds
The 12h reply-suppression stamp regressed from wrap-immune 64-bit seconds to
raw 32-bit milliseconds, and pruneLastNodeInfoCache() evicts only by node count
and DB membership - never by age. A stable mesh under the node cap therefore
keeps every stamp indefinitely, and once uptime passes 49.7 days an old one
aliases back into the window: `now - stamp` computes as ~0 and a legitimate
NodeInfo request goes unanswered for up to 12h. It self-heals and repeats once
per wrap cycle.
Store Time::getUptimeSecs() instead, which does not wrap for 136 years, and
drop the millisecond conversion the previous shape needed. Entries past the
window are now evicted too: they can only ever decide "don't suppress".
N8-N11 cover the window from both sides, and N10 pins the regression - it needs
a full 2^32 ms of uptime to elapse, not merely a crossing of the boundary,
because that is when a millisecond stamp reads as "answered this instant".
tearDown() now restores the injected clock and C14's region and TX bucket. A
failing assertion aborts the test body, so restoring at the end of it leaked
that state into every later case.
* Update the agent docs for the single-writer clock and sentinel direction
Two rules the preceding three commits changed.
The monotonic clock is no longer maintained by whoever happens to read it:
serviceMonotonic() is the only writer, readers are pure, and calling it from
anywhere but the main loop reintroduces the double-count.
The sentinel guidance gained the half it was missing. It named UINT32_MAX as a
sentinel while prescribing an idiom that only covers 0, and it assumed the
sentinel always means "suppress" - at the GPS fix-hold site it meant "fire",
which is how that regression passed review looking like the rule.
* Name the fix-hold expiry predicate and arm it from the injected clock
holdJustExpired() gives the second reading of the fixHoldEnds sentinel a
name beside the first, so both are pinned by test/test_gps_fix_hold/ and
neither can be respelled at the call site. The old inline form could not
be tested: written as a literal, its guard folds at compile time and the
assertion asserts nothing.
The arm site used bare millis() while the evaluation reads the Throttle
clock; same value in production, but it kept that write out of reach of
Time::setTestMillis(). Remap a deadline that lands on 0, which would
otherwise read as no hold at all.
* Share the extend formula between the clock's reader and writer
getMillisMonotonic() and serviceMonotonic() carried byte-identical wrap
arithmetic. A one-sided edit to either would drift the published carry
from what readers report, so keep one copy.
* Trim the NodeInfo dedup comment to the house limit
* todo note for potential future imrpovments
* fix some simple deadlines
* Trim the hold-expiry test comment to the house limit
* Fix non-blocking uptime publication and pre-clock recency edges (#29)
* fix(time): avoid blocking monotonic readers
* test(time): make paused-publisher check deterministic
* fix(time): address review portability gaps
* Init the eviction sentinel to the newest possible recency
EvictionRecency{} is {0, false}, which evictionRecencyOlder() ranks as older than
every candidate: without the oldestIndex/oldestBoringIndex guards nothing would
ever be selected and a full node DB would stop evicting entirely.
Init to the genuine maximum instead, so the sentinel is correct on its own. The
index guards stay: two independent reasons the scan is right beats one.
* Keep the deadline-guard check name branch protection matches
The guard was widened to cover Time::getMillis() and unqualified getMillis(),
and renamed to suit. Upstream branch protection matches required checks by name,
so a rename means the old name never reports and merges block on a check that
will never arrive.
Widen the guard, keep the name; the descriptive text carries the broader scope.
* Correct native-suite-count to 47 after the develop merge
Upstream #11293 added test_nmea_wpl and took develop's count to 43; this branch
had independently reached 46. Merging develop resolved the counter textually,
keeping 46, while the directory set became the union of both sides at 47.
The suite-count CI gate fails on the mismatch, and it gates the native test jobs,
so the tests themselves were being skipped.
* test(uptime): make the wrap fall where the comment says it does
The concurrent-reader case started at 0xFFFFF000, leaving 0x1000 to the wrap, so
the 0x800 advance annotated "cross the wrap" fell short and the wrap actually
happened during the following 60s advance.
Start at 0xFFFFF800 instead, so the first advance lands exactly on the wrap while
the readers are running and the second is the ordinary time after it - the shape
both comments already described. Total elapsed is unchanged, so the closing
assertion still holds.
* Respond to human comments
* Did I ever tell you about the time I went to Shelbyville? I wore an onion on my belt, which was the style at the time.
* Convert the I2S nag deadline develop dragged in
The HAS_I2S_SPEAKER_NRF52 RTTTL block arrived from develop with a raw
nagCycleCutoff >= millis(), which the deadline guard rejects. Use the same
Throttle::deadlinePassed() form as the two sibling paths in this function.
* Arm the LittleFS format guard with a flag, not a zero timestamp
preFSBegin() runs in the first millisecond of boot, so millis() can legitimately
return 0 there. Both readers of last_format_ms treated 0 as "nothing formatted
this boot", which would skip the repeat-corruption escalation and let a dead
flash reformat-loop instead of reporting FLASH_CORRUPTION_UNRECOVERABLE.
* Note the single-thread contract on AirTime
* Note the AirTime locking TODO, and tighten the thread note
The two constant getters are not constrained, and getSilentMinutes() reads the
buckets without rotating them, so "the accessors mutate" was not accurate.
* trunk: ignore trufflehog false positives on millis-wrap test constants
test_throttle and test_uptime_clock pin dense clusters of hex boundary
constants (0xFFFFFF00u and neighbors) to exercise 32-bit millis()
rollover. trufflehog's Lob detector stitches nearby hex literals into
one candidate string, and the result happens to match a Lob API key
shape - not a secret, just test fixtures.
Same pattern already used for the gitleaks/nodedb-fixture false
positive in this file.
---------
Co-authored-by: nightjoker7 <mattdeering7@gmail.com>
Co-authored-by: Clive Blackledge <clive@ansible.org>
Co-authored-by: Benjamin Faershtein <119711889+RCGV1@users.noreply.github.com>
Co-authored-by: Thomas Göttgens <tgoettgens@gmail.com>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
|