Software

Software

IOWeaver, the protocol analyzer and grapher; RockFace, the firmware debugger; mx3 for parts and assembly.

Three tools for the bench: one plots what a device says, one debugs the firmware that says it, one keeps the parts and the boards in order. draft

IOWeaver

FIG. 1IOWeaver: protocol analyzer and grapher

A protocol analyzer and grapher for the desktop: it records, decodes, shows on a timeline and sends the traffic of a device. It is the GUI for Pinsmith and Portwright and ships with them.

Record
I2C, CAN, UART, GPIO, WireWeaver
Decode
the same protocols
Show
on a timeline, and plot
Send
back to the device

RockFace

RockFace window: timeline view, log view and defmt counters
FIG. 2RockFace: firmware debugger, timeline and logs

A tracer and debugger for embedded firmware, for systems that are running. Commercial release: TBD.

Logs
from one board or a fleet
Graphs
stack and CPU use
Profiling
zero-cost
Timeline
sleep spans and events

mx3

The electronics side of an in-house lab system: from a BOM to parts on the shelf to an assembled board.

mx3 PCB Assemble tab: each BOM line with its designators, the part to place and the stock left
FIG. 3mx3: guided PCB assembly
mx3 BOM Editor tab: a capacitor line with its part choices, stock and Digi-Key parameters
FIG. 4mx3: BOM editor
mx3 BOM Cost tab: a chart of the cost per BOM line and a table of each line's unit price from 1 to 5k pieces
FIG. 5mx3: BOM cost at every price break
  • BOM import from CSV or XLSX, with each line mapped to a part automatically.
  • A BOM editor: part choices, variants, stock and box per line, cost with a price chart, a purchase list.
  • A parts database with typed parameters and parametric search.
  • Digi-Key pricing and stock for every part, with their history, kept fresh by a background crawler.
  • A warehouse: boxes in boxes, stock, photos and a barcode scanner.
  • Guided manual PCB assembly: it plans the placement and records every part that moves.

mx3 runs in VHRD’s own lab today. It may be opened up or offered as a product later.

From the terminal

Real runs of the command-line tools — cnt, WireWeaver’s ww, embedded_bedrock’s bedrock-lint and ExactUnits — recorded once and kept as text, colours and all. Under each one: the version, what it ran on and the day it was recorded.

cnt tui --chip STM32H725IGKx110×32

                                         Counters · 11 of 11 non-zero
╭ 📍  RAM counters at 0x24000178, 68 of 128 B used ───────────────────────────────────────────────────────────╮
│  Severity  Name           Value         Type  Location                                                     │
│                                                                                                            │
│▶ ● INFO    adc/charge        25686 mAs  u64   b135_demo @ src/main.rs:153                                  │
│  ◆ DEBUG   adc/samples          212600  u64   b135_demo @ src/main.rs:150                                  │
│  ✖ ERROR   can/bus_off               1  u32   b135_demo @ src/main.rs:127                                  │
│  ● INFO    can/frames             9592  u64   b135_demo @ src/main.rs:139                                  │
│  ◇ TRACE   ctrl/busy        4672452 µs  u64   b135_demo @ src/main.rs:186                                  │
│  ▲ WARN    ctrl/overruns             9  u32   b135_demo @ src/main.rs:187                                  │
│  ● INFO    ctrl/ticks            21269  u64   b135_demo @ src/main.rs:179                                  │
│  ● INFO    usb/RxPackets         25108  u64   b135_demo UsbEvent @ src/main.rs:40                          │
│  ◆ DEBUG   usb/RxBytes          12.7MB  u64   b135_demo UsbEvent @ src/main.rs:40                          │
│  ▲ WARN    usb/CrcError              3  u32   b135_demo UsbEvent @ src/main.rs:40                          │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
╭ 📍  BKP counters at 0x58004050, 4 of 16 B used ─────────────────────────────────────────────────────────────╮
│  Severity  Name           Value         Type  Location                                                     │
│                                                                                                            │
│  ● INFO    boots                     2  u32   b135_demo @ src/main.rs:50                                   │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
╭ 📜  defmt logs ─────────────────────────────────────────────────────────────────────────────────────────────╮
│0.000016 INFO  b135_demo up: USB, CAN at 500 kbit/s, ADC at 10 kS/s, control loop at 1 kHz @ src/main.rs:51 │
│4.000144 INFO  USB: host opened the bulk endpoint, firmware image transfer started @ src/main.rs:88         │
│12.000217 INFO  ADC: +5V0_SYS 4.98 V, load 146 mA @ src/main.rs:158                                         │
│17.000178 ERROR CAN: bus-off after 248 TX errors, recovering @ src/main.rs:128                              │
│19.000175 INFO  CAN: back on the bus after 128 x 11 recessive bits @ src/main.rs:131                        │
│                                                                                                            │
│                                                                                                            │
│                                                                                                            │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

q quit  ↑↓/jk select  g/G first/last  Tab switch table  r reset RAM  R reset BKP  c clear logs
FIG. 6cnt tui: a board's event counters and its firmware's defmt logs through a debug probe — here, cnt's demo firmware on a B135: a USB transfer, a CAN bus-off and a USB bus reset that clears its group · cnt_cli 0.5.1 (8b1e1db) · a B135 board through a J-Link, running cnt's demo firmware (examples/b135_demo) · recorded 6 Nov 2026

ww api tree -d examples/all_gpio_api160×1644×30

$ ww api tree -d examples/all_gpio_api
trait AllGpioApi all_gpio_api@0.1.0
└─ 0 #[claim(exclusive)] impl port[]: ww_gpio::Bank
   ├─ 0 impl pin[]: ww_gpio::Pin
   │  ├─ 0 fn set_output_level(level: Level)
   │  ├─ 1 fn toggle()
   │  ├─ 2 fn output_level() -> Level
   │  ├─ 3 fn input_level() -> Level
   │  ├─ 7 #[errors(Common(Unsupported))] #[counted] stream event(enabled: IoPinEnabledEvents): IoPinEvent -> Result<(), GpioError>
   │  ├─ 8 #[errors(Common(Unsupported, InUse, NoPower, Refused), DifferentModes)] fn set_mode(mode: Mode, initial: Option<Level>) -> Result<(), GpioError>
   │  ├─ 9 fn mode() -> Mode
   │  ├─ 10 #[errors(Common(Unsupported))] rw property pull: Pull, write error: GpioError
   │  ├─ 11 #[errors(Common(Unsupported))] rw property speed: Speed, write error: GpioError
   │  └─ 12 fn capabilities() -> PinCapabilities
   ├─ 8 fn capabilities() -> BankCapabilities
   ├─ 9 #[errors(Common(Unsupported, Refused))] fn configure(config: BankConfig) -> Result<BankConfig, GpioError>
   └─ 10 fn name() -> String
$ ww api tree -d examples/all_gpio_api
trait AllGpioApi all_gpio_api@0.1.0
└─ 0 #[claim(exclusive)] impl port[]: ww_gpi
o::Bank
   ├─ 0 impl pin[]: ww_gpio::Pin
   │  ├─ 0 fn set_output_level(level: Level)
   │  ├─ 1 fn toggle()
   │  ├─ 2 fn output_level() -> Level
   │  ├─ 3 fn input_level() -> Level
   │  ├─ 7 #[errors(Common(Unsupported))] #[
counted] stream event(enabled: IoPinEnabledE
vents): IoPinEvent -> Result<(), GpioError>
   │  ├─ 8 #[errors(Common(Unsupported, InUs
e, NoPower, Refused), DifferentModes)] fn se
t_mode(mode: Mode, initial: Option<Level>) -
> Result<(), GpioError>
   │  ├─ 9 fn mode() -> Mode
   │  ├─ 10 #[errors(Common(Unsupported))] r
w property pull: Pull, write error: GpioErro
r
   │  ├─ 11 #[errors(Common(Unsupported))] r
w property speed: Speed, write error: GpioEr
ror
   │  └─ 12 fn capabilities() -> PinCapabili
ties
   ├─ 8 fn capabilities() -> BankCapabilitie
s
   ├─ 9 #[errors(Common(Unsupported, Refused
))] fn configure(config: BankConfig) -> Resu
lt<BankConfig, GpioError>
   └─ 10 fn name() -> String
FIG. 7WireWeaver's ww api tree: every method, stream and property of a device API with its number on the wire — here, the GPIO example's ports of pins · ww 0.5.0 (46d60964, committed 5 Nov 2026 14:11) · wire_weaver at v0.5.0-1044-g5be6ed30 · recorded 6 Nov 2026

ww api check compatible; ww api check breaking220×1744×55

$ ww api check compatible
blinky_api 0.1.0 (compatible/api_snapshots/blinky_api_0_1_0.ron) -> 0.2.0 (source): breaking changes
  needs: wire w1.0 -> w1.1 or later (a new wire revision), crate 0.1.0 -> 0.2.0 or later (breaking position (minor before 1.0, major after))
  breaking (Rust only, the wire is compatible): BlinkyApi::led_on: takes 0 argument(s) on the old version, 1 on the new version: the same bytes on the wire, but other parameters for handlers and client calls
      rule: arguments added at the end with a default, or promoted to a #[flatten] struct, keep the wire, but change the Rust signatures of handlers and client calls (docs/evolution/rules.md#arguments)
Error: blinky_api declared wire version w1.0, the new version declares none.
Declare the wire version in `[package.metadata.wire_weaver] wire` of compatible/Cargo.toml, w1.1 or later (docs/evolution/versions.md#the-wire-version).

$ ww api check breaking
blinky_api 0.1.0 (breaking/api_snapshots/blinky_api_0_1_0.ron) -> 0.1.1 (source): breaking changes
  needs: wire w1.0 -> w2.0 or later (a new wire line), crate 0.1.0 -> 0.2.0 or later (breaking position (minor before 1.0, major after))
  breaking: BlinkyApi::led_on: argument `brightness` is not sent by the old version and has no #[default]
      rule: a resource works between the old and the new version, both ways (old client with a new device, new client with an old device) (docs/evolution/rules.md#api)
Error: blinky_api 0.1.0 -> 0.1.1 has breaking changes, which requires a bump of the breaking position (minor before 1.0, major after): 0.2.0 or later.
Bump `version` in breaking/Cargo.toml to 0.2.0 or later (docs/evolution/versions.md#your-apis-version); once that version is released, record it with `ww api save breaking` and commit api_snapshots/.
blinky_api declared wire version w1.0, the new version declares none.
Declare the wire version in `[package.metadata.wire_weaver] wire` of breaking/Cargo.toml, w2.0 or later (docs/evolution/versions.md#the-wire-version).
$ ww api check compatible
blinky_api 0.1.0 (compatible/api_snapshots/b
linky_api_0_1_0.ron) -> 0.2.0 (source): brea
king changes
  needs: wire w1.0 -> w1.1 or later (a new w
ire revision), crate 0.1.0 -> 0.2.0 or later
 (breaking position (minor before 1.0, major
 after))
  breaking (Rust only, the wire is compatibl
e): BlinkyApi::led_on: takes 0 argument(s) o
n the old version, 1 on the new version: the
 same bytes on the wire, but other parameter
s for handlers and client calls
      rule: arguments added at the end with
a default, or promoted to a #[flatten] struc
t, keep the wire, but change the Rust signat
ures of handlers and client calls (docs/evol
ution/rules.md#arguments)
Error: blinky_api declared wire version w1.0
, the new version declares none.
Declare the wire version in `[package.metada
ta.wire_weaver] wire` of compatible/Cargo.to
ml, w1.1 or later (docs/evolution/versions.m
d#the-wire-version).

$ ww api check breaking
blinky_api 0.1.0 (breaking/api_snapshots/bli
nky_api_0_1_0.ron) -> 0.1.1 (source): breaki
ng changes
  needs: wire w1.0 -> w2.0 or later (a new w
ire line), crate 0.1.0 -> 0.2.0 or later (br
eaking position (minor before 1.0, major aft
er))
  breaking: BlinkyApi::led_on: argument `bri
ghtness` is not sent by the old version and
has no #[default]
      rule: a resource works between the old
 and the new version, both ways (old client
with a new device, new client with an old de
vice) (docs/evolution/rules.md#api)
Error: blinky_api 0.1.0 -> 0.1.1 has breakin
g changes, which requires a bump of the brea
king position (minor before 1.0, major after
): 0.2.0 or later.
Bump `version` in breaking/Cargo.toml to 0.2
.0 or later (docs/evolution/versions.md#your
-apis-version); once that version is release
d, record it with `ww api save breaking` and
 commit api_snapshots/.
blinky_api declared wire version w1.0, the n
ew version declares none.
Declare the wire version in `[package.metada
ta.wire_weaver] wire` of breaking/Cargo.toml
, w2.0 or later (docs/evolution/versions.md#
the-wire-version).
FIG. 8ww api check: checks an API against the snapshot from its last release and judges whether the version bump is enough — a trailing argument on led_on with a default stays wire-compatible even though the Rust signature changed, so the check still asks for a bump but says the wire itself is fine; one without a default breaks an old client's bytes outright, and the check names the bump it needs · ww 0.5.0 (43f8e1b, committed 18 Nov 2026 09:04) · wire_weaver's blinky_api example, saved at 0.1.0: a trailing #[default] argument bumped to 0.2.0, a required one left at 0.1.1 (tools/term/ww_evolution) · recorded 18 Nov 2026

bedrock-lint check240×8444×195

$ bedrock-lint check
error[STK-1]: no linker set for thumbv7em-none-eabihf: it links with rust-lld, not flip-link; a stack overflow runs into .data and .bss unnoticed.
  --> .cargo/config.toml:4
   | [target.thumbv7em-none-eabihf]
   = fix: add `linker = "flip-link"` under `[target.<triple>]` in .cargo/config.toml (cargo install flip-link); a `cargo:rustc-linker` line in build.rs does nothing (STK-2)
   = STK-1: flip-link not the linker: linker = "flip-link" under [target.<triple>] in .cargo/config.toml (bedrock-lint explain STK-1)

error[FLT-1]: no HardFault handler: cortex-m-rt's default spins in `loop {}`, so a fault hangs the board until someone power-cycles it
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: save a crash record (bedrock_crash::capture_fault) and reset (bedrock_crash::reset(), SCB::sys_reset()) instead of spinning
   = FLT-1: HardFault handler that spins (loop {}) in a shipped build: record (crash capture) and reset (bedrock-lint explain FLT-1)

error[FLT-7]: no watchdog: a hang in a task or a stuck bus stays hung until someone power-cycles the board
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: start the independent watchdog at boot and feed it from a supervisor that checks the tasks made progress
   = FLT-7: No watchdog, or a watchdog fed from an interrupt or from a loop not bounded by a timeout; window not calibrated to the real period (bedrock-lint explain FLT-7)

warning[IRQ-8]: no #[exception] fn DefaultHandler in the source seen: an interrupt that fires on an unbound vector goes unnoticed instead of being counted (estimate)
  --> Cargo.toml:13
   | cortex-m-rt = "0.7"
   = fix: add #[exception] fn DefaultHandler(irqn: i16) that records irqn raw (a static, the crash record or cnt!)
   = IRQ-8: Unused interrupt vectors not caught: DefaultHandler must record the IRQ number (raw: a static, the crash record or a counter; not logging it, FLT-2) (bedrock-lint explain IRQ-8)

warning[FLT-4]: no UsageFault, MemoryManagement, BusFault handler: these faults escalate to HardFault, and the crash record cannot tell them apart
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: add `#[exception]` handlers for UsageFault, MemoryManagement and BusFault that record and reset, and enable them in SHCSR (`scb.enable(Exception::UsageFault)`, ...)
   = FLT-4: No UsageFault, MemManage and BusFault handlers (or not enabled in SHCSR) where the core has them (bedrock-lint explain FLT-4)

warning[FLT-5]: the reset cause is never read: a brown-out, a watchdog or a lockup reset looks like a normal start
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: read the reset flags at boot (RCC_CSR / RCC_RSR, RP CHIP_RESET and WATCHDOG.REASON), clear them, and report the cause (the template's last_reset.rs, `bedrock_crash::boot`)
   = FLT-5: Reset cause not read and reported at boot (power-on, brown-out, watchdog, software, pin, lockup) (bedrock-lint explain FLT-5)

warning[FLT-8]: no boot-loop rescue: a crash at start-up repeats on every boot, and only a probe gets the board back
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: count boots without a healthy mark in a backup register and fall back to the ROM bootloader after a few (bedrock_rom_boot::early, the template's boot.rs)
   = FLT-8: No boot-loop rescue: count boots without a healthy mark in a backup register, fall back to the ROM bootloader after a few (bedrock-lint explain FLT-8)

warning[FLS-4]: nothing checks the image in flash: a flipped bit runs until the code it is in executes
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: compute a CRC of the image at boot and periodically in a low-priority task, compare it with the one the build stored, and count a mismatch
   = FLS-4: No periodic CRC of the image in flash, no ECC on flash and RAM where the part has it (bedrock-lint explain FLS-4)

warning[CNT-1]: `Err(winner)` is handled in `cnt::buffers::Index::allocate` without a counter (candidate)
  --> ../../cnt/src/buffers.rs:116:13 error branch
   | Err(winner) => winner,
   = fix: count it: `cnt!(what_happened)` in the branch (cnt crate), so it shows in `cnt read`
   = CNT-1: Error branch (an Err handled, a timeout, a retry, a dropped frame, an overrun) without a cnt! counter: unexpected events must be counted (bedrock-lint explain CNT-1)

warning[CFG-7]: no bedrock_build dependency: the build carries no git SHA or dirty flag, so a dirty build flashed by mistake can't be told apart from a clean one by reading the device (estimate)
  --> Cargo.toml
   = fix: add bedrock_build as a build-dependency and call serialize_build_info() from build.rs (book/build_info.md)
   = CFG-7: Dirty build flashed without a mark: build info carries the git SHA and a dirty flag, read back from the device (bedrock-lint explain CFG-7)

suggestion[SIZ-4]: release profile is missing lto, panic = "abort": opt-level alone is the smallest lever (z vs 3 is 10-25 % speed, measure before choosing) (estimate)
  --> Cargo.toml:16
   | [profile.release]
   = fix: add lto, panic = "abort" to [profile.release]
   = SIZ-4: Release profile without codegen-units = 1, lto, panic = "abort", and an opt-level chosen by measurement (z vs 3 is the largest lever: Donguru 501.7 → 321.8 KiB, 10-25 % slower USB) (bedrock-lint explain SIZ-4)

Not run:
  DMA-1 no ELF given: pass --elf target/<triple>/release/<name>
  DMA-2 bedrock_fw.json names no chip (answers.chip)
  DMA-6 bedrock_fw.json names no chip (answers.chip)
  STK-3 no ELF given: pass --elf target/<triple>/release/<name>
  STK-5 no ELF given: pass --elf target/<triple>/release/<name>
  STK-6 no ELF given: pass --elf target/<triple>/release/<name>
  MEM-3 no ELF given: pass --elf target/<triple>/release/<name>
  MEM-6 no ELF given: pass --elf target/<triple>/release/<name>
  MEM-10 no ELF given: pass --elf target/<triple>/release/<name>
  SIZ-1 no ELF given: pass --elf target/<triple>/release/<name>
  SIZ-6 no ELF given: pass --elf target/<triple>/release/<name>
  ERR-3 no ELF given: pass --elf target/<triple>/release/<name>
  CFG-2 bedrock_fw.json names no chip (answers.chip)
  CFG-5 bedrock_fw.json names no chip (answers.chip)
  PERF-2 no ELF given: pass --elf target/<triple>/release/<name>
  PERF-3 no ELF given: pass --elf target/<triple>/release/<name>

counters_fw: 3 errors, 7 warnings, 1 suggestion · 81 checks ran, 16 not run · 97 of 116 automatic catalogue checks implemented
$ bedrock-lint check
error[STK-1]: no linker set for thumbv7em-no
ne-eabihf: it links with rust-lld, not flip-
link; a stack overflow runs into .data and .
bss unnoticed.
  --> .cargo/config.toml:4
   | [target.thumbv7em-none-eabihf]
   = fix: add `linker = "flip-link"` under `
[target.<triple>]` in .cargo/config.toml (ca
rgo install flip-link); a `cargo:rustc-linke
r` line in build.rs does nothing (STK-2)
   = STK-1: flip-link not the linker: linker
 = "flip-link" under [target.<triple>] in .c
argo/config.toml (bedrock-lint explain STK-1
)

error[FLT-1]: no HardFault handler: cortex-m
-rt's default spins in `loop {}`, so a fault
 hangs the board until someone power-cycles
it
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: save a crash record (bedrock_crash
::capture_fault) and reset (bedrock_crash::r
eset(), SCB::sys_reset()) instead of spinnin
g
   = FLT-1: HardFault handler that spins (lo
op {}) in a shipped build: record (crash cap
ture) and reset (bedrock-lint explain FLT-1)

error[FLT-7]: no watchdog: a hang in a task
or a stuck bus stays hung until someone powe
r-cycles the board
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: start the independent watchdog at
boot and feed it from a supervisor that chec
ks the tasks made progress
   = FLT-7: No watchdog, or a watchdog fed f
rom an interrupt or from a loop not bounded
by a timeout; window not calibrated to the r
eal period (bedrock-lint explain FLT-7)

warning[IRQ-8]: no #[exception] fn DefaultHa
ndler in the source seen: an interrupt that
fires on an unbound vector goes unnoticed in
stead of being counted (estimate)
  --> Cargo.toml:13
   | cortex-m-rt = "0.7"
   = fix: add #[exception] fn DefaultHandler
(irqn: i16) that records irqn raw (a static,
 the crash record or cnt!)
   = IRQ-8: Unused interrupt vectors not cau
ght: DefaultHandler must record the IRQ numb
er (raw: a static, the crash record or a cou
nter; not logging it, FLT-2) (bedrock-lint e
xplain IRQ-8)

warning[FLT-4]: no UsageFault, MemoryManagem
ent, BusFault handler: these faults escalate
 to HardFault, and the crash record cannot t
ell them apart
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: add `#[exception]` handlers for Us
ageFault, MemoryManagement and BusFault that
 record and reset, and enable them in SHCSR
(`scb.enable(Exception::UsageFault)`, ...)
   = FLT-4: No UsageFault, MemManage and Bus
Fault handlers (or not enabled in SHCSR) whe
re the core has them (bedrock-lint explain F
LT-4)

warning[FLT-5]: the reset cause is never rea
d: a brown-out, a watchdog or a lockup reset
 looks like a normal start
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: read the reset flags at boot (RCC_
CSR / RCC_RSR, RP CHIP_RESET and WATCHDOG.RE
ASON), clear them, and report the cause (the
 template's last_reset.rs, `bedrock_crash::b
oot`)
   = FLT-5: Reset cause not read and reporte
d at boot (power-on, brown-out, watchdog, so
ftware, pin, lockup) (bedrock-lint explain F
LT-5)

warning[FLT-8]: no boot-loop rescue: a crash
 at start-up repeats on every boot, and only
 a probe gets the board back
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: count boots without a healthy mark
 in a backup register and fall back to the R
OM bootloader after a few (bedrock_rom_boot:
:early, the template's boot.rs)
   = FLT-8: No boot-loop rescue: count boots
 without a healthy mark in a backup register
, fall back to the ROM bootloader after a fe
w (bedrock-lint explain FLT-8)

warning[FLS-4]: nothing checks the image in
flash: a flipped bit runs until the code it
is in executes
  --> src/main.rs:20:4
   | fn main() -> ! {
   = fix: compute a CRC of the image at boot
 and periodically in a low-priority task, co
mpare it with the one the build stored, and
count a mismatch
   = FLS-4: No periodic CRC of the image in
flash, no ECC on flash and RAM where the par
t has it (bedrock-lint explain FLS-4)

warning[CNT-1]: `Err(winner)` is handled in
`cnt::buffers::Index::allocate` without a co
unter (candidate)
  --> ../../cnt/src/buffers.rs:116:13 error
branch
   | Err(winner) => winner,
   = fix: count it: `cnt!(what_happened)` in
 the branch (cnt crate), so it shows in `cnt
 read`
   = CNT-1: Error branch (an Err handled, a
timeout, a retry, a dropped frame, an overru
n) without a cnt! counter: unexpected events
 must be counted (bedrock-lint explain CNT-1
)

warning[CFG-7]: no bedrock_build dependency:
 the build carries no git SHA or dirty flag,
 so a dirty build flashed by mistake can't b
e told apart from a clean one by reading the
 device (estimate)
  --> Cargo.toml
   = fix: add bedrock_build as a build-depen
dency and call serialize_build_info() from b
uild.rs (book/build_info.md)
   = CFG-7: Dirty build flashed without a ma
rk: build info carries the git SHA and a dir
ty flag, read back from the device (bedrock-
lint explain CFG-7)

suggestion[SIZ-4]: release profile is missin
g lto, panic = "abort": opt-level alone is t
he smallest lever (z vs 3 is 10-25 % speed,
measure before choosing) (estimate)
  --> Cargo.toml:16
   | [profile.release]
   = fix: add lto, panic = "abort" to [profi
le.release]
   = SIZ-4: Release profile without codegen-
units = 1, lto, panic = "abort", and an opt-
level chosen by measurement (z vs 3 is the l
argest lever: Donguru 501.7 → 321.8 KiB, 10-
25 % slower USB) (bedrock-lint explain SIZ-4
)

Not run:
  DMA-1 no ELF given: pass --elf target/<tri
ple>/release/<name>
  DMA-2 bedrock_fw.json names no chip (answe
rs.chip)
  DMA-6 bedrock_fw.json names no chip (answe
rs.chip)
  STK-3 no ELF given: pass --elf target/<tri
ple>/release/<name>
  STK-5 no ELF given: pass --elf target/<tri
ple>/release/<name>
  STK-6 no ELF given: pass --elf target/<tri
ple>/release/<name>
  MEM-3 no ELF given: pass --elf target/<tri
ple>/release/<name>
  MEM-6 no ELF given: pass --elf target/<tri
ple>/release/<name>
  MEM-10 no ELF given: pass --elf target/<tr
iple>/release/<name>
  SIZ-1 no ELF given: pass --elf target/<tri
ple>/release/<name>
  SIZ-6 no ELF given: pass --elf target/<tri
ple>/release/<name>
  ERR-3 no ELF given: pass --elf target/<tri
ple>/release/<name>
  CFG-2 bedrock_fw.json names no chip (answe
rs.chip)
  CFG-5 bedrock_fw.json names no chip (answe
rs.chip)
  PERF-2 no ELF given: pass --elf target/<tr
iple>/release/<name>
  PERF-3 no ELF given: pass --elf target/<tr
iple>/release/<name>

counters_fw: 3 errors, 7 warnings, 1 suggest
ion · 81 checks ran, 16 not run · 97 of 116
automatic catalogue checks implemented
FIG. 9embedded_bedrock's bedrock-lint reads a firmware crate and its build files and reports what could hang, crash or brick the board in the field, each with a fix — here, checking a small test firmware · bedrock-lint 0.1.0 (46ba0ee, commit 3 Nov 2026 14:43 UTC) · the cnt test firmware (cnt/fixtures/counters_fw) · recorded 6 Nov 2026

src/main.rsline 5

use exact_units::{Fixed, prefix::One, units::{Joule, Torque}};

fn main() {
    // angle is a dimension: a torque is N·m per radian, not an energy
    let _ = Fixed::<Torque, One, i32>::new(5).convert::<Joule, One, i32>();
}

cargo build -q300×3644×69

$ cargo build -q
error[E0080]: evaluation panicked: a Fixed converts only between units of the same dimension: Torque (N·m) is kg·m²·s⁻²·rad⁻¹, not Joule (J) which is kg·m²·s⁻²; they differ by rad⁻¹ — cross it with a named step: Torque × Radian is Joule
   --> ~/.cargo/git/checkouts/exact_units-0e04409aaab251a0/90d25ff/src/fixed.rs:445:25
    |
445 |         let c = const { Conversion::between::<U, P, U2, P2, T, T2>(false) };
    |                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ evaluation of `exact_units::Fixed::<exact_units::units::Torque, exact_units::prefix::One, i32>::convert::<exact_units::units::Joule, exact_units::prefix::One, i32>::{constant#0}` failed inside this call
    |
note: inside `exact_units::fixed::Conversion::between::<Torque, One, Joule, One, i32, i32>`
   --> ~/.cargo/git/checkouts/exact_units-0e04409aaab251a0/90d25ff/src/fixed.rs:907:13
    |
907 |             crate::conv_error::dim_mismatch::<U, U2>("a Fixed");
    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
note: inside `exact_units::conv_error::dim_mismatch::<Torque, Joule>`
   --> /rustc/58cf8f9c35b355199727627d07ae34a4e2846d67/library/core/src/panic.rs:57:9
    |
 57 |         $crate::panicking::panic_display(&$arg);
    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the failure occurred here
    |
   ::: ~/.cargo/git/checkouts/exact_units-0e04409aaab251a0/90d25ff/src/conv_error.rs:388:5
    |
388 |     panic!("{}", msg.as_str())
    |     -------------------------- in this macro invocation

note: erroneous constant encountered
   --> ~/.cargo/git/checkouts/exact_units-0e04409aaab251a0/90d25ff/src/fixed.rs:445:17
    |
445 |         let c = const { Conversion::between::<U, P, U2, P2, T, T2>(false) };
    |                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

note: the above error was encountered while instantiating `fn Fixed::<Torque, One, i32>::convert::<Joule, One, i32>`
 --> src/main.rs:5:13
  |
5 |     let _ = Fixed::<Torque, One, i32>::new(5).convert::<Joule, One, i32>();
  |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

For more information about this error, try `rustc --explain E0080`.
error: could not compile `units_demo` (bin "units_demo") due to 1 previous error
$ cargo build -q
error[E0080]: evaluation panicked: a Fixed c
onverts only between units of the same dimen
sion: Torque (N·m) is kg·m²·s⁻²·rad⁻¹, not J
oule (J) which is kg·m²·s⁻²; they differ by
rad⁻¹ — cross it with a named step: Torque ×
 Radian is Joule
   --> ~/.cargo/git/checkouts/exact_units-0e
04409aaab251a0/90d25ff/src/fixed.rs:445:25
    |
445 | ... { Conversion::between::<U, P, U2,
P2, T, T2>(false) };
    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^ evaluation of `exact_units
::Fixed::<exact_units::units::Torque, exact_
units::prefix::One, i32>::convert::<exact_un
its::units::Joule, exact_units::prefix::One,
 i32>::{constant#0}` failed inside this call
    |
note: inside `exact_units::fixed::Conversion
::between::<Torque, One, Joule, One, i32, i3
2>`
   --> ~/.cargo/git/checkouts/exact_units-0e
04409aaab251a0/90d25ff/src/fixed.rs:907:13
    |
907 | ...   crate::conv_error::dim_mismatch:
:<U, U2>("a Fixed");
    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^
note: inside `exact_units::conv_error::dim_m
ismatch::<Torque, Joule>`
   --> /rustc/58cf8f9c35b355199727627d07ae34
a4e2846d67/library/core/src/panic.rs:57:9
    |
 57 | ...   $crate::panicking::panic_display
(&$arg);
    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^ the failure occurred here
    |
   ::: ~/.cargo/git/checkouts/exact_units-0e
04409aaab251a0/90d25ff/src/conv_error.rs:388
:5
    |
388 |     panic!("{}", msg.as_str())
    |     -------------------------- in this
 macro invocation

note: erroneous constant encountered
   --> ~/.cargo/git/checkouts/exact_units-0e
04409aaab251a0/90d25ff/src/fixed.rs:445:17
    |
445 | ... = const { Conversion::between::<U,
 P, U2, P2, T, T2>(false) };
    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^

note: the above error was encountered while
instantiating `fn Fixed::<Torque, One, i32>:
:convert::<Joule, One, i32>`
 --> src/main.rs:5:13
  |
5 | ... = Fixed::<Torque, One, i32>::new(5).
convert::<Joule, One, i32>();
  |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^

For more information about this error, try `
rustc --explain E0080`.
error: could not compile `units_demo` (bin "
units_demo") due to 1 previous error
FIG. 10ExactUnits: converting a torque in N·m to an energy in joules does not compile — the angle is a dimension, and the message says so · rustc 1.99.1 (58cf8f9c3 2026-10-13) · tools/term/units_demo against exact_units at 90d25ff · recorded 6 Nov 2026