espp::CoreDump wraps the ESP-IDF flash core dump (espcoredump) in an
idiomatic espp API: has_core_dump(), summary() (esp_core_dump_summary_t),
format_report() (reset reason, panic reason, crashed task + PC, raw
backtrace addresses / RISC-V stack dump, addr2line decode hint with the
right toolchain prefix for CONFIG_IDF_TARGET, and brownout/watchdog hints
for resets that write no dump), plus raw image access (image_size(),
read_image(), erase()) for downloading the full ELF core dump over any
transport. All failures via std::error_code; degrades gracefully when
core-dump-to-flash is disabled.

espp::CoreDumpService serves that over any byte stream by reusing the ota
component's CRC-32 stream framing, with message types in a dedicated range
(0x40..0x43 requests / 0xC0..0xC4 replies: GET_SUMMARY/GET_SIZE/READ/ERASE
-> SUMMARY/SIZE/DATA/OK/ERROR). Construct with a send function; feed(bytes)
runs a resynchronizing parser (so the stream can also carry console text)
and handle_frame() supports an external parser. Unknown frame types are
ignored so the service coexists with other protocols on one stream.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

- coredump.hpp: explicitly include format.hpp (fmt::format is used in the
  header) and add the format component to REQUIRES
- coredump.hpp: validate the core-dump image address range lies within the
  coredump partition before computing the partition-relative offset, so an
  inconsistent address cannot underflow or read the wrong flash region
- coredump_service.hpp: normalize non-generic-category error codes to a
  std::errc value (default_error_condition, io_error fallback) so the
  on-wire ERROR code always matches the wire spec
- web console: default WebUSB filter now matches VID and PID (the 'show all
  USB devices' checkbox still lifts the filter), removing the unused
  DEFAULT_PID
- web console: extractAddresses() now includes the faulting PC first
  (deduplicated) alongside the backtrace addresses, matching its comment

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…config-constant findings

The static-analysis CI flagged: the example's INTENTIONAL crash triggers
(null-pointer store, divide-by-zero - crashing is their purpose), two
knownConditionTrueFalse findings that are constants only in the
coredump-disabled configuration cppcheck analyzes, and the
CONFIG_IDF_TARGET string-paste it cannot resolve. All suppressed inline
with justifications; cppcheck now reports 0 findings on the component and
the example still builds clean.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…utside lock, README API names)

- CoreDump::format_report(): report ALL abnormal reset reasons even when no
  core dump image is present (PANIC with a missing dump, UNKNOWN, power
  glitch, CPU lockup, ... in addition to the existing brownout/watchdog
  hints); empty string is now reserved for genuinely clean reasons
  (power-on / EXT / SW / deep-sleep / SDIO / USB / JTAG) via a new
  is_clean_reset_reason() helper, and the docs state precisely which.
- CoreDumpService: never invoke the user send callback under the internal
  mutex. handle_frame_locked() now performs the flash access and BUILDS the
  reply frame under the lock (send_error -> build_error); feed()/
  handle_frame() transmit the collected replies after unlocking, so a
  re-entrant transport (send path that triggers RX -> feed()) cannot
  deadlock. Documented the threading / send-callback contract.
- READMEs: the example routes the console with
  tinyusb_console_init(TINYUSB_CDC_ACM_0), not esp_tusb_init_console.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…B FIFO sizing, IDF >=5.2, webapp label)

- CoreDump: serialize all flash-touching methods (has_core_dump / summary /
  format_report / image_size / read_image / erase) with an internal mutex so
  one CoreDump can be shared by several CoreDumpService instances /
  transports without a READ on one racing an ERASE on the other; documented
  on the class and in CoreDumpService's shared-CoreDump note.
- example sdkconfig.defaults: size the TinyUSB vendor RX/TX and CDC TX FIFOs
  (4096) to hold a complete DATA reply frame (2,063 bytes with the web
  console's 2 KiB read chunks) so the stock write path cannot truncate
  replies; PR #742 additionally makes write_vendor wait-for-drain.
- idf_component.yml: require idf >=5.2 (first release with
  esp_core_dump_get_panic_reason(), used for the ELF crash report).
- example: replace the now-stale knownConditionTrueFalse inline suppression
  (which CI reports as unmatched since b7871e9) with a
  CONFIG_ESP_COREDUMP_ENABLE_TO_FLASH guard, robust across cppcheck versions.
- CoreDumpService: make build_error() const (cppcheck functionConst).
- webapp: correct the download-panel label — core.elf drops the flash header
  before the ELF magic; the trailing checksum bytes are retained.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…ough marker, handle_frame doc)

- webapp: navigator.usb.requestDevice() has no acceptAllDevices option
  (that's Web Bluetooth); use the required filters member with an empty
  list to show all devices
- example: end the Assert case with an explicit __builtin_unreachable()
  so there is no implicit fallthrough into DivideByZero
- coredump_service: handle_frame() @return doc now states the reply is
  only delivered when a send callback is configured (dropped otherwise)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>