pureboot: review-pass fixes to the host tool and device runner

pureboot.py: reject an empty image file with a clear error instead of
an IndexError deep in the vector-surgery planner; tighten the erase
docstring (order is irrelevant there — every target byte is the same
value, unlike a real flash where page 0 must go last).

pureboot_device.c: the GPIO bridge's bit_cycles used plain truncating
division where the firmware computes its own bit period with
round-to-nearest (uart.hpp: (Clock.hz + Baud.bd/2)/Baud.bd) — one
cycle off per bit on both tinies, harmless in practice but needless
drift against a firmware built to a different constant. Matched
exactly. Also clear the queued-bytes/decode-in-progress bridge state
on the test-only reset signal, so a future reset-mid-transfer scenario
can't feed a freshly reset chip bytes queued for its previous life.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AReSwkWkPX2A9Ym6grxRAh

.clang-format

@@ -7,9 +7,8 @@ TabWidth: 4
 UseTab: ForIndentation
 AlignEscapedNewlines: DontAlign
 AllowShortFunctionsOnASingleLine: Empty
-BreakTemplateDeclarations: Yes
+AlwaysBreakTemplateDeclarations: true
 BreakBeforeBraces: Custom
 BraceWrapping:
   AfterFunction: true
-InsertBraces: true
 ...

.clangd

@@ -1,38 +0,0 @@
-# Editor accommodations for the second frontend. No compilation database is
-# named here: this repo rides as a submodule in its consumers, and this file
-# travels with it — a consumer's own database then covers these sources, with
-# that project's loader flags. The checkout that is opened as a folder names
-# its build tree in .vscode/settings.json instead.
-CompileFlags:
-  Add:
-    # clang has no 24-bit integer and GCC's are keywords, not macros, so the
-    # editor needs a stand-in for avr::uint24_t. The next width up is the only
-    # one available — clang rejects _BitInt(24) on this target.
-    - -D__uint24=unsigned long
-    - -D__int24=long
-    # clangd forwards the driver's system includes but not its own header
-    # directory, so <stdint.h> resolves to avr-libc's, which still gates the
-    # limit and constant macros on the C++98 opt-in.
-    - -D__STDC_LIMIT_MACROS
-    - -D__STDC_CONSTANT_MACROS
-    # isr::emit spells a vector number into [[gnu::signal(N)]], which clang
-    # rejects rather than ignores — enough of them in one TU to reach the
-    # default limit of 19 inside the headers and truncate the parse.
-    - -ferror-limit=0
-  Remove:
-    # Codegen shaping the loader TUs carry and clang has no spelling for.
-    - -fira-algorithm=*
-    - -fno-split-wide-types
-    - -fno-tree-ter
-    - -fno-ivopts
-    - -fno-move-loop-invariants
-    # The build promotes warnings for the compiler that has to be right about
-    # them; in the editor the flag paints a second frontend's opinions in the
-    # colour reserved for things that do not compile.
-    - -Werror
-Diagnostics:
-  Suppress:
-    # clang's AVR `signal` attribute takes no arguments and it knows none of
-    # progmem, naked or OS_main. A misspelling is what the build is for.
-    - attribute_wrong_number_arguments
-    - unknown-attributes

.gitattributes

@@ -1,11 +1,8 @@
-# Line endings are the repository's, not the editing machine's: this checkout
-# is reached from two hosts, and a file rewritten by a Windows tool comes back
-# with every line changed unless something says otherwise. Naming the source
-# extensions left Markdown, Python, shell and CMake to whatever the writing
-# tool defaulted to, which is CRLF on one of the two.
-* text=auto eol=lf
-
-# Atmel Studio writes these and expects them back.
+*.h         eol=lf
+*.hpp       eol=lf
+*.c         eol=lf
+*.cpp       eol=lf
+.git*       eol=lf
 *.vcxproj*  eol=crlf
 *.cppproj   eol=crlf
 *.sln       eol=crlf

.gitignore

@@ -12,7 +12,6 @@ Debug
 
 # CMake / clangd
 /build/
-/local/
 compile_commands.json
 .cache/
 

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "libavr"]
-	path = libavr
-	url = ../libavr.git

.vscode/extensions.json

@@ -1,6 +0,0 @@
-{
-	"recommendations": [
-		"llvm-vs-code-extensions.vscode-clangd",
-		"ms-vscode.cmake-tools"
-	]
-}

.vscode/settings.json

@@ -1,36 +0,0 @@
-{
-	// clangd is the language server; the cpptools engine would parse every file
-	// a second time and disagree, since nothing tells it about a cross
-	// compiler.
-	"C_Cpp.intelliSenseEngine": "disabled",
-
-	// --query-driver lets clangd ask the cross compiler for its own system
-	// includes and target. The database is named here rather than in .clangd
-	// because that file travels with the driver into a consumer's submodule,
-	// where a build tree of this repo's own need not exist.
-	"clangd.arguments": [
-		"--compile-commands-dir=${workspaceFolder}/build/atmega328p-generated",
-		"--query-driver=**avr-g++*",
-		"--header-insertion=never"
-	],
-
-	// The presets are the build interface, and the toolchain file inside the
-	// libavr submodule is the one place the compiler is chosen. **No prefix is
-	// named here**: a committed file may not name a path that is true of one
-	// machine (libavr guidance rule 50), so the gitignored local/machine.cmake
-	// at this repository's root is where a checkout says where its toolchain
-	// is - one file, and it answers for both hosts.
-	"cmake.useCMakePresets": "always",
-	"cmake.configureOnOpen": true,
-	"cmake.options.statusBarVisibility": "compact",
-
-	"files.watcherExclude": {
-		"**/build/**": true,
-		"**/libavr/**": true
-	},
-
-	"files.associations": {
-		".clangd": "yaml",
-		".clang-format": "yaml"
-	}
-}

CMakeLists.txt

@@ -2,119 +2,84 @@ cmake_minimum_required(VERSION 3.28)
 
 project(tsb_libavr LANGUAGES CXX)
 
-# libavr rides as the pinned submodule; LIBAVR_ROOT (cache or environment)
-# overrides it for tandem development against a working tree. The toolchain
-# file comes from the submodule via CMakePresets.json either way.
+# libavr from a local checkout (LIBAVR_ROOT) or the forge; the toolchain file
+# comes from the same checkout via CMakePresets.json.
+include(FetchContent)
 if(NOT LIBAVR_ROOT AND DEFINED ENV{LIBAVR_ROOT})
 	set(LIBAVR_ROOT $ENV{LIBAVR_ROOT})
 endif()
-if(NOT LIBAVR_ROOT)
-	set(LIBAVR_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/libavr)
+if(LIBAVR_ROOT)
+	FetchContent_Declare(libavr SOURCE_DIR ${LIBAVR_ROOT})
+else()
+	FetchContent_Declare(libavr GIT_REPOSITORY git@git.blackmark.me:avr/libavr.git GIT_TAG main)
 endif()
-if(NOT EXISTS ${LIBAVR_ROOT}/CMakeLists.txt)
-	message(FATAL_ERROR "libavr not found at ${LIBAVR_ROOT} - run: git submodule update --init libavr")
-endif()
-add_subdirectory(${LIBAVR_ROOT} libavr-build)
-
-include(${LIBAVR_ROOT}/cmake/checks.cmake)
+FetchContent_MakeAvailable(libavr)
 
 if(PROJECT_IS_TOP_LEVEL)
 	add_compile_options(-Werror) # warnings are errors for the port's own code
 	enable_testing()
 
-	# Rules 11 and 33 over this repo's own sources. The oracle's assembly needs
-	# no exclusion: it is neither formatted nor ASCII-checked, being in neither
-	# glob, which is the right answer for a vendored reference whose text is
-	# the artifact.
-	libavr_format_test()
-
 	# The behavioral tests drive the real wire protocols over a simavr pty
-	# (as the host tools do) and actually flash the device. The runner is a
-	# host program built at configure time against libsimavr (C++23 - what the
-	# distribution's compiler speaks in full).
-	#
-	# **A host that cannot build it registers those tests anyway and skips
-	# them.** They used to be left out, which makes the suite a different size
-	# on a different machine - and a suite whose size is a property of the
-	# machine is one nothing can be compared against.
-	set(TSB_DEVICE ${CMAKE_BINARY_DIR}/tsb_device)
-	find_program(_host_cxx NAMES c++ g++)
-	set(_tsb_absent "${LIBAVR_NO_PYTHON}")
-	if(NOT _host_cxx)
-		set(_tsb_absent "no host C++ compiler on PATH, and the simavr device is a host program")
-	elseif(NOT _tsb_absent)
+	# (as the host tools do) and actually flash the device. The runners are
+	# host programs built at configure time against libsimavr; if they or
+	# Python are missing, only the size tests run.
+	find_program(_host_cc NAMES cc gcc)
+	find_package(Python3 COMPONENTS Interpreter)
+	if(_host_cc AND Python3_FOUND)
+		set(PB_DEVICE ${CMAKE_BINARY_DIR}/pureboot_device)
 		execute_process(
-		    COMMAND ${_host_cxx} -std=c++23 -Wall -Wextra -O2
-		            -I/usr/include/simavr -I/usr/include/simavr/parts
-		            -o ${TSB_DEVICE} ${CMAKE_CURRENT_SOURCE_DIR}/test/device.cpp
-		            -lsimavr -lsimavrparts -lelf
-		    RESULT_VARIABLE _dev_res ERROR_VARIABLE _dev_err)
-		if(NOT _dev_res EQUAL 0)
-			# One bounded line of it: this becomes a single argument on a
-			# command line, and the reading has to say what stopped the build
-			# rather than that something did.
-			string(REGEX REPLACE "[\r\n\t]+" " " _dev_err "${_dev_err}")
-			string(REPLACE ";" "," _dev_err "${_dev_err}")
-			string(LENGTH "${_dev_err}" _dev_len)
-			if(_dev_len GREATER 240)
-				string(SUBSTRING "${_dev_err}" 0 240 _dev_err)
-			endif()
-			set(_tsb_absent "test/device.cpp does not build here: ${_dev_err}")
+		    COMMAND ${_host_cc} -O2 -I/usr/include/simavr -I/usr/include/simavr/parts
+		            -o ${PB_DEVICE} ${CMAKE_CURRENT_SOURCE_DIR}/test/pureboot_device.c
+		            -lsimavr -lsimavrparts -lelf -lutil
+		    RESULT_VARIABLE _pbdev_res ERROR_VARIABLE _pbdev_err)
+		if(NOT _pbdev_res EQUAL 0)
+			message(STATUS "pureboot_device not built (${_pbdev_err}) — protocol tests skipped")
+			unset(PB_DEVICE)
+		endif()
+		if(LIBAVR_MCU STREQUAL "atmega328p")
+			set(TSB_DEVICE ${CMAKE_BINARY_DIR}/tsb_device)
+			execute_process(
+			    COMMAND ${_host_cc} -O2 -I/usr/include/simavr -I/usr/include/simavr/parts
+			            -o ${TSB_DEVICE} ${CMAKE_CURRENT_SOURCE_DIR}/test/device.c
+			            -lsimavr -lsimavrparts -lelf
+			    RESULT_VARIABLE _dev_res ERROR_VARIABLE _dev_err)
+			if(NOT _dev_res EQUAL 0)
+				message(STATUS "tsb_device not built (${_dev_err}) — protocol tests skipped")
+				unset(TSB_DEVICE)
+			endif()
 		endif()
-	endif()
-	libavr_launcher(_tsb_python "${_tsb_absent}" ${Python3_EXECUTABLE})
-	if(_tsb_absent)
-		message(STATUS "the protocol tests skip here - ${_tsb_absent}")
 	endif()
 endif()
 
-# The ELF is only a container (symbols, section headers) and is never flashed -
-# and the host tool's load_image() dispatches on extension, so handing it one
-# would silently program the header bytes. Every loader image therefore gets
-# both flashable forms beside it at link time: .hex for avrdude, and .bin for
-# the host tool's raw path (which is what the reloc and update tests convert to
-# on the fly). .eeprom is dropped - EEPROM content is its own update.
-function(add_image_outputs name)
-	add_custom_command(TARGET ${name} POST_BUILD
-	                   COMMAND ${CMAKE_OBJCOPY} -O ihex -R .eeprom
-	                           $<TARGET_FILE:${name}> $<TARGET_FILE:${name}>.hex
-	                   COMMAND ${CMAKE_OBJCOPY} -O binary -R .eeprom
-	                           $<TARGET_FILE:${name}> $<TARGET_FILE:${name}>.bin)
-endfunction()
-
-# The TinySafeBoot protocol reimplemented on libavr in variants that trade
+# The TinySafeBoot protocol reimplemented on libavr in three variants that trade
 # clarity for size. Each links into the ATmega328P boot section (BOOTSZ selects
-# its size; BOOTRST vectors a reset to its base) with -nostartfiles - a polled
-# loader has no use for the crt or the vector table. The entry sits in
-# .vectors, laid first, and runs - avr::startup::entry on the policy tier,
-# the experiment tiers' own naked stubs elsewhere, each documented in its
-# source. The boot base is FLASHEND+1 minus the section size; the linker
-# section-start and the source's boot_bytes agree. tsb_app is
+# its size; BOOTRST vectors a reset to its base) with -nostartfiles — a polled
+# loader has no use for the crt or the vector table. The naked entry sits in
+# .vectors, laid first, and runs. The boot base is FLASHEND+1 minus the section
+# size; the linker section-start and the source's boot_bytes agree. tsb_app is
 # the application's reset vector, pinned to 0 here so the loaders jump to a
 # named function; --pmem-wrap-around lets relaxation turn that absolute jump
 # into the wrapped rjmp AVR's modulo-flash PC actually executes.
-# All four implement the full oracle feature set (see oracle/README.md):
+# All three implement the full oracle feature set (see oracle/README.md):
 # watchdog bail, one-wire half-duplex, config-page activation timeout, password
 # gate, emergency erase, config/flash/EEPROM read-write. They differ only in how,
-# and the size gradient is the cost of that "how".
-#   tsb_asm    - the tricks tier's C++ with exactly two routines in asm: the
-#                 bounded rx and the page-store loop, the two whose remaining
-#                 cost is the C ABI itself. Everything else, bring-up to
-#                 dispatch, is C++ on libavr.
-#   tsb_tricks - no asm at all: the whole-loader register allocation lives in
+# and the size gradient is the cost of that "how" — see dev/lessons.md.
+#   tsb_asm   — the tricks tier's C++ with exactly two routines in asm (the
+#                 bounded rx and the page-store loop — the two whose remaining
+#                 cost is the C ABI itself): 510 B in the 512 B section the
+#                 hand-written 500 B oracle occupies. Everything else, from
+#                 bring-up to dispatch, is C++ on libavr.
+#   tsb_tricks — no asm at all: the whole-loader register allocation lives in
 #                 global register variables (Y walks the page pointer), every
 #                 helper is a tiny noinline primitive placed by the
 #                 global-register store rules, pages stream straight to
-#                 SPM/EEPROM.
-#   tsb_pure   - pure idiomatic libavr, one function per command, TU-local
-#                 (internal linkage), streaming (no SRAM page buffer).
-#   tsb_policy - the policy floor: no inline assembly and no global register
-#                 variables, which is philosophy #5's own bound, and the
-#                 measured evidence that the 512 B fit is a property of the
-#                 mechanisms it bans.
-#
-# What each measures is oracle/README.md's table, which is the one place the
-# four numbers and the hand-written loader's own are compared.
+#                 SPM/EEPROM, and the bring-up is the two reset-non-default
+#                 registers only. 526 B in the 1 KB section (BOOTSZ=10) — 14
+#                 over the oracle's section, from 168 over at this tier's first
+#                 floor.
+#   tsb_pure  — pure idiomatic libavr, one function per command, TU-local
+#                 (internal linkage), streaming (no SRAM page buffer): 836 B in
+#                 the 1 KB section.
 #
 # add_tsb_variant(<name> <boot-section-bytes>)
 function(add_tsb_variant name bytes)
@@ -125,33 +90,88 @@ function(add_tsb_variant name bytes)
 	target_link_options(${name} PRIVATE -nostartfiles -Wl,--section-start=.text=${base_hex}
 	                    -Wl,--defsym=tsb_app=0 -Wl,--pmem-wrap-around=32k)
 	add_custom_command(TARGET ${name} POST_BUILD COMMAND ${CMAKE_SIZE} $<TARGET_FILE:${name}>)
-	add_image_outputs(${name})
 	if(PROJECT_IS_TOP_LEVEL)
 		add_test(NAME ${name}.size
 		         COMMAND ${CMAKE_COMMAND} -DSIZE_TOOL=${CMAKE_SIZE} -DELF=$<TARGET_FILE:${name}>
 		                 -DLIMIT=${bytes} -P ${CMAKE_CURRENT_SOURCE_DIR}/test/check_size.cmake)
-		add_test(NAME ${name}.protocol
-		         COMMAND ${_tsb_python} ${CMAKE_CURRENT_SOURCE_DIR}/test/tsbtest.py
-		                 ${TSB_DEVICE} $<TARGET_FILE:${name}> ${base_hex})
+		if(DEFINED TSB_DEVICE)
+			add_test(NAME ${name}.protocol
+			         COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/test/tsbtest.py
+			                 ${TSB_DEVICE} $<TARGET_FILE:${name}> ${base_hex})
+		endif()
 	endif()
 endfunction()
 
-# The tiers reimplement the ATmega328P-only reference protocol, so the guard is
-# the whole of what this repo builds.
+# The tsb tiers reimplement the ATmega328P-only reference protocol; the other
+# chips build pureboot alone.
 if(LIBAVR_MCU STREQUAL "atmega328p")
 	add_tsb_variant(tsb_asm 512)
-	add_tsb_variant(tsb_policy 1024)
 	add_tsb_variant(tsb_pure 1024)
 	add_tsb_variant(tsb_tricks 1024)
-	# The policy tier's floor needs these two: a loader's loop bodies all
-	# contain calls, which is what makes hoisting an invariant out of one cost
-	# more than it saves. The other tiers keep the flag set their recorded
-	# floors were measured with - none.
-	target_compile_options(tsb_policy PRIVATE -fno-move-loop-invariants -fno-tree-ter)
 endif()
 
-# Every test registered above carries the marker a stubbed launcher prints, so
-# a check this host cannot run reads as Skipped rather than Failed.
-if(PROJECT_IS_TOP_LEVEL)
-	libavr_skip_unverified()
+# pureboot — the pure-constraint port (see pureboot/README.md): one source,
+# no inline assembly, no global register variables, every libavr chip, 512
+# bytes each. The loader owns the top 512 bytes of flash on every chip; the
+# application entry symbol is address 0 on the mega (reset re-vectors to the
+# loader through BOOTRST, so word 0 stays the application's own vector) and
+# the trampoline word just below the loader on the tinies (host-side vector
+# surgery points it at the application). --pmem-wrap-around models AVR's
+# modulo-flash PC where the flash is big enough to need it.
+if(LIBAVR_MCU STREQUAL "attiny13a")
+	set(_pb_flash 1024)
+	set(_pb_wrap "")
+	set(_pb_page 32)
+	set(_pb_hz 9600000)
+	set(_pb_baud 57600)
+	set(_pb_eeprom 64)
+elseif(LIBAVR_MCU STREQUAL "attiny85")
+	set(_pb_flash 8192)
+	set(_pb_wrap -Wl,--pmem-wrap-around=8k)
+	set(_pb_page 64)
+	set(_pb_hz 8000000)
+	set(_pb_baud 57600)
+	set(_pb_eeprom 512)
+else()
+	set(_pb_flash 32768)
+	set(_pb_wrap -Wl,--pmem-wrap-around=32k)
+	set(_pb_page 128)
+	set(_pb_hz 16000000)
+	set(_pb_baud 115200)
+	set(_pb_eeprom 1024)
+endif()
+math(EXPR _pb_base "${_pb_flash} - 512")
+math(EXPR _pb_base_hex "${_pb_base}" OUTPUT_FORMAT HEXADECIMAL)
+if(LIBAVR_MCU STREQUAL "atmega328p")
+	set(_pb_app 0)
+else()
+	math(EXPR _pb_app "${_pb_base} - 2")
+endif()
+
+add_executable(pureboot pureboot/pureboot.cpp)
+target_link_libraries(pureboot PRIVATE libavr)
+target_link_options(pureboot PRIVATE -nostartfiles -Wl,--section-start=.text=${_pb_base_hex}
+                    -Wl,--defsym=pureboot_app=${_pb_app} ${_pb_wrap})
+add_custom_command(TARGET pureboot POST_BUILD COMMAND ${CMAKE_SIZE} $<TARGET_FILE:pureboot>)
+if(PROJECT_IS_TOP_LEVEL)
+	add_test(NAME pureboot.size
+	         COMMAND ${CMAKE_COMMAND} -DSIZE_TOOL=${CMAKE_SIZE} -DELF=$<TARGET_FILE:pureboot>
+	                 -DLIMIT=512 -P ${CMAKE_CURRENT_SOURCE_DIR}/test/check_size.cmake)
+
+	# The protocol test flashes this fixture through the loader with the real
+	# host tool and expects its banner after the hand-over; a normally linked
+	# application whose reset vector is what the tinies' surgery re-homes.
+	if(DEFINED PB_DEVICE)
+		add_executable(pbapp test/pbapp.cpp)
+		target_link_libraries(pbapp PRIVATE libavr)
+		add_custom_command(TARGET pbapp POST_BUILD
+		                   COMMAND ${CMAKE_OBJCOPY} -O binary $<TARGET_FILE:pbapp> $<TARGET_FILE:pbapp>.bin)
+		add_test(NAME pureboot.protocol
+		         COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/test/pbtest.py
+		                 ${PB_DEVICE} $<TARGET_FILE:pureboot> ${LIBAVR_MCU} ${_pb_hz} ${_pb_base_hex}
+		                 ${_pb_page} ${_pb_baud} ${_pb_eeprom} $<TARGET_FILE:pbapp>.bin
+		                 ${CMAKE_CURRENT_SOURCE_DIR}/pureboot/pureboot.py
+		                 ${CMAKE_BINARY_DIR}/pbtest-work)
+		set_tests_properties(pureboot.protocol PROPERTIES TIMEOUT 180)
+	endif()
 endif()

CMakePresets.json

@@ -1,84 +1,86 @@
 {
- "version": 8,
- "configurePresets": [
-  {
-   "name": "base",
-   "hidden": true,
-   "generator": "Ninja",
-   "binaryDir": "${sourceDir}/build/${presetName}",
-   "toolchainFile": "${sourceDir}/libavr/cmake/avr-toolchain.cmake",
-   "cacheVariables": {
-    "CMAKE_BUILD_TYPE": "Release",
-    "CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
-    "CMAKE_COLOR_DIAGNOSTICS": "ON"
-   }
-  },
-  {
-   "name": "atmega328p-generated",
-   "inherits": "base",
-   "cacheVariables": {
-    "LIBAVR_MCU": "atmega328p",
-    "LIBAVR_REFLECT": "OFF"
-   }
-  },
-  {
-   "name": "atmega328p-reflect",
-   "inherits": "base",
-   "cacheVariables": {
-    "LIBAVR_MCU": "atmega328p",
-    "LIBAVR_REFLECT": "ON"
-   }
-  }
- ],
- "buildPresets": [
-  {
-   "name": "atmega328p-generated",
-   "configurePreset": "atmega328p-generated"
-  },
-  {
-   "name": "atmega328p-reflect",
-   "configurePreset": "atmega328p-reflect"
-  }
- ],
- "testPresets": [
-  {
-   "name": "atmega328p-generated",
-   "configurePreset": "atmega328p-generated",
-   "output": {
-    "outputOnFailure": true
-   }
-  }
- ],
- "workflowPresets": [
-  {
-   "name": "atmega328p-generated",
-   "steps": [
-    {
-     "type": "configure",
-     "name": "atmega328p-generated"
-    },
-    {
-     "type": "build",
-     "name": "atmega328p-generated"
-    },
-    {
-     "type": "test",
-     "name": "atmega328p-generated"
-    }
-   ]
-  },
-  {
-   "name": "atmega328p-reflect",
-   "steps": [
-    {
-     "type": "configure",
-     "name": "atmega328p-reflect"
-    },
-    {
-     "type": "build",
-     "name": "atmega328p-reflect"
-    }
-   ]
-  }
- ]
+	"version": 8,
+	"configurePresets": [
+		{
+			"name": "base",
+			"hidden": true,
+			"generator": "Ninja",
+			"binaryDir": "${sourceDir}/build/${presetName}",
+			"toolchainFile": "$env{LIBAVR_ROOT}/cmake/avr-toolchain.cmake",
+			"cacheVariables": {
+				"CMAKE_BUILD_TYPE": "Release",
+				"CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
+				"CMAKE_COLOR_DIAGNOSTICS": "ON"
+			}
+		},
+		{
+			"name": "atmega328p-generated",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "atmega328p", "LIBAVR_REFLECT": "OFF" }
+		},
+		{
+			"name": "atmega328p-reflect",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "atmega328p", "LIBAVR_REFLECT": "ON" }
+		},
+		{
+			"name": "attiny85-generated",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "attiny85", "LIBAVR_REFLECT": "OFF" }
+		},
+		{
+			"name": "attiny85-reflect",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "attiny85", "LIBAVR_REFLECT": "ON" }
+		},
+		{
+			"name": "attiny13a-generated",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "attiny13a", "LIBAVR_REFLECT": "OFF" }
+		},
+		{
+			"name": "attiny13a-reflect",
+			"inherits": "base",
+			"cacheVariables": { "LIBAVR_MCU": "attiny13a", "LIBAVR_REFLECT": "ON" }
+		}
+	],
+	"buildPresets": [
+		{ "name": "atmega328p-generated", "configurePreset": "atmega328p-generated" },
+		{ "name": "atmega328p-reflect", "configurePreset": "atmega328p-reflect" },
+		{ "name": "attiny85-generated", "configurePreset": "attiny85-generated" },
+		{ "name": "attiny85-reflect", "configurePreset": "attiny85-reflect" },
+		{ "name": "attiny13a-generated", "configurePreset": "attiny13a-generated" },
+		{ "name": "attiny13a-reflect", "configurePreset": "attiny13a-reflect" }
+	],
+	"workflowPresets": [
+		{
+			"name": "atmega328p-generated",
+			"steps": [
+				{ "type": "configure", "name": "atmega328p-generated" },
+				{ "type": "build", "name": "atmega328p-generated" },
+				{ "type": "test", "name": "atmega328p-generated" }
+			]
+		},
+		{
+			"name": "attiny85-generated",
+			"steps": [
+				{ "type": "configure", "name": "attiny85-generated" },
+				{ "type": "build", "name": "attiny85-generated" },
+				{ "type": "test", "name": "attiny85-generated" }
+			]
+		},
+		{
+			"name": "attiny13a-generated",
+			"steps": [
+				{ "type": "configure", "name": "attiny13a-generated" },
+				{ "type": "build", "name": "attiny13a-generated" },
+				{ "type": "test", "name": "attiny13a-generated" }
+			]
+		}
+	],
+	"testPresets": [
+		{ "name": "atmega328p-generated", "configurePreset": "atmega328p-generated", "output": { "outputOnFailure": true } },
+		{ "name": "attiny85-generated", "configurePreset": "attiny85-generated", "output": { "outputOnFailure": true } },
+		{ "name": "attiny13a-generated", "configurePreset": "attiny13a-generated", "output": { "outputOnFailure": true } }
+	]
 }

ide/README.md

@@ -1,67 +0,0 @@
-# Atmel Studio
-
-`master` carries `bootloader.atsln`, so this branch does too: `ide/bootloader.atsln`
-builds the loader from the same source Ninja does, to a **byte-identical
-`.text`** — the `tsb_asm` tier in its 512-byte section (`check-flags.py` below
-is what holds the flag sets equal, so the size is Ninja's own). CMake remains
-the build system; the solution is here so the port opens in Studio as its
-predecessor did.
-
-## One project, of four tiers
-
-A `.cppproj` is one binary at one set of flags. `tsb_asm` is the tier that
-occupies the same 512-byte section `master`'s `tsb` project targeted, which is
-the one worth opening in Studio.
-
-The other three tiers (`tsb_pure`, `tsb_tricks`, `tsb_policy`) are not here.
-They differ from `tsb_asm` in their source file, their section size, and — for
-`tsb_policy` — two loop flags; nothing about that is a Studio concern, and what
-they exist to demonstrate is a size gradient only the CMake size tests measure.
-Adding one is a copy of `tsb_asm/tsb_asm.cppproj` in its own directory, with its
-name, its GUID, its source path and its `--section-start` changed (`0x7c00` for
-the 1 KiB tiers), plus four lines in the solution.
-
-`avrdevice` is a project property, so each project gets its own directory:
-Studio builds into `<project dir>/<Configuration>` whatever `OutputDirectory`
-says, and two projects sharing a directory would share one object file.
-
-## Debug keeps `-Os`
-
-Both configurations compile at `-Os`; Debug adds only `-gdwarf-4`. The `.text`
-is therefore identical in both, which is the point — a loader's section is a
-**correctness** bound and not a budget. A debug configuration that silently
-overruns the section is worse than none, and DWARF costs no flash, so the
-optimisation level stays where correctness needs it.
-
-## What Studio needs from the machine
-
-libavr from the **submodule**, found at
-`$(MSBuildProjectDirectory)\..\..\libavr\include` — correct by construction, and
-anchored to the project because a plain relative path resolves against the
-generated makefile's directory (the configuration's output directory), not the
-project's. There is no `LIBAVR_ROOT` escape hatch: a variable exported in a
-shell is invisible to Studio launched from the Start menu, and the failure reads
-as a missing `libavr/libavr.hpp` — which is what the submodule answers.
-
-A GCC 16.1 toolchain registered as flavour `avr-g++-16.1.0`, nothing older
-reaching `-std=c++26`.
-
-## Generating and gating
-
-One generated file is required before the project will load at all, and one
-command checks the flags have not drifted (both from libavr's
-`tools/atmelstudio/`):
-
-```sh
-python libavr/tools/atmelstudio/componentinfo.py \
-    ide/tsb_asm/tsb_asm.componentinfo.xml --device ATmega328P
-python libavr/tools/atmelstudio/check-flags.py \
-    --solution ide/bootloader.atsln --project tsb_asm --target tsb_asm \
-    --compile-commands build/atmega328p-generated/compile_commands.json \
-    --log build/as-tsb_asm.log
-```
-
-Release is what the gate compares — the presets define no debug build, and
-Debug differs from Release only in `-gdwarf-4`.
-
-Legacy (the yazoalfa-era submodules) stays on `master`.

ide/bootloader.atsln

@@ -1,22 +0,0 @@
-
-Microsoft Visual Studio Solution File, Format Version 12.00
-# Atmel Studio Solution File, Format Version 11.00
-VisualStudioVersion = 14.0.23107.0
-MinimumVisualStudioVersion = 10.0.40219.1
-Project("{E66E83B9-2572-4076-B26E-6BE79FF3018A}") = "tsb_asm", "tsb_asm\tsb_asm.cppproj", "{6618D3BE-7EB3-49A2-9113-F128E396FF06}"
-EndProject
-Global
-	GlobalSection(SolutionConfigurationPlatforms) = preSolution
-		Debug|AVR = Debug|AVR
-		Release|AVR = Release|AVR
-	EndGlobalSection
-	GlobalSection(ProjectConfigurationPlatforms) = postSolution
-		{6618D3BE-7EB3-49A2-9113-F128E396FF06}.Debug|AVR.ActiveCfg = Debug|AVR
-		{6618D3BE-7EB3-49A2-9113-F128E396FF06}.Debug|AVR.Build.0 = Debug|AVR
-		{6618D3BE-7EB3-49A2-9113-F128E396FF06}.Release|AVR.ActiveCfg = Release|AVR
-		{6618D3BE-7EB3-49A2-9113-F128E396FF06}.Release|AVR.Build.0 = Release|AVR
-	EndGlobalSection
-	GlobalSection(SolutionProperties) = preSolution
-		HideSolutionNode = FALSE
-	EndGlobalSection
-EndGlobal

ide/tsb_asm/tsb_asm.cppproj

@@ -1,112 +0,0 @@
-<?xml version="1.0" encoding="utf-8"?>
-<Project DefaultTargets="Build" xmlns="http://schemas.microsoft.com/developer/msbuild/2003" ToolsVersion="14.0">
-  <PropertyGroup>
-    <SchemaVersion>2.0</SchemaVersion>
-    <ProjectVersion>7.0</ProjectVersion>
-    <ToolchainName>com.Atmel.AVRGCC8.CPP</ToolchainName>
-    <ProjectGuid>6618d3be-7eb3-49a2-9113-f128e396ff06</ProjectGuid>
-    <avrdevice>ATmega328P</avrdevice>
-    <avrdeviceseries>none</avrdeviceseries>
-    <OutputType>Executable</OutputType>
-    <Language>CPP</Language>
-    <OutputFileName>$(MSBuildProjectName)</OutputFileName>
-    <OutputFileExtension>.elf</OutputFileExtension>
-    <OutputDirectory>$(MSBuildProjectDirectory)\$(Configuration)</OutputDirectory>
-    <AssemblyName>tsb_asm</AssemblyName>
-    <Name>tsb_asm</Name>
-    <RootNamespace>tsb_asm</RootNamespace>
-    <ToolchainFlavour>avr-g++-16.1.0</ToolchainFlavour>
-    <KeepTimersRunning>true</KeepTimersRunning>
-    <OverrideVtor>false</OverrideVtor>
-    <CacheFlash>true</CacheFlash>
-    <ProgFlashFromRam>true</ProgFlashFromRam>
-    <RamSnippetAddress>0x20000000</RamSnippetAddress>
-    <UncachedRange />
-    <preserveEEPROM>true</preserveEEPROM>
-    <OverrideVtorValue>exception_table</OverrideVtorValue>
-    <BootSegment>2</BootSegment>
-    <ResetRule>0</ResetRule>
-    <eraseonlaunchrule>0</eraseonlaunchrule>
-    <EraseKey />
-    <AsfFrameworkConfig>
-      <framework-data xmlns="">
-  <options />
-  <configurations />
-  <files />
-  <documentation help="" />
-  <offline-documentation help="" />
-  <dependencies>
-    <content-extension eid="atmel.asf" uuidref="Atmel.ASF" version="3.52.0" />
-  </dependencies>
-</framework-data>
-    </AsfFrameworkConfig>
-  </PropertyGroup>
-  <PropertyGroup Condition=" '$(Configuration)' == 'Release' ">
-    <ToolchainSettings>
-      <AvrGccCpp>
-        <avrgcc.common.Device>-mmcu=atmega328p</avrgcc.common.Device>
-        <avrgcc.common.outputfiles.hex>True</avrgcc.common.outputfiles.hex>
-        <avrgcc.common.outputfiles.lss>True</avrgcc.common.outputfiles.lss>
-        <avrgcc.common.outputfiles.eep>True</avrgcc.common.outputfiles.eep>
-        <avrgcc.common.outputfiles.srec>True</avrgcc.common.outputfiles.srec>
-        <avrgcc.common.outputfiles.usersignatures>False</avrgcc.common.outputfiles.usersignatures>
-        <avrgcccpp.compiler.symbols.DefSymbols>
-          <ListValues>
-            <Value>NDEBUG</Value>
-          </ListValues>
-        </avrgcccpp.compiler.symbols.DefSymbols>
-        <avrgcccpp.compiler.directories.IncludePaths>
-          <ListValues>
-            <Value>$(MSBuildProjectDirectory)\..\..\libavr\include</Value>
-          </ListValues>
-        </avrgcccpp.compiler.directories.IncludePaths>
-        <avrgcccpp.compiler.optimization.level>Optimize for size (-Os)</avrgcccpp.compiler.optimization.level>
-        <avrgcccpp.compiler.optimization.PrepareFunctionsForGarbageCollection>True</avrgcccpp.compiler.optimization.PrepareFunctionsForGarbageCollection>
-        <avrgcccpp.compiler.optimization.PrepareDataForGarbageCollection>True</avrgcccpp.compiler.optimization.PrepareDataForGarbageCollection>
-        <avrgcccpp.compiler.warnings.AllWarnings>True</avrgcccpp.compiler.warnings.AllWarnings>
-        <avrgcccpp.compiler.miscellaneous.OtherFlags>-std=c++26 -Wextra -Werror -mrelax -fno-exceptions -fno-rtti -fno-threadsafe-statics</avrgcccpp.compiler.miscellaneous.OtherFlags>
-        <avrgcccpp.linker.optimization.GarbageCollectUnusedSections>True</avrgcccpp.linker.optimization.GarbageCollectUnusedSections>
-        <avrgcccpp.linker.miscellaneous.LinkerFlags>-mrelax -nostartfiles -Wl,--section-start=.text=0x7e00 -Wl,--defsym=tsb_app=0 -Wl,--pmem-wrap-around=32k</avrgcccpp.linker.miscellaneous.LinkerFlags>
-      </AvrGccCpp>
-    </ToolchainSettings>
-  </PropertyGroup>
-  <PropertyGroup Condition=" '$(Configuration)' == 'Debug' ">
-    <ToolchainSettings>
-      <AvrGccCpp>
-        <avrgcc.common.Device>-mmcu=atmega328p</avrgcc.common.Device>
-        <avrgcc.common.outputfiles.hex>True</avrgcc.common.outputfiles.hex>
-        <avrgcc.common.outputfiles.lss>True</avrgcc.common.outputfiles.lss>
-        <avrgcc.common.outputfiles.eep>True</avrgcc.common.outputfiles.eep>
-        <avrgcc.common.outputfiles.srec>True</avrgcc.common.outputfiles.srec>
-        <avrgcc.common.outputfiles.usersignatures>False</avrgcc.common.outputfiles.usersignatures>
-        <avrgcccpp.compiler.symbols.DefSymbols>
-          <ListValues>
-            <Value>DEBUG</Value>
-          </ListValues>
-        </avrgcccpp.compiler.symbols.DefSymbols>
-        <avrgcccpp.compiler.directories.IncludePaths>
-          <ListValues>
-            <Value>$(MSBuildProjectDirectory)\..\..\libavr\include</Value>
-          </ListValues>
-        </avrgcccpp.compiler.directories.IncludePaths>
-        <avrgcccpp.compiler.optimization.level>Optimize for size (-Os)</avrgcccpp.compiler.optimization.level>
-        <avrgcccpp.compiler.optimization.PrepareFunctionsForGarbageCollection>True</avrgcccpp.compiler.optimization.PrepareFunctionsForGarbageCollection>
-        <avrgcccpp.compiler.optimization.PrepareDataForGarbageCollection>True</avrgcccpp.compiler.optimization.PrepareDataForGarbageCollection>
-        <avrgcccpp.compiler.warnings.AllWarnings>True</avrgcccpp.compiler.warnings.AllWarnings>
-        <avrgcccpp.compiler.miscellaneous.OtherFlags>-std=c++26 -Wextra -Werror -mrelax -fno-exceptions -fno-rtti -fno-threadsafe-statics -gdwarf-4</avrgcccpp.compiler.miscellaneous.OtherFlags>
-        <avrgcccpp.linker.optimization.GarbageCollectUnusedSections>True</avrgcccpp.linker.optimization.GarbageCollectUnusedSections>
-        <avrgcccpp.linker.miscellaneous.LinkerFlags>-mrelax -nostartfiles -Wl,--section-start=.text=0x7e00 -Wl,--defsym=tsb_app=0 -Wl,--pmem-wrap-around=32k</avrgcccpp.linker.miscellaneous.LinkerFlags>
-      </AvrGccCpp>
-    </ToolchainSettings>
-  </PropertyGroup>
-  <ItemGroup>
-    <Compile Include="..\..\tsb\tsb_asm.cpp">
-      <SubType>compile</SubType>
-      <Link>tsb\tsb_asm.cpp</Link>
-    </Compile>
-  </ItemGroup>
-  <ItemGroup>
-    <Folder Include="tsb" />
-  </ItemGroup>
-  <Import Project="$(AVRSTUDIO_EXE_PATH)\Vs\Compiler.targets" />
-</Project>

oracle/README.md

@@ -38,23 +38,11 @@ avra -I /usr/share/avra tsb-fixedbaud.asm   # after uncommenting .include "m328P
 ```
 
 **500 bytes with every feature** — the proof that ≤512 B and full feature parity
-are simultaneously reachable. The port's four tiers reach it from the other
-side, and the gradient between them is the cost of the mechanisms each is
-allowed:
-
-| tier | bytes | section | what it is allowed |
-|---|---|---|---|
-| oracle | 500 | 512 B | hand-written assembly, the reference |
-| `tsb_asm` | 512 | 512 B | C++ on libavr, two routines in asm |
-| `tsb_tricks` | 528 | 1 KB | no asm; global register variables |
-| `tsb_policy` | 630 | 1 KB | pureboot's rules: no asm, no register variables |
-| `tsb_pure` | 776 | 1 KB | idiomatic libavr throughout |
-
-The two routines `tsb_asm` keeps are the ones whose remaining cost is the
-calling convention itself: the bounded rx and the page-store loop. It fills
-its section exactly, with the same one-bit-time turn-around guard the oracle
-spends six bytes on - every tier implements the whole feature set, which is
-what makes the column a gradient rather than four different loaders.
+are simultaneously reachable. The port's `tsb_asm` tier meets the same bar at
+510 B in the same 512 B section, written in C++ on libavr except the two
+routines whose remaining cost is the calling convention itself (the bounded rx
+and the page-store loop); `tsb_tricks` needs no assembly at all at 526 B, and
+`tsb_pure` stays fully idiomatic at 836 B, both in the 1 KB section.
 
 The oracle targets 20 MHz / 33333 baud; the port targets 16 MHz / 115200 baud
 (what the simavr protocol test drives). Baud and geometry differ, code size and

pureboot/README.md

@@ -0,0 +1,123 @@
+# pureboot
+
+A serial bootloader on [libavr](https://git.blackmark.me/avr/libavr), pure by
+constraint: one C++ source, no inline assembly, no global register variables
+(attributes allowed), built for every chip libavr targets, **512 bytes on
+each** — 490 B on the ATtiny13A, 510 B on the ATtiny85, 484 B on the
+ATmega328P. The device speaks primitives; every composite — verify, erase,
+reset-vector surgery, timeout configuration — lives in the host tool
+(`pureboot.py`).
+
+## Link
+
+| Chip | Serial | Baud | Clock assumed |
+|---|---|---|---|
+| ATmega328P | USART0, RXD/TXD = PD0/PD1 | 115200 8N1 | 16 MHz crystal |
+| ATtiny85 | software UART, RX = PB0, TX = PB1 | 57600 8N1 | 8 MHz internal RC |
+| ATtiny13A | software UART, RX = PB0, TX = PB1 | 57600 8N1 | 9.6 MHz internal RC |
+
+The tiny RX pin has its pull-up enabled; TX idles high. All multi-byte
+quantities on the wire are little-endian.
+
+## Activation
+
+Reset enters the loader (BOOTRST on the mega, the patched reset vector on the
+tinies) — except a watchdog reset, which hands straight to the application
+(the application owns its watchdog; it must clear WDRF itself, which also
+releases the WDRF-forced WDE).
+
+The host then has one activation window per awaited byte to knock: `p` then
+`b`. Each awaited byte gets a fresh window; any other byte is discarded and
+awaited again (line noise cannot lock the loader, only delay it). A window
+expiring with an idle line boots the application.
+
+The window length in seconds is the **last EEPROM cell** (address
+`eeprom_size - 1`); `0x00` and the erased `0xff` both mean the 4 s default,
+so a full EEPROM erase resets the timeout rather than maxing it. The host
+changes it with the ordinary EEPROM-write command.
+
+## Session
+
+After the knock the loader stays in its command loop until `G` or a reset.
+Before reading each command it waits for any pending EEPROM write to finish
+and sends the prompt `+` (0x2b) — the prompt is therefore also the completion
+ack of the previous command. A session is: await `+`, send a command, read
+its reply, repeat.
+
+| Cmd | Arguments | Reply |
+|---|---|---|
+| `b` | — | the 12-byte info block |
+| `R` | addr16, n8 | n flash bytes (n = 0 means 256) |
+| `W` | addr16, then one page of data | — (completion = next prompt) |
+| `r` | addr16, n8 | n EEPROM bytes (n = 0 means 256) |
+| `w` | addr16, n8, then n data bytes | `+` per byte, sent once its write has begun |
+| `F` | — | 4 bytes: low fuse, lock, extended fuse, high fuse |
+| `G` | — | `+`, then the application runs |
+| other | — | ignored; the loop re-prompts (send a junk byte, await `+`, to resync) |
+
+`W` streams exactly one SPM page (size from the info block) into the buffer,
+then erases and programs; the address must be page-aligned. Pages inside the
+loader's own 512 bytes are drained but never programmed — a broken host
+cannot brick the chip. `w` is host-paced: send the next byte only after the
+previous byte's `+`. `F` returns the bytes in the hardware's Z order; on a chip without an
+extended fuse byte (the ATtiny13A) that slot carries no meaning. Fuse *writing* does not
+exist: SPM reaches flash (and, on the mega, lock bits) only — fuse bytes are
+external-programming territory by hardware.
+
+The info block (`b`):
+
+| Offset | Content |
+|---|---|
+| 0–2 | `'P'`, `'B'`, protocol version (1) |
+| 3–5 | device signature |
+| 6 | SPM page size in bytes |
+| 7–8 | loader base — application flash ends here |
+| 9–10 | EEPROM size |
+| 11 | bit 0 set: host must patch the reset vector (no hardware boot section) |
+
+Composites are the host's job: verify = read back and compare, erase =
+write `0xff` (per page for flash, per byte for EEPROM), timeout = EEPROM
+write to the last cell.
+
+## Deployment
+
+**ATmega328P**: program the loader at 0x7e00 with an external programmer;
+fuses BOOTSZ = 11 (256 words) and BOOTRST programmed. Applications are
+flashed unmodified — reset re-vectors to the loader in hardware, word 0
+stays the application's own reset vector, and `G` jumps to 0.
+
+**Tinies** (no boot section): program the loader at `flash - 512`; erased
+flash below it walks up into the loader, so a virgin chip activates. When
+flashing an application the host performs reset-vector surgery: the
+application's own `rjmp` target is re-encoded as a trampoline `rjmp` in the
+word just below the loader (`base - 2`, where `G` jumps), and word 0 is
+rewritten to `rjmp` to the loader base. Every other vector stays the
+application's. Page 0 is written last, so an interrupted flash leaves word 0
+erased and the chip still falls through to the loader on the next reset.
+
+## Host tool
+
+`pureboot.py` — Python 3, standard library only (termios drives any tty,
+a USB adapter as well as a simavr pty):
+
+    pureboot.py --port /dev/ttyUSB0 --baud 57600 \
+        --info --fuses --flash app.hex --timeout 10
+
+Operations run in a fixed order within one session: info, fuses, flash
+(erase / program / read / verify), EEPROM (erase / program / read / verify),
+timeout — then the loader hands over to the application; `--stay` keeps the
+session alive instead, and a later invocation reconnects into it (the knock
+converges there too). `--flash` and `--eeprom` verify by read-back unless
+`--no-verify`; images are raw binary, or Intel HEX by extension.
+
+## Tests
+
+Per chip preset, `ctest` runs the 512-byte size gate and the end-to-end
+protocol test: a simavr device (`test/pureboot_device.c` — the mega's USART
+as a pty; on the tinies a cycle-timed GPIO⇄pty bridge for the software UART,
+plus the SPM/NVM module simavr's tiny cores lack) driven by the real host
+tool through knock-from-reset, program + verify of both memories, timeout
+configuration, session reconnect, an external reset through the patched
+vector, and the hand-over to a fixture application whose banner proves the
+launch — cross-checked against the simulator's ground-truth memory dumps and
+an independent decode of the surgery's rjmp words.

pureboot/pureboot.cpp

@@ -0,0 +1,326 @@
+// pureboot — a serial bootloader on libavr, pure by constraint: one C++
+// source with no inline assembly and no global register variables, built for
+// every chip libavr targets, 512 bytes on each. The device speaks primitives
+// — read/program flash, read/write EEPROM, fuse bytes, an info block, run —
+// and everything composite (verify, erase, reset-vector surgery on the
+// tinies, timeout configuration) lives in the host tool. Protocol reference:
+// README.md next to this file.
+//
+// Entry: reset lands in avr::startup::entry below (BOOTRST on the mega; the
+// patched reset vector — or erased flash walking up into the loader — on the
+// tinies). A watchdog reset hands straight to the application. Otherwise the
+// host has one activation window — EEPROM's last cell, in seconds — to knock
+// ("pb"); an idle line boots the application. A session then stays in the
+// command loop until 'G' hands over or the chip resets.
+
+#include <libavr/libavr.hpp>
+
+using namespace avr::literals;
+namespace spm = avr::spm;
+namespace ee = avr::eeprom;
+
+namespace pureboot {
+namespace {
+
+// Purely polled — interrupts stay off, every guard folds to nothing.
+constexpr auto off = avr::irq::guard_policy::unused;
+
+constexpr std::uint8_t ack = '+';
+
+// Per-chip personality, from the chip database: the clocks the dogfood
+// boards run (16 MHz crystal on the mega, calibrated RC on the tinies) and
+// the device signature (compile-time data — the tiny13A cannot even read its
+// signature row from code).
+consteval avr::hertz_t clock()
+{
+	if (avr::hw::db.name == "ATtiny13A")
+		return 9.6_MHz;
+	if (avr::hw::db.name == "ATtiny85")
+		return 8_MHz;
+	return 16_MHz;
+}
+
+consteval std::array<std::uint8_t, 3> signature()
+{
+	if (avr::hw::db.name == "ATtiny13A")
+		return {0x1e, 0x90, 0x07};
+	if (avr::hw::db.name == "ATtiny85")
+		return {0x1e, 0x93, 0x0b};
+	return {0x1e, 0x95, 0x0f};
+}
+
+using dev = avr::device<{.clock = clock()}>;
+
+// Geometry: the loader owns the top 512 bytes of flash; the byte below it is
+// the trampoline word (the application's relocated reset vector) on chips
+// without a hardware boot section. The RWWSRE bit marks a separate boot
+// section — on classic AVR the two capabilities coincide.
+constexpr std::uint16_t boot_bytes = 512;
+constexpr std::uint16_t base = static_cast<std::uint16_t>(spm::flash_bytes - boot_bytes);
+constexpr std::uint16_t page = spm::page_bytes;
+constexpr bool boot_section = avr::hw::db.field_index("SPMCSR", "RWWSRE") >= 0;
+
+// The activation timeout lives in EEPROM's last cell, in seconds; the host
+// rewrites it with the ordinary EEPROM-write command. An unprogrammed cell —
+// 0x00 or the erased 0xff — means the 4 s default: a stray value can never
+// floor the window to nothing and lock the loader out, and erasing the whole
+// EEPROM resets the timeout instead of maxing it to 255 s.
+constexpr std::uint16_t timeout_cell = avr::hw::db.mem.eeprom_size - 1;
+constexpr std::uint8_t default_seconds = 4;
+
+// The 12-byte info block the host reads with the 'b' command; flash-resident
+// (there is no crt to copy a .data image).
+inline constexpr std::array<std::uint8_t, 12> info_data = {
+    'P',
+    'B',
+    1, // magic, protocol version
+    signature()[0],
+    signature()[1],
+    signature()[2],
+    static_cast<std::uint8_t>(page),
+    base & 0xff,
+    base >> 8, // app flash ends here; loader base
+    avr::hw::db.mem.eeprom_size & 0xff,
+    avr::hw::db.mem.eeprom_size >> 8,
+    boot_section ? 0 : 1, // bit 0: host must patch the reset vector (no hardware boot section)
+};
+using info = avr::flash_table<info_data>;
+
+// The serial link: the hardware USART where the chip has one, the polled
+// software UART (no vector — the table belongs to the application) on PB0/PB1
+// elsewhere. Both are class templates on the clock so only the selected
+// backend is ever instantiated. pending() is the cheap line test the
+// activation window polls; rx() then picks the byte up.
+template <avr::hertz_t C>
+consteval std::int16_t rxc_field()
+{
+	return avr::hw::db.field_index("UCSR0A", "RXC0");
+}
+
+template <avr::hertz_t C>
+struct hardware_link {
+	using uart = avr::uart::usart0<C, {.baud = 115200_Bd, .max_baud_error = 2.5_pct}>;
+
+	static void init()
+	{
+		avr::init<uart>();
+	}
+
+	static bool pending()
+	{
+		return avr::hw::field_impl<rxc_field<C>()>::test();
+	}
+
+	static std::uint8_t rx()
+	{
+		return uart::read_blocking();
+	}
+
+	static void tx(std::uint8_t byte)
+	{
+		uart::write(byte);
+	}
+};
+
+template <avr::hertz_t C>
+struct software_link {
+	using rx_t = avr::uart::software_rx_polled<C, avr::pb0, 57600_Bd>;
+	using tx_t = avr::uart::software_tx<C, avr::pb1, 57600_Bd>;
+
+	static void init()
+	{
+		avr::init<rx_t, tx_t>();
+	}
+
+	static bool pending()
+	{
+		return !avr::io::input<avr::pb0>::read(); // a start bit has begun
+	}
+
+	static std::uint8_t rx()
+	{
+		return rx_t::template read_blocking<off>();
+	}
+
+	static void tx(std::uint8_t byte)
+	{
+		tx_t::template write<off>(byte);
+	}
+};
+
+using link = std::conditional_t<avr::hw::db.has_reg("UDR0"), hardware_link<dev::clock>, software_link<dev::clock>>;
+
+// The application's entry: the linker pins pureboot_app to 0x0000 on the
+// mega (reset re-vectors here through BOOTRST, so address 0 stays the
+// application's own vector) and to the trampoline word at base - 2 on the
+// tinies (--defsym in CMakeLists.txt).
+extern "C" [[noreturn]] void pureboot_app();
+
+[[noreturn]] void run_app()
+{
+	pureboot_app();
+}
+
+// One activation tick is 65536 pending() polls — a pin (or flag) test plus a
+// 16-bit countdown, about 8 cycles. Whole-second precision is all the
+// timeout cell promises; the seconds count stays a loop bound (a runtime
+// multiply would drag libgcc's __mulhi3 into the MUL-less tinies).
+consteval std::uint16_t ticks_per_second()
+{
+	return static_cast<std::uint16_t>(dev::clock.hz / (65536ull * 8u));
+}
+static_assert(ticks_per_second() >= 1);
+
+bool pending_before(std::uint8_t seconds)
+{
+	do {
+		std::uint16_t ticks = ticks_per_second();
+		do {
+			std::uint16_t spins = 0; // wraps first, so 65536 polls per tick
+			do {
+				if (link::pending())
+					return true;
+			} while (--spins);
+		} while (--ticks);
+	} while (--seconds);
+	return false;
+}
+
+// A knock byte under the activation deadline: an idle line means no host is
+// there, and the application runs.
+std::uint8_t rx_deadline(std::uint8_t seconds)
+{
+	if (!pending_before(seconds))
+		run_app();
+	return link::rx();
+}
+
+std::uint16_t rx16()
+{
+	std::uint8_t low = link::rx();
+	return static_cast<std::uint16_t>(low | (link::rx() << 8));
+}
+
+const std::uint8_t *flash_ptr(std::uint16_t address)
+{
+	return reinterpret_cast<const std::uint8_t *>(address);
+}
+
+// The streamers take the count in the wire's 8-bit form: 0 means 256.
+void send_flash(std::uint16_t address, std::uint8_t count)
+{
+	do
+		link::tx(avr::flash_load(flash_ptr(address++)));
+	while (--count);
+}
+
+void send_eeprom(std::uint16_t address, std::uint8_t count)
+{
+	do
+		link::tx(ee::read(address++));
+	while (--count);
+}
+
+// EEPROM write, host-paced: each ack goes out once the byte's write has
+// begun, so the next byte arrives while it completes and the following
+// write's own ready-wait sees an idle line. Nothing is ever missed, on
+// either serial backend, without a buffer.
+void store_eeprom(std::uint16_t address, std::uint8_t count)
+{
+	do {
+		ee::write<off>(address++, link::rx());
+		link::tx(ack);
+	} while (--count);
+}
+
+// One flash page: stream the bytes into the SPM buffer as little-endian
+// words, then erase and program. Addresses in the loader's own 512 bytes
+// are drained but never programmed — a broken host cannot brick the chip.
+// On the mega the RWW section is re-enabled so reads work immediately.
+void program_flash(std::uint16_t address)
+{
+	for (std::uint16_t i = 0; i < page; i += 2) {
+		std::uint8_t low = link::rx();
+		std::uint8_t high = link::rx();
+		spm::fill<off>(address + i, static_cast<std::uint16_t>(low | (high << 8)));
+	}
+	if (address < base) {
+		spm::erase_page<off>(address);
+		spm::wait();
+		spm::write_page<off>(address);
+		spm::wait();
+		if constexpr (boot_section)
+			spm::rww_enable<off>();
+	}
+}
+
+// The four fuse/lock bytes in the hardware's own Z order: low, lock,
+// extended, high. Writing fuses is not a thing self-programming can do on
+// AVR — SPM reaches flash (and boot lock bits) only.
+void send_fuses()
+{
+	for (std::uint8_t which = 0; which < 4; ++which)
+		link::tx(spm::read_fuse<off>(static_cast<spm::fuse>(which)));
+}
+
+[[noreturn]] void run()
+{
+	// A watchdog reset belongs to the application (whose watchdog stays
+	// forced on until it clears WDRF) — no activation window in its way.
+	if (avr::hw::mcusr::wdrf.test())
+		run_app();
+
+	link::init();
+
+	std::uint8_t seconds = ee::read(timeout_cell);
+	if (seconds == 0 || seconds == 0xff)
+		seconds = default_seconds;
+
+	// The knock: 'p' then 'b', each under a fresh window; any other byte is
+	// line noise and waits again. Falling out of a window runs the app.
+	while (rx_deadline(seconds) != 'p' || rx_deadline(seconds) != 'b') {
+	}
+
+	for (;;) {
+		// No prompt while an EEPROM write runs: a pending write blocks SPM
+		// and fuse reads (§26.2.1), and the ack tells the host all is done.
+		ee::wait();
+		link::tx(ack);
+		switch (link::rx()) {
+		case 'b': // info block
+			send_flash(reinterpret_cast<std::uint16_t>(info::storage.data()), info::size());
+			break;
+		case 'R': { // read flash: addr16, n8 (0 = 256)
+			std::uint16_t address = rx16();
+			send_flash(address, link::rx());
+			break;
+		}
+		case 'W': // program one flash page: addr16, page bytes
+			program_flash(rx16());
+			break;
+		case 'r': { // read EEPROM: addr16, n8
+			std::uint16_t address = rx16();
+			send_eeprom(address, link::rx());
+			break;
+		}
+		case 'w': { // write EEPROM: addr16, n8, then n bytes each acked
+			std::uint16_t address = rx16();
+			store_eeprom(address, link::rx());
+			break;
+		}
+		case 'F': // fuse and lock bytes
+			send_fuses();
+			break;
+		case 'G': // hand over to the application
+			link::tx(ack);
+			run_app();
+		default: // unknown bytes are ignored; the loop re-acks
+			break;
+		}
+	}
+}
+
+} // namespace
+} // namespace pureboot
+
+template struct avr::startup::entry<pureboot::run>;

pureboot/pureboot.py

@@ -0,0 +1,450 @@
+#!/usr/bin/env python3
+"""pureboot host tool — the smart half of the pureboot protocol (README.md).
+
+The device exposes primitives; this tool composes them: image loading (raw
+binary or Intel HEX), flash programming with read-back verification, erase as
+writing 0xff, EEPROM programming, fuse and info readout, activation-timeout
+configuration, and — on chips without a hardware boot section — the
+reset-vector surgery that re-homes the application's entry through the
+trampoline word below the loader, writing page 0 last so an interrupted
+flash still falls through to the loader.
+
+Python standard library only; the serial port is driven with termios, so any
+tty works — a USB adapter as well as a simavr pty.
+"""
+
+import argparse
+import os
+import select
+import sys
+import termios
+import time
+
+PROMPT = b"+"
+PROTOCOL_VERSION = 1
+
+
+class Error(Exception):
+    pass
+
+
+# ---------------------------------------------------------------- serial ---
+
+
+class Port:
+    """A raw serial port with deadline-based reads."""
+
+    def __init__(self, path, baud):
+        self.fd = os.open(path, os.O_RDWR | os.O_NOCTTY)
+        attrs = termios.tcgetattr(self.fd)
+        attrs[0] = 0  # iflag
+        attrs[1] = 0  # oflag
+        attrs[2] = termios.CREAD | termios.CLOCAL | termios.CS8  # cflag
+        attrs[3] = 0  # lflag
+        try:
+            speed = getattr(termios, f"B{baud}")
+        except AttributeError:
+            raise Error(f"unsupported baud rate {baud}") from None
+        attrs[4] = attrs[5] = speed
+        attrs[6][termios.VMIN] = 0
+        attrs[6][termios.VTIME] = 0
+        termios.tcsetattr(self.fd, termios.TCSANOW, attrs)
+
+    def close(self):
+        os.close(self.fd)
+
+    def write(self, data):
+        os.write(self.fd, data)
+
+    def flush_input(self):
+        termios.tcflush(self.fd, termios.TCIFLUSH)
+
+    def read_available(self, wait):
+        """Everything that arrives within `wait` seconds of quiet start."""
+        ready, _, _ = select.select([self.fd], [], [], wait)
+        return os.read(self.fd, 4096) if ready else b""
+
+    def read_exact(self, count, timeout):
+        data = b""
+        deadline = time.monotonic() + timeout
+        while len(data) < count:
+            remaining = deadline - time.monotonic()
+            if remaining <= 0:
+                raise Error(f"timeout: got {len(data)} of {count} bytes")
+            ready, _, _ = select.select([self.fd], [], [], remaining)
+            if ready:
+                data += os.read(self.fd, count - len(data))
+        return data
+
+
+# -------------------------------------------------------------- protocol ---
+
+
+class Info:
+    """The 12-byte info block."""
+
+    def __init__(self, raw):
+        if len(raw) != 12 or raw[0:2] != b"PB":
+            raise Error(f"bad info block: {raw.hex()}")
+        if raw[2] != PROTOCOL_VERSION:
+            raise Error(f"protocol version {raw[2]}, tool speaks {PROTOCOL_VERSION}")
+        self.signature = raw[3:6]
+        self.page = raw[6]
+        self.base = raw[7] | (raw[8] << 8)
+        self.eeprom_size = raw[9] | (raw[10] << 8)
+        self.patch_vector = bool(raw[11] & 1)
+        self.flash_size = self.base + 512
+
+    def describe(self):
+        sig = " ".join(f"{b:02x}" for b in self.signature)
+        vector = "host-patched reset vector" if self.patch_vector else "hardware boot section"
+        return (
+            f"signature {sig}, page {self.page} B, "
+            f"app flash {self.base} B (loader at {self.base:#06x}), "
+            f"EEPROM {self.eeprom_size} B, {vector}"
+        )
+
+
+class Loader:
+    """A pureboot session. Between commands the loader has prompted `+` and
+    awaits a command byte; every method restores that invariant."""
+
+    def __init__(self, port):
+        self.port = port
+        self.info = None
+
+    def connect(self, wait):
+        """Knock until the activation window answers, then read the info
+        block. Also converges when the loader already sits in its command
+        loop: the knock bytes are ignored-or-executed there, and the drain
+        absorbs whatever they produced."""
+        self.port.flush_input()
+        deadline = time.monotonic() + wait
+        while True:
+            self.port.write(b"pb")
+            if PROMPT in self.port.read_available(0.4):
+                break
+            if time.monotonic() > deadline:
+                raise Error("no answer — reset the device within its activation window")
+        while self.port.read_available(0.3):
+            pass
+        self.port.write(b"b")
+        self.info = Info(self.port.read_exact(12, 2.0))
+        self._expect_prompt()
+        return self.info
+
+    def _expect_prompt(self, timeout=2.0):
+        byte = self.port.read_exact(1, timeout)
+        if byte != PROMPT:
+            raise Error(f"expected prompt, got {byte.hex()}")
+
+    def _command(self, tx, reply_len=0, timeout=2.0):
+        self.port.write(tx)
+        reply = self.port.read_exact(reply_len, timeout) if reply_len else b""
+        self._expect_prompt(timeout)
+        return reply
+
+    def _stream_read(self, command, address, count):
+        data = b""
+        while count:
+            chunk = min(count, 256)
+            head = bytes((ord(command), address & 0xFF, address >> 8, chunk & 0xFF))
+            data += self._command(head, chunk, 5.0)
+            address += chunk
+            count -= chunk
+        return data
+
+    def read_flash(self, address, count):
+        return self._stream_read("R", address, count)
+
+    def read_eeprom(self, address, count):
+        return self._stream_read("r", address, count)
+
+    def write_page(self, address, data):
+        assert len(data) == self.info.page and address % self.info.page == 0
+        head = bytes((ord("W"), address & 0xFF, address >> 8))
+        self._command(head + data, 0, 2.0)
+
+    def write_eeprom(self, address, data):
+        offset = 0
+        while offset < len(data):
+            chunk = data[offset : offset + 256]
+            head = bytes((ord("w"), address & 0xFF, address >> 8, len(chunk) & 0xFF))
+            self.port.write(head)
+            for byte in chunk:
+                self.port.write(bytes((byte,)))
+                self._expect_prompt()  # per-byte ack: the write has begun
+            self._expect_prompt()  # the next command prompt
+            address += len(chunk)
+            offset += len(chunk)
+
+    def read_fuses(self):
+        return self._command(b"F", 4, 2.0)
+
+    def run_application(self):
+        self.port.write(b"G")
+        self._expect_prompt()
+
+
+# ---------------------------------------------------------------- images ---
+
+
+def load_image(path):
+    """Raw binary, or Intel HEX by extension (.hex/.ihx/.ihex)."""
+    data = open(path, "rb").read()
+    if not path.lower().endswith((".hex", ".ihx", ".ihex")):
+        if not data:
+            raise Error(f"{path}: empty image")
+        return data
+    memory = {}
+    for number, line in enumerate(data.decode("ascii", "replace").splitlines(), 1):
+        line = line.strip()
+        if not line:
+            continue
+        if not line.startswith(":"):
+            raise Error(f"{path}:{number}: not an Intel HEX record")
+        record = bytes.fromhex(line[1:])
+        if sum(record) & 0xFF:
+            raise Error(f"{path}:{number}: checksum mismatch")
+        count, address, kind = record[0], (record[1] << 8) | record[2], record[3]
+        payload = record[4 : 4 + count]
+        if kind == 0:
+            for i, byte in enumerate(payload):
+                memory[address + i] = byte
+        elif kind == 1:
+            break
+        elif kind in (2, 4) and not any(payload):
+            continue  # a zero base extends nothing
+        elif kind in (3, 5):
+            continue  # start address: irrelevant, reset is the entry
+        else:
+            raise Error(f"{path}:{number}: record type {kind} reaches beyond the 16-bit space")
+    if not memory:
+        raise Error(f"{path}: empty image")
+    return bytes(memory.get(i, 0xFF) for i in range(max(memory) + 1))
+
+
+# --------------------------------------------------------------- surgery ---
+
+
+def rjmp_target(word_address, opcode, flash_words):
+    return (word_address + 1 + (opcode & 0x0FFF)) % flash_words
+
+
+def rjmp_to(word_address, destination, flash_words):
+    return 0xC000 | ((destination - word_address - 1) % flash_words % 0x1000)
+
+
+def plan_flash(image, info):
+    """The pages to program, as {page_address: bytes}, already carrying the
+    reset-vector surgery where the chip needs it. Page 0 must go last —
+    callers get it separated."""
+    page = info.page
+    limit = info.base - (2 if info.patch_vector else 0)
+    if len(image) > limit:
+        raise Error(f"image is {len(image)} B, application flash ends at {limit}")
+    final = bytearray(image) + bytearray([0xFF] * (-len(image) % page))
+
+    if info.patch_vector:
+        flash_words = info.flash_size // 2
+        word0 = final[0] | (final[1] << 8)
+        if word0 & 0xF000 != 0xC000:
+            raise Error(
+                "the image's reset vector is not an rjmp — pureboot's vector "
+                "surgery cannot re-home it (crt-less entry at address 0?)"
+            )
+        entry = rjmp_target(0, word0, flash_words)
+        if entry >= info.base // 2:
+            raise Error(
+                "the image's reset vector already targets the loader — this "
+                "is a read-back of a patched image; flash the original"
+            )
+        trampoline_word = (info.base - 2) // 2
+        patch = rjmp_to(0, info.base // 2, flash_words)
+        final[0], final[1] = patch & 0xFF, patch >> 8
+        trampoline_page = info.base - page
+        if len(final) < trampoline_page + page:
+            final += bytearray([0xFF] * (trampoline_page + page - len(final)))
+        jump = rjmp_to(trampoline_word, entry, flash_words)
+        final[info.base - 2], final[info.base - 1] = jump & 0xFF, jump >> 8
+
+    pages = {a: bytes(final[a : a + page]) for a in range(0, len(final), page)}
+    return pages
+
+
+def covered(pages, skip_blank):
+    """Pages in programming order: ascending, page 0 last; optionally
+    dropping all-0xff pages (sound only over erased flash) — never the
+    load-bearing page 0."""
+    rest = [a for a in sorted(pages) if a != 0]
+    if skip_blank:
+        rest = [a for a in rest if pages[a].count(0xFF) != len(pages[a])]
+    return rest + [0]
+
+
+# ------------------------------------------------------------ operations ---
+
+
+def op_erase_flash(loader):
+    """0xff over the whole application area. Order does not matter here —
+    every target byte is the same value — and a blank word 0 still falls
+    through to the loader, so an interruption is harmless."""
+    blank = bytes([0xFF] * loader.info.page)
+    for address in range(0, loader.info.base, loader.info.page):
+        loader.write_page(address, blank)
+    print(f"erase: {loader.info.base // loader.info.page} pages")
+
+
+def op_erase_eeprom(loader):
+    loader.write_eeprom(0, bytes([0xFF] * loader.info.eeprom_size))
+    print(f"erase: {loader.info.eeprom_size} B of EEPROM")
+
+
+def op_flash(loader, path, erase, verify):
+    image = load_image(path)
+    pages = plan_flash(image, loader.info)
+    if erase:
+        op_erase_flash(loader)
+    order = covered(pages, skip_blank=erase)
+    for address in order:
+        loader.write_page(address, pages[address])
+    print(f"flash: {path}: {len(order)} pages")
+    if verify:
+        verify_pages(loader, pages)
+
+
+def verify_pages(loader, pages):
+    for address in sorted(pages):
+        got = loader.read_flash(address, loader.info.page)
+        if got != pages[address]:
+            first = next(i for i in range(len(got)) if got[i] != pages[address][i])
+            raise Error(
+                f"verify failed at {address + first:#06x}: "
+                f"wrote {pages[address][first]:02x}, read {got[first]:02x}"
+            )
+    print(f"verify: {len(pages)} pages ok")
+
+
+def op_verify_flash(loader, path):
+    verify_pages(loader, plan_flash(load_image(path), loader.info))
+
+
+def op_read_flash(loader, path):
+    data = loader.read_flash(0, loader.info.base)
+    open(path, "wb").write(data)
+    print(f"read flash: {len(data)} B -> {path}")
+
+
+def op_eeprom(loader, path, erase, verify):
+    image = load_image(path)
+    if len(image) > loader.info.eeprom_size:
+        raise Error(f"EEPROM image is {len(image)} B, device has {loader.info.eeprom_size}")
+    if erase:
+        op_erase_eeprom(loader)
+    loader.write_eeprom(0, image)
+    print(f"eeprom: {path}: {len(image)} B")
+    if verify:
+        got = loader.read_eeprom(0, len(image))
+        if got != image:
+            first = next(i for i in range(len(got)) if got[i] != image[i])
+            raise Error(f"verify failed at EEPROM {first:#06x}: wrote {image[first]:02x}, read {got[first]:02x}")
+        print(f"verify: {len(image)} B ok")
+
+
+def op_verify_eeprom(loader, path):
+    image = load_image(path)
+    got = loader.read_eeprom(0, len(image))
+    if got != image:
+        first = next(i for i in range(len(got)) if got[i] != image[i])
+        raise Error(f"verify failed at EEPROM {first:#06x}: expected {image[first]:02x}, read {got[first]:02x}")
+    print(f"verify: {len(image)} B of EEPROM ok")
+
+
+def op_read_eeprom(loader, path):
+    data = loader.read_eeprom(0, loader.info.eeprom_size)
+    open(path, "wb").write(data)
+    print(f"read EEPROM: {len(data)} B -> {path}")
+
+
+def op_timeout(loader, seconds):
+    loader.write_eeprom(loader.info.eeprom_size - 1, bytes((seconds,)))
+    label = f"{seconds} s" if seconds else "the device default"
+    print(f"activation timeout: {label}")
+
+
+def op_fuses(loader):
+    low, lock, extended, high = loader.read_fuses()
+    print(f"fuses: low {low:02x} high {high:02x} extended {extended:02x} lock {lock:02x}")
+
+
+# -------------------------------------------------------------------- cli ---
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="pureboot host tool", epilog="operations run in the order listed above"
+    )
+    parser.add_argument("--port", required=True, help="serial device (or simavr pty)")
+    parser.add_argument("--baud", type=int, default=115200, help="115200 mega, 57600 tinies")
+    parser.add_argument("--wait", type=float, default=30.0, help="seconds to keep knocking")
+    parser.add_argument("--info", action="store_true", help="print the device info block")
+    parser.add_argument("--fuses", action="store_true", help="read the fuse and lock bytes")
+    parser.add_argument("--erase-flash", action="store_true", help="0xff over the application flash")
+    parser.add_argument("--flash", metavar="FILE", help="program an application (bin or ihex)")
+    parser.add_argument("--no-verify", action="store_true", help="skip read-back after writes")
+    parser.add_argument("--read-flash", metavar="FILE", help="dump the application flash")
+    parser.add_argument("--verify-flash", metavar="FILE", help="compare flash against an image")
+    parser.add_argument("--erase-eeprom", action="store_true", help="0xff over the EEPROM")
+    parser.add_argument("--eeprom", metavar="FILE", help="program the EEPROM (bin or ihex)")
+    parser.add_argument("--read-eeprom", metavar="FILE", help="dump the EEPROM")
+    parser.add_argument("--verify-eeprom", metavar="FILE", help="compare EEPROM against an image")
+    parser.add_argument("--timeout", type=int, metavar="S", help="activation window, 1-254 s (0: default)")
+    parser.add_argument("--stay", action="store_true", help="leave the loader in its session")
+    args = parser.parse_args()
+
+    if args.timeout is not None and not 0 <= args.timeout <= 254:
+        parser.error("--timeout must be 0..254 (255 is the erased cell)")
+
+    port = Port(args.port, args.baud)
+    try:
+        loader = Loader(port)
+        info = loader.connect(args.wait)
+        if args.info:
+            print(f"device: {info.describe()}")
+        if args.fuses:
+            op_fuses(loader)
+        if args.flash:
+            op_flash(loader, args.flash, args.erase_flash, not args.no_verify)
+        elif args.erase_flash:
+            op_erase_flash(loader)
+        if args.read_flash:
+            op_read_flash(loader, args.read_flash)
+        if args.verify_flash:
+            op_verify_flash(loader, args.verify_flash)
+        if args.eeprom:
+            op_eeprom(loader, args.eeprom, args.erase_eeprom, not args.no_verify)
+        elif args.erase_eeprom:
+            op_erase_eeprom(loader)
+        if args.read_eeprom:
+            op_read_eeprom(loader, args.read_eeprom)
+        if args.verify_eeprom:
+            op_verify_eeprom(loader, args.verify_eeprom)
+        if args.timeout is not None:
+            op_timeout(loader, args.timeout)
+        if args.stay:
+            print("loader stays in its session (reset to leave)")
+        else:
+            loader.run_application()
+            print("application running")
+    finally:
+        port.close()
+
+
+if __name__ == "__main__":
+    try:
+        main()
+    except Error as error:
+        print(f"error: {error}", file=sys.stderr)
+        sys.exit(1)
+    except KeyboardInterrupt:
+        sys.exit(130)

test/check_size.cmake

@@ -4,9 +4,6 @@ if(NOT _res EQUAL 0)
 endif()
 # avr-size line 2 is "<text> <data> <bss> <dec> <hex> <file>".
 string(REGEX MATCH "\n[ \t]*([0-9]+)" _m "${_out}")
-if(NOT _m)
-	message(FATAL_ERROR "could not read a .text size out of ${SIZE_TOOL}'s output for ${ELF}:\n${_out}")
-endif()
 set(_text ${CMAKE_MATCH_1})
 if(_text GREATER LIMIT)
 	message(FATAL_ERROR ".text is ${_text} bytes, over the ${LIMIT}-byte boot section")

test/device.c

@@ -7,122 +7,103 @@
 // SPM genuinely writes avr->flash on the mega cores, so on exit (or SIGTERM)
 // we dump the flash image to a file for a ground-truth cross-check against
 // what the client read back through the bootloader.
-#include <array>
-#include <csignal>
-#include <cstdint>
-#include <cstdio>
-#include <cstdlib>
-#include <cstring>
-#include <print>
-
+#include <signal.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
 #include <unistd.h>
 
-// The parts headers (uart_pty.h) carry no C++ linkage guards of their own,
-// unlike simavr's core headers - the block covers both harmlessly.
-extern "C" {
 #include "avr_uart.h"
 #include "sim_avr.h"
 #include "sim_elf.h"
 #include "uart_pty.h"
-}
 
-namespace {
+static avr_t *avr;
+static uart_pty_t uart_pty;
+static const char *dump_path;
 
-avr_t *avr;
-uart_pty_t uart_pty;
-const char *dump_path;
-
-[[noreturn]] void finish(int)
+static void finish(int sig)
 {
+	(void)sig;
 	if (dump_path) {
-		std::FILE *f = std::fopen(dump_path, "wb");
+		FILE *f = fopen(dump_path, "wb");
 		if (f) {
-			std::fwrite(avr->flash, 1, avr->flashend + 1, f);
-			std::fclose(f);
+			fwrite(avr->flash, 1, avr->flashend + 1, f);
+			fclose(f);
 		}
 	}
 	uart_pty_stop(&uart_pty);
 	_exit(0);
 }
 
-} // namespace
-
 int main(int argc, char *argv[])
 {
 	if (argc < 3) {
-		std::println(stderr, "usage: {} <tsb.elf> <boot_base_hex> [flash_dump.bin]", argv[0]);
+		fprintf(stderr, "usage: %s <tsb.elf> <boot_base_hex> [flash_dump.bin]\n", argv[0]);
 		return 2;
 	}
-	auto boot_base = static_cast<std::uint32_t>(std::strtoul(argv[2], nullptr, 0));
-	dump_path = argc >= 4 ? argv[3] : nullptr;
+	uint32_t boot_base = (uint32_t)strtoul(argv[2], NULL, 0);
+	dump_path = argc >= 4 ? argv[3] : NULL;
 
 	avr = avr_make_mcu_by_name("atmega328p");
 	if (!avr) {
-		std::println(stderr, "device: no ATmega328P core");
+		fprintf(stderr, "device: no ATmega328P core\n");
 		return 1;
 	}
 	avr_init(avr);
 	avr->frequency = 16000000;
 	// Real flash powers up erased (0xff); the app region must look erased
 	// before the bootloader programs it.
-	std::memset(avr->flash, 0xff, avr->flashend + 1);
+	memset(avr->flash, 0xff, avr->flashend + 1);
 
 	// simavr's ELF loader flattens the flash base to 0 (it expects an app at
 	// 0x0), but it hands back the boot code in fw.flash; place it at the boot
 	// section base ourselves and enter there (BOOTRST is not modelled).
-	elf_firmware_t fw{};
+	elf_firmware_t fw = {0};
 	if (elf_read_firmware(argv[1], &fw) != 0) {
-		std::println(stderr, "device: cannot read {}", argv[1]);
+		fprintf(stderr, "device: cannot read %s\n", argv[1]);
 		return 1;
 	}
-	// An image that runs past flash end cannot execute on hardware, and a
-	// naive copy of it would smash the heap beyond avr->flash - after which
-	// the simulation misbehaves in ways that point everywhere but here.
-	// Refuse it loudly instead.
-	if (boot_base + fw.flashsize > avr->flashend + 1) {
-		std::println(stderr, "device: {} B at {:#x} runs past flash end {:#x} - image does not fit its slot",
-		             fw.flashsize, boot_base, avr->flashend);
-		return 1;
-	}
-	std::memcpy(avr->flash + boot_base, fw.flash, fw.flashsize);
+	memcpy(avr->flash + boot_base, fw.flash, fw.flashsize);
 	avr->pc = boot_base;
 	avr->codeend = avr->flashend;
 
 	// Optional: seed the config page (one page below the boot section) with a
 	// hex byte string, so the password gate and emergency erase can be tested.
 	// Layout: [appjump lo][appjump hi][timeout][password...][0xff].
-	const char *cfg = std::getenv("TSB_CONFIG");
+	const char *cfg = getenv("TSB_CONFIG");
 	if (cfg) {
-		std::uint32_t app_end = boot_base - 128; // config page sits directly below the boot code
+		uint32_t app_end = boot_base - 128; // config page sits directly below the boot code
 		for (int i = 0; cfg[i] && cfg[i + 1]; i += 2) {
-			const std::array pair{cfg[i], cfg[i + 1], '\0'};
-			avr->flash[app_end + i / 2] = static_cast<std::uint8_t>(std::strtoul(pair.data(), nullptr, 16));
+			char b[3] = {cfg[i], cfg[i + 1], 0};
+			avr->flash[app_end + i / 2] = (uint8_t)strtoul(b, NULL, 16);
 		}
 	}
 
 	// POLL_SLEEP makes simavr usleep(1) on every status-register read while the
-	// UART is idle - a host-CPU-saving hack that models no hardware and paces a
+	// UART is idle — a host-CPU-saving hack that models no hardware and paces a
 	// tight-polling loader (one that releases TX between bytes, as one-wire does)
 	// in real time, distorting protocol timing. Clear it so the loader runs at
 	// true cycle speed.
-	std::uint32_t uflags = 0;
+	uint32_t uflags = 0;
 	avr_ioctl(avr, AVR_IOCTL_UART_GET_FLAGS('0'), &uflags);
 	uflags &= ~AVR_UART_FLAG_POLL_SLEEP;
 	avr_ioctl(avr, AVR_IOCTL_UART_SET_FLAGS('0'), &uflags);
 
 	uart_pty_init(avr, &uart_pty);
 	uart_pty_connect(&uart_pty, '0');
-	std::println("TSB_PTY {}", uart_pty.pty.slavename);
-	std::fflush(stdout);
+	printf("TSB_PTY %s\n", uart_pty.pty.slavename);
+	fflush(stdout);
 
-	std::signal(SIGTERM, finish);
-	std::signal(SIGINT, finish);
+	signal(SIGTERM, finish);
+	signal(SIGINT, finish);
 
 	for (;;) {
 		int state = avr_run(avr);
-		if (state == cpu_Done || state == cpu_Crashed) {
+		if (state == cpu_Done || state == cpu_Crashed)
 			break;
-		}
 	}
 	finish(0);
+	return 0;
 }

test/pbapp.cpp

@@ -0,0 +1,51 @@
+// Test-fixture application for the pureboot protocol test: prints "APP" on
+// the chip's serial link (the same link the loader uses) and idles — the
+// proof that the loader's hand-over, and on the tinies the host's
+// reset-vector surgery, actually launched it. Linked normally (crt, vectors
+// at 0); on the tinies its reset vector is the rjmp the host re-homes.
+#include <libavr/libavr.hpp>
+
+using namespace avr::literals;
+
+namespace {
+
+consteval avr::hertz_t clock()
+{
+	if (avr::hw::db.name == "ATtiny13A")
+		return 9.6_MHz;
+	if (avr::hw::db.name == "ATtiny85")
+		return 8_MHz;
+	return 16_MHz;
+}
+
+using dev = avr::device<{.clock = clock()}>;
+
+template <avr::hertz_t C, bool Hardware = avr::hw::db.has_reg("UDR0")>
+struct link {
+	using tx_t = avr::uart::usart0<C, {.baud = 115200_Bd, .max_baud_error = 2.5_pct}>;
+	static void tx(char c)
+	{
+		tx_t::write(static_cast<std::uint8_t>(c));
+	}
+};
+
+template <avr::hertz_t C>
+struct link<C, false> {
+	using tx_t = avr::uart::software_tx<C, avr::pb1, 57600_Bd>;
+	static void tx(char c)
+	{
+		tx_t::write(static_cast<std::uint8_t>(c));
+	}
+};
+
+} // namespace
+
+int main()
+{
+	avr::init<typename link<dev::clock>::tx_t>();
+	link<dev::clock>::tx('A');
+	link<dev::clock>::tx('P');
+	link<dev::clock>::tx('P');
+	while (true) {
+	}
+}

test/pbtest.py

@@ -0,0 +1,178 @@
+#!/usr/bin/env python3
+"""End-to-end pureboot protocol test: spawn the simavr device, then drive it
+with the real host tool (pureboot.py, as a subprocess over the device's pty)
+through flash + EEPROM + timeout + fuse + hand-over scenarios, and cross-check
+the tool's view against the simulator's ground-truth memory dumps.
+
+Usage: pbtest.py <device_bin> <pureboot_elf> <mcu> <hz> <base_hex> <page>
+                 <baud> <eeprom_size> <app_bin> <tool_py> <workdir>
+Exits 0 if every scenario passes.
+"""
+
+import os
+import signal
+import subprocess
+import sys
+import time
+
+
+def fail(message):
+    print(f"FAIL: {message}")
+    sys.exit(1)
+
+
+def rjmp_decode(word, at, flash_words):
+    """Where an rjmp word at word-address `at` lands — deliberately written
+    against the instruction-set definition (12-bit signed offset), not with
+    the host tool's encoder, so an encoding bug cannot verify itself."""
+    if word & 0xF000 != 0xC000:
+        fail(f"word at {at * 2:#06x} is {word:#06x}, not an rjmp")
+    offset = word & 0x0FFF
+    if offset >= 0x800:
+        offset -= 0x1000
+    return (at + 1 + offset) % flash_words
+
+
+class Device:
+    def __init__(self, binary, elf, mcu, hz, base, page, baud, dump):
+        self.proc = subprocess.Popen(
+            [binary, elf, mcu, hz, base, str(page), str(baud), dump],
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+            text=True,
+        )
+        self.dump = dump
+        self.pty = None
+        deadline = time.time() + 5
+        while time.time() < deadline:
+            line = self.proc.stdout.readline()
+            if not line:
+                break
+            if line.startswith("PB_PTY"):
+                self.pty = line.split()[1]
+                break
+        if not self.pty:
+            self.stop()
+            raise RuntimeError("device did not report a pty")
+
+    def stop(self):
+        self.proc.terminate()
+        try:
+            self.proc.wait(timeout=3)
+        except subprocess.TimeoutExpired:
+            self.proc.kill()
+
+
+def run_tool(tool, pty, baud, *args):
+    result = subprocess.run(
+        [sys.executable, tool, "--port", pty, "--baud", str(baud), "--wait", "20", *args],
+        capture_output=True,
+        text=True,
+        timeout=120,
+    )
+    print(result.stdout, end="")
+    if result.returncode != 0:
+        fail(f"tool exited {result.returncode}: {result.stderr.strip()}")
+    return result.stdout
+
+
+def main():
+    (device_bin, elf, mcu, hz, base_hex, page, baud, eeprom_size, app_bin, tool, workdir) = sys.argv[1:]
+    base, page, baud, eeprom_size = int(base_hex, 0), int(page), int(baud), int(eeprom_size)
+    sys.path.insert(0, os.path.dirname(os.path.abspath(tool)))
+    import pureboot as pb
+
+    os.makedirs(workdir, exist_ok=True)
+    ee_image = bytes(range(0xA0, 0xB0))
+    ee_path = os.path.join(workdir, "ee.bin")
+    open(ee_path, "wb").write(ee_image)
+    dump = os.path.join(workdir, "flash_dump.bin")
+    read_flash = os.path.join(workdir, "readback_flash.bin")
+    read_eeprom = os.path.join(workdir, "readback_eeprom.bin")
+
+    # The geometry the host will discover, for computing the expected image.
+    info = pb.Info(
+        bytes([ord("P"), ord("B"), 1, 0, 0, 0, page])
+        + bytes([base & 0xFF, base >> 8, eeprom_size & 0xFF, eeprom_size >> 8])
+        + bytes([0 if mcu == "atmega328p" else 1])
+    )
+
+    device = Device(device_bin, elf, mcu, hz, base_hex, page, baud, dump)
+    try:
+        # Session 1: knock from reset, identify, program everything, stay.
+        out = run_tool(tool, device.pty, baud, "--info", "--fuses", "--flash", app_bin,
+                       "--eeprom", ee_path, "--timeout", "8", "--stay")
+        for needed in ("device: signature", "fuses:", "verify:", "activation timeout: 8 s", "stays"):
+            if needed not in out:
+                fail(f"session 1 output lacks {needed!r}")
+
+        # Session 2: reconnect into the live session, verify, dump, hand over
+        # is deferred — the pty must be reopened for the APP banner first.
+        out = run_tool(tool, device.pty, baud, "--verify-flash", app_bin, "--verify-eeprom", ee_path,
+                       "--read-flash", read_flash, "--read-eeprom", read_eeprom, "--stay")
+        if out.count("verify:") != 2:
+            fail("session 2 did not verify both memories")
+
+        eeprom_back = open(read_eeprom, "rb").read()
+        if eeprom_back[: len(ee_image)] != ee_image:
+            fail("EEPROM read-back mismatch")
+        if eeprom_back[-1] != 8:
+            fail(f"timeout cell reads {eeprom_back[-1]}, expected 8")
+
+        # The expected post-surgery flash, straight from the tool's planner.
+        pages = pb.plan_flash(open(app_bin, "rb").read(), info)
+        flash_back = open(read_flash, "rb").read()
+        for address, data in pages.items():
+            if flash_back[address : address + page] != data:
+                fail(f"flash read-back mismatch in page {address:#06x}")
+
+        # An external reset re-enters through the patched word 0 (tinies; the
+        # runner resets them to address 0 like silicon) or BOOTRST (mega).
+        # The loader must answer a fresh knock, and 'G' must land in the
+        # application, which banners on the same link.
+        device.proc.send_signal(signal.SIGUSR1)
+        port = pb.Port(device.pty, baud)
+        try:
+            loader = pb.Loader(port)
+            loader.connect(15)
+            port.write(b"G")
+            if port.read_exact(1, 5.0) != pb.PROMPT:
+                fail("no ack for G")
+            banner = port.read_exact(3, 5.0)
+            if banner != b"APP":
+                fail(f"application banner was {banner!r}")
+        finally:
+            port.close()
+    finally:
+        device.stop()
+
+    # Ground truth: the simulator's own memories, against the host's view.
+    flash_true = open(dump, "rb").read()
+    if flash_true[:base] != flash_back:
+        fail("host flash read-back differs from the simulator's flash")
+    if flash_true[base] == 0xFF and flash_true[base + 1] == 0xFF:
+        fail("loader region looks erased in the ground-truth dump")
+
+    # The surgery, decoded independently: the patched vector must land on the
+    # loader, the trampoline on the application's own entry.
+    if mcu != "atmega328p":
+        flash_words = (base + 512) // 2
+        app = open(app_bin, "rb").read()
+        word0 = flash_true[0] | (flash_true[1] << 8)
+        if rjmp_decode(word0, 0, flash_words) != base // 2:
+            fail("patched reset vector does not land on the loader base")
+        trampoline = flash_true[base - 2] | (flash_true[base - 1] << 8)
+        original = app[0] | (app[1] << 8)
+        if rjmp_decode(trampoline, (base - 2) // 2, flash_words) != rjmp_decode(original, 0, flash_words):
+            fail("trampoline does not land on the application's own entry")
+    ee_true_path = dump + ".eeprom"
+    if os.path.exists(ee_true_path):
+        ee_true = open(ee_true_path, "rb").read()
+        if ee_true[: len(ee_image)] != ee_image or ee_true[-1] != 8:
+            fail("ground-truth EEPROM does not match what was programmed")
+
+    print("pbtest: all scenarios pass")
+
+
+if __name__ == "__main__":
+    main()

test/pureboot_device.c

@@ -0,0 +1,323 @@
+// simavr "device" for the pureboot protocol tests, all three chips. Loads
+// the boot-linked ELF at the loader base, starts execution there (BOOTRST /
+// the patched vector are not what is under test), and exposes the loader's
+// serial link as a pty for the real host tool:
+//
+//   - ATmega328P: the hardware USART0 through simavr's uart_pty.
+//   - Tinies: an 8N1 bridge between a pty and the GPIO software UART
+//     (drives PB0, the loader's RX; decodes PB1, its TX), timed against the
+//     simulated cycle counter.
+//
+// simavr's tiny cores decode the SPM opcode but attach no NVM module — SPM
+// is a silent no-op (the mega's boot section has one, avr_flash). The
+// missing module is supplied here: the SPM ioctl reads SPMCSR/Z/r1:r0 and
+// implements buffer fill, page erase, page write, and CTPB, completing
+// instantly. RFLB's LPM diversion (fuse readout) stays unmodeled, so the
+// 'F' command answers with flash bytes — the tests assert transport only.
+//
+// On exit (or SIGTERM) the flash and EEPROM are dumped to files for a
+// ground-truth cross-check against what the host read back.
+#include <fcntl.h>
+#include <pty.h>
+#include <signal.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <termios.h>
+#include <unistd.h>
+
+#include "avr_eeprom.h"
+#include "avr_flash.h"
+#include "avr_ioport.h"
+#include "avr_uart.h"
+#include "sim_avr.h"
+#include "sim_elf.h"
+#include "sim_io.h"
+#include "uart_pty.h"
+
+static avr_t *avr;
+static uart_pty_t uart_pty;
+static int use_uart_pty;
+static const char *dump_path;
+static uint32_t reset_pc;
+static volatile sig_atomic_t reset_requested;
+
+static void request_reset(int sig)
+{
+	(void)sig;
+	reset_requested = 1;
+}
+
+// ------------------------------------------------------------- tiny NVM ---
+
+typedef struct {
+	avr_io_t io;
+	uint8_t buffer[128];
+	unsigned page;
+} tiny_nvm_t;
+
+static tiny_nvm_t nvm;
+
+static int nvm_ioctl(avr_io_t *io, uint32_t ctl, void *param)
+{
+	(void)param;
+	if (ctl != AVR_IOCTL_FLASH_SPM)
+		return -1;
+	tiny_nvm_t *n = (tiny_nvm_t *)io;
+	avr_t *mcu = io->avr;
+	uint8_t command = mcu->data[0x57] & 0x1f; // SPMCSR, both tinies
+	uint16_t z = (uint16_t)(mcu->data[30] | (mcu->data[31] << 8));
+	uint32_t page_base = (uint32_t)(z & ~(n->page - 1)) % (mcu->flashend + 1);
+	if (command == 0x01) { // SPMEN alone: buffer fill from r1:r0
+		unsigned offset = z & (n->page - 1) & ~1u;
+		n->buffer[offset] = mcu->data[0];
+		n->buffer[offset + 1] = mcu->data[1];
+	} else if (command == 0x03) { // PGERS
+		memset(mcu->flash + page_base, 0xff, n->page);
+	} else if (command == 0x05) { // PGWRT: programming only clears bits
+		for (unsigned i = 0; i < n->page; i++)
+			mcu->flash[page_base + i] &= n->buffer[i];
+		memset(n->buffer, 0xff, n->page);
+	} else if (command == 0x11) { // CTPB
+		memset(n->buffer, 0xff, n->page);
+	}
+	mcu->data[0x57] &= (uint8_t)~0x1f; // the operation completes instantly
+	return 0;
+}
+
+// ----------------------------------------------------------- GPIO bridge ---
+
+static int pty_master = -1;
+static avr_irq_t *rx_pin; // the loader's RX (PB0), driven from the pty
+static avr_cycle_count_t bit_cycles;
+
+static int tx_level = 1, tx_active, tx_bit;
+static uint8_t tx_shift;
+
+static avr_cycle_count_t tx_sample(avr_t *mcu, avr_cycle_count_t when, void *param)
+{
+	(void)mcu;
+	(void)param;
+	tx_shift = (uint8_t)((tx_shift >> 1) | (tx_level ? 0x80 : 0));
+	if (++tx_bit < 8)
+		return when + bit_cycles;
+	if (write(pty_master, &tx_shift, 1) != 1)
+		fprintf(stderr, "device: pty write lost a byte\n");
+	tx_active = 0;
+	return 0;
+}
+
+static void tx_hook(avr_irq_t *irq, uint32_t value, void *param)
+{
+	(void)irq;
+	(void)param;
+	int level = value & 1;
+	if (!tx_active && tx_level == 1 && level == 0) { // start edge
+		tx_active = 1;
+		tx_bit = 0;
+		avr_cycle_timer_register(avr, bit_cycles + bit_cycles / 2, tx_sample, NULL);
+	}
+	tx_level = level;
+}
+
+static uint8_t rx_queue[8192];
+static unsigned rx_head, rx_tail; // ring: head = next to send
+static int rx_active, rx_bit;
+static uint8_t rx_byte;
+
+static void rx_start_next(void);
+
+static avr_cycle_count_t rx_step(avr_t *mcu, avr_cycle_count_t when, void *param)
+{
+	(void)mcu;
+	(void)param;
+	if (rx_bit < 8) {
+		avr_raise_irq(rx_pin, (rx_byte >> rx_bit) & 1);
+		rx_bit++;
+		return when + bit_cycles;
+	}
+	if (rx_bit == 8) { // stop bit, plus one idle bit of margin
+		avr_raise_irq(rx_pin, 1);
+		rx_bit++;
+		return when + 2 * bit_cycles;
+	}
+	rx_active = 0;
+	rx_start_next();
+	return 0;
+}
+
+static void rx_start_next(void)
+{
+	if (rx_active || rx_head == rx_tail)
+		return;
+	rx_byte = rx_queue[rx_head];
+	rx_head = (rx_head + 1) % sizeof(rx_queue);
+	rx_active = 1;
+	rx_bit = 0;
+	avr_raise_irq(rx_pin, 0); // start bit
+	avr_cycle_timer_register(avr, bit_cycles, rx_step, NULL);
+}
+
+// A reset abandons whatever the bridge was mid-transfer: bytes still queued
+// for a chip that no longer has the context to receive them meaningfully,
+// and a decode in progress on a TX line the reset may have already changed.
+static void bridge_reset(void)
+{
+	rx_head = rx_tail = 0;
+	rx_active = 0;
+	tx_active = 0;
+	tx_level = 1;
+	avr_raise_irq(rx_pin, 1); // idle line
+}
+
+static void poll_pty(void)
+{
+	uint8_t chunk[256];
+	ssize_t got = read(pty_master, chunk, sizeof(chunk));
+	for (ssize_t i = 0; i < got; i++) {
+		unsigned next = (rx_tail + 1) % sizeof(rx_queue);
+		if (next == rx_head)
+			break; // full: the host will retry on timeout
+		rx_queue[rx_tail] = chunk[i];
+		rx_tail = next;
+	}
+	if (got > 0)
+		rx_start_next();
+}
+
+// ------------------------------------------------------------------ main ---
+
+static void finish(int sig)
+{
+	(void)sig;
+	if (dump_path) {
+		FILE *f = fopen(dump_path, "wb");
+		if (f) {
+			fwrite(avr->flash, 1, avr->flashend + 1, f);
+			fclose(f);
+		}
+		avr_eeprom_desc_t ee = {.ee = NULL, .offset = 0, .size = 0};
+		if (avr_ioctl(avr, AVR_IOCTL_EEPROM_GET, &ee) == 0 && ee.ee && ee.size) {
+			char path[512];
+			snprintf(path, sizeof(path), "%s.eeprom", dump_path);
+			f = fopen(path, "wb");
+			if (f) {
+				fwrite(ee.ee, 1, ee.size, f);
+				fclose(f);
+			}
+		}
+	}
+	if (use_uart_pty)
+		uart_pty_stop(&uart_pty);
+	_exit(0);
+}
+
+int main(int argc, char *argv[])
+{
+	if (argc != 8) {
+		fprintf(stderr, "usage: %s <pureboot.elf> <mcu> <hz> <base_hex> <page> <baud> <flash_dump>\n", argv[0]);
+		return 2;
+	}
+	const char *mcu_name = argv[2];
+	uint32_t base = (uint32_t)strtoul(argv[4], NULL, 0);
+	unsigned page = (unsigned)atoi(argv[5]);
+	unsigned baud = (unsigned)atoi(argv[6]);
+	dump_path = argv[7];
+	use_uart_pty = strcmp(mcu_name, "atmega328p") == 0;
+
+	avr = avr_make_mcu_by_name(mcu_name);
+	if (!avr) {
+		fprintf(stderr, "device: no %s core\n", mcu_name);
+		return 1;
+	}
+	avr_init(avr);
+	avr->frequency = (uint32_t)strtoul(argv[3], NULL, 0);
+	memset(avr->flash, 0xff, avr->flashend + 1); // real flash powers up erased
+
+	elf_firmware_t fw = {0};
+	if (elf_read_firmware(argv[1], &fw) != 0) {
+		fprintf(stderr, "device: cannot read %s\n", argv[1]);
+		return 1;
+	}
+	memcpy(avr->flash + base, fw.flash, fw.flashsize);
+	// The mega enters the loader in hardware (BOOTRST, not modeled); the
+	// tinies reset to word 0 like silicon — erased flash walks up into the
+	// loader, and after the host's surgery the patched vector routes there.
+	reset_pc = use_uart_pty ? base : 0;
+	avr->pc = reset_pc;
+	avr->codeend = avr->flashend;
+
+	// Erased EEPROM, as hardware powers up (simavr zeroes it).
+	uint8_t blank[1024];
+	memset(blank, 0xff, sizeof(blank));
+	avr_eeprom_desc_t seed = {.ee = blank, .offset = 0, .size = 0};
+	if (avr_ioctl(avr, AVR_IOCTL_EEPROM_GET, &seed) == 0 && seed.size <= sizeof(blank)) {
+		seed.ee = blank;
+		avr_ioctl(avr, AVR_IOCTL_EEPROM_SET, &seed);
+	}
+
+	if (use_uart_pty) {
+		// POLL_SLEEP paces an idle-polling loader in host real time (a
+		// no-hardware CPU-saving hack); clear it so cycles run free.
+		uint32_t flags = 0;
+		avr_ioctl(avr, AVR_IOCTL_UART_GET_FLAGS('0'), &flags);
+		flags &= ~AVR_UART_FLAG_POLL_SLEEP;
+		avr_ioctl(avr, AVR_IOCTL_UART_SET_FLAGS('0'), &flags);
+		uart_pty_init(avr, &uart_pty);
+		uart_pty_connect(&uart_pty, '0');
+		printf("PB_PTY %s\n", uart_pty.pty.slavename);
+	} else {
+		nvm.page = page;
+		memset(nvm.buffer, 0xff, sizeof(nvm.buffer));
+		nvm.io.kind = "tiny_nvm";
+		nvm.io.ioctl = nvm_ioctl;
+		avr_register_io(avr, &nvm.io);
+
+		bit_cycles = (avr->frequency + baud / 2) / baud; // matches uart.hpp's own rounding exactly
+		rx_pin = avr_io_getirq(avr, AVR_IOCTL_IOPORT_GETIRQ('B'), 0);
+		avr_irq_register_notify(avr_io_getirq(avr, AVR_IOCTL_IOPORT_GETIRQ('B'), 1), tx_hook, NULL);
+		avr_raise_irq(rx_pin, 1); // idle line
+
+		int slave;
+		struct termios raw;
+		cfmakeraw(&raw);
+		if (openpty(&pty_master, &slave, NULL, &raw, NULL) != 0) {
+			fprintf(stderr, "device: openpty failed\n");
+			return 1;
+		}
+		fcntl(pty_master, F_SETFL, O_NONBLOCK);
+		printf("PB_PTY %s\n", ttyname(slave));
+	}
+	fflush(stdout);
+
+	signal(SIGTERM, finish);
+	signal(SIGINT, finish);
+	signal(SIGUSR1, request_reset); // an external reset line, for the tests
+
+	long since_poll = 0;
+	for (;;) {
+		int state = avr_run(avr);
+		if (state == cpu_Done || state == cpu_Crashed)
+			break;
+		if (reset_requested) {
+			reset_requested = 0;
+			avr_reset(avr);
+			avr->pc = reset_pc;
+			if (use_uart_pty) { // reset restores the pacing hack; re-clear it
+				uint32_t flags = 0;
+				avr_ioctl(avr, AVR_IOCTL_UART_GET_FLAGS('0'), &flags);
+				flags &= ~AVR_UART_FLAG_POLL_SLEEP;
+				avr_ioctl(avr, AVR_IOCTL_UART_SET_FLAGS('0'), &flags);
+			} else {
+				bridge_reset();
+			}
+		}
+		if (!use_uart_pty && ++since_poll >= 2000) {
+			since_poll = 0;
+			poll_pty();
+		}
+	}
+	finish(0);
+	return 0;
+}

test/tsbtest.py

@@ -144,7 +144,7 @@ class Host:
         self._expect(CONFIRM, "C end")
         return echo
 
-    # Activation when the config page carries a password: 3x'@' then the
+    # Activation when the config page carries a password: 3×'@' then the
     # password bytes, then the info block + mainloop '!'.
     def activate_password(self, password):
         self.s.reset_input_buffer()
@@ -167,20 +167,6 @@ class Host:
         self._expect(CONFIRM, "emergency mainloop ready")
 
 
-    # A wrong password byte hangs the loader, still draining the line. Two
-    # things must not happen: it must not activate, and it must not fall
-    # through to the emergency erase - a byte the gate has already refused
-    # reaching the erase would let a guess wipe the part.
-    def refuse_password(self, byte):
-        self.s.reset_input_buffer()
-        self.s.write(bytes([KNOCK, KNOCK, KNOCK, byte]))
-        return self.s.read(1)
-
-    def say(self, byte):
-        self.s.write(bytes([byte]))
-        return self.s.read(1)
-
-
 def check(cond, msg):
     if not cond:
         raise AssertionError(msg)
@@ -195,7 +181,7 @@ PW_BYTES = bytes([0x50, 0x57])
 
 def scenario_roundtrip(host):
     """Activation + info block + flash/EEPROM/config read-write round-trips, on
-    a device with a blank (erased) config page - the usual no-password case."""
+    a device with a blank (erased) config page — the usual no-password case."""
     info = host.activate()
     check(info[0:3] == b"TSB", f"magic 'TSB' (got {info[0:3]!r})")
     check(info[6:9] == bytes([0x1E, 0x95, 0x0F]), f"signature 1E 95 0F (got {info[6:9].hex()})")
@@ -233,15 +219,6 @@ def scenario_emergency(host):
     check(host.read_eeprom(1) == b"\xff" * PAGE, "EEPROM wiped")
 
 
-def scenario_wrong_password(host):
-    """A wrong password byte neither activates the loader nor opens the
-    emergency erase behind it - the oracle carries a dedicated fix for the
-    second, and nothing here exercised either half."""
-    check(host.refuse_password(PW_BYTES[0] ^ 1) == b"", "a wrong password byte draws no reply")
-    check(host.say(0x00) == b"", "a 0 byte after it does not request the erase")
-    check(host.say(CONFIRM) == b"", "and neither does a confirm")
-
-
 def main():
     binary, elf, boot_base = sys.argv[1], sys.argv[2], sys.argv[3]
     failures = []
@@ -252,7 +229,6 @@ def main():
         ("round-trip", None, scenario_roundtrip),
         ("password activation", PW_CONFIG, scenario_password),
         ("emergency erase", PW_CONFIG, scenario_emergency),
-        ("wrong password", PW_CONFIG, scenario_wrong_password),
     ]
     for name, config, fn in groups:
         print(f"--- {name} ---")

tools/check.sh

@@ -1,60 +0,0 @@
-#!/bin/bash
-# The port's gate: the generated workflow - build, size tests, and the
-# simulator-driven protocol suite. --full adds the reflect build, which
-# compiles the same TUs through libavr's other producer. libavr resolves from
-# the `libavr/` submodule; LIBAVR_ROOT overrides it for a working tree.
-set -e
-cd "$(dirname "$0")/.."
-
-full=0
-[[ "$1" == "--full" ]] && { full=1; shift; }
-
-# The chip lists come from the presets rather than being spelled a second time
-# here: a chip added to make_presets.py and missed in a copy of its list would
-# be a gate that silently never builds it, which is the one failure mode a gate
-# cannot report. tools/make_presets.py is the single source, CMakePresets.json
-# is its output, and this reads that.
-readarray -t WORKFLOWS < <(python3 -c '
-import json, sys
-presets = json.load(open("CMakePresets.json"))["workflowPresets"]
-print("\n".join(p["name"] for p in presets))')
-if ((${#WORKFLOWS[@]} == 0)); then
-	echo "no workflow presets in CMakePresets.json - run tools/make_presets.py" >&2
-	exit 1
-fi
-CHIPS=()
-REFLECT_SPOT=()
-for workflow in "${WORKFLOWS[@]}"; do
-	case $workflow in
-	*-generated) CHIPS+=("${workflow%-generated}") ;;
-	*-reflect) REFLECT_SPOT+=("${workflow%-reflect}") ;;
-	esac
-done
-
-# Every preset runs even after one goes red, and the gate fails at the end
-# naming all of them: stopping at the first failure turns a red - a stale size
-# canary above all - into an alibi for every chip behind it, and a loader can
-# ship on a chip this gate has not compiled since.
-red=()
-run_preset() {
-	echo "==== $1 ===="
-	cmake --workflow --preset "$1" "${@:2}" || red+=("$1")
-}
-
-for chip in "${CHIPS[@]}"; do
-	run_preset "$chip-generated" "$@"
-done
-
-if ((full)); then
-	for chip in "${REFLECT_SPOT[@]}"; do
-		run_preset "$chip-reflect" "$@"
-	done
-fi
-
-if ((${#red[@]})); then
-	printf '==== red presets ====\n' >&2
-	printf '  %s\n' "${red[@]}" >&2
-	exit 1
-fi
-
-echo "check: every chip green"

tools/make_presets.py

@@ -1,90 +0,0 @@
-#!/usr/bin/env python3
-"""Regenerate CMakePresets.json - one uniform pipeline per chip.
-
-The tiers reimplement the ATmega328P-only reference protocol, so that is the
-whole chip list. It stays a generated file rather than a hand-written one
-because the shape - configure, build, test, workflow, and a reflect pair
-without tests - is the shape a second chip would need too.
-
-Run from the repo root: tools/make_presets.py - or with --check, which
-verifies the committed file matches this generator and edits nothing (the
-ctest entry `presets.generated` runs that, so drift reds the gate).
-"""
-
-import json
-import os
-import sys
-
-CHIPS = [
-    "atmega328p",
-]
-
-# The reflect pair: the same chip, built in libavr's other mode.
-REFLECT_SPOT = [
-    "atmega328p",
-]
-
-
-def main():
-    configure = [{
-        "name": "base",
-        "hidden": True,
-        "generator": "Ninja",
-        "binaryDir": "${sourceDir}/build/${presetName}",
-        "toolchainFile": "${sourceDir}/libavr/cmake/avr-toolchain.cmake",
-        "cacheVariables": {
-            "CMAKE_BUILD_TYPE": "Release",
-            "CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
-            "CMAKE_COLOR_DIAGNOSTICS": "ON",
-        },
-    }]
-    build, test, workflows = [], [], []
-
-    def add(chip, mode):
-        name = f"{chip}-{mode}"
-        configure.append({
-            "name": name,
-            "inherits": "base",
-            "cacheVariables": {
-                "LIBAVR_MCU": chip,
-                "LIBAVR_REFLECT": "ON" if mode == "reflect" else "OFF",
-            },
-        })
-        build.append({"name": name, "configurePreset": name})
-        steps = [{"type": "configure", "name": name}, {"type": "build", "name": name}]
-        if mode == "generated":
-            test.append({"name": name, "configurePreset": name, "output": {"outputOnFailure": True}})
-            steps.append({"type": "test", "name": name})
-        workflows.append({"name": name, "steps": steps})
-
-    for chip in CHIPS:
-        add(chip, "generated")
-    for chip in REFLECT_SPOT:
-        add(chip, "reflect")
-
-    # CMake rejects unknown fields in the presets root, $comment included, so
-    # the file cannot carry a generated-file marker; the --check ctest is the
-    # whole of rule 10's guard here.
-    presets = {
-        "version": 8,
-        "configurePresets": configure,
-        "buildPresets": build,
-        "testPresets": test,
-        "workflowPresets": workflows,
-    }
-    rendered = json.dumps(presets, indent=1) + "\n"
-    path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "CMakePresets.json")
-    if "--check" in sys.argv[1:]:
-        current = open(path).read() if os.path.exists(path) else ""
-        if current != rendered:
-            print("CMakePresets.json does not match its generator - run tools/make_presets.py")
-            return 1
-        return 0
-    with open(path, "w") as f:
-        f.write(rendered)
-    print(f"{len(CHIPS)} chips, {len(REFLECT_SPOT)} reflect: {os.path.normpath(path)}")
-    return 0
-
-
-if __name__ == "__main__":
-    sys.exit(main())

tsb/tsb_asm.cpp

@@ -1,15 +1,14 @@
-// TinySafeBoot on libavr - tier 3: full feature parity in the 512-byte boot
+// TinySafeBoot on libavr — tier 3: full feature parity in the 512-byte boot
 // section, in C++ except where the C ABI itself is the cost.
 //
-// The complete TinySafeBoot feature set - watchdog-reset bail, one-wire
+// The complete TinySafeBoot feature set — watchdog-reset bail, one-wire
 // half-duplex UART, a config-page activation timeout, the password gate,
-// emergency erase, and config/flash/EEPROM read-write - inside the 512-byte
-// BOOTSZ=11 section the hand-written oracle occupies (oracle/README.md holds
-// what each tier measures, in one table rather than four). The
-// body is the tricks tier's C++ (same register protocol, same structure - see
-// tsb_tricks.cpp, including the global-register miscompile rules) with exactly
-// two routines kept in assembly, the two whose remaining cost *is* the calling
-// convention:
+// emergency erase, and config/flash/EEPROM read-write — at 510 bytes in the
+// 512-byte BOOTSZ=11 section the hand-written oracle occupies (500 B). This tier used to be one
+// monolithic inline-asm routine; it is now the tricks tier's C++ (same
+// register protocol, same structure — see tsb_tricks.cpp, including the
+// global-register miscompile rules) with exactly two routines kept in
+// assembly, the two whose remaining cost *is* the calling convention:
 //
 //   rx     the bounded receive: C++ must re-floor the timeout window on every
 //          call (the global-register-store miscompile) and split it across
@@ -17,12 +16,12 @@
 //          countdown.
 //   store  the page-store loop: C++ cannot hold the receive byte pair and the
 //          walked Z pointer across the rx calls without call-saved staging
-//          (push/pop + a Y->Z copy per word); the asm calls rx knowing exactly
+//          (push/pop + a Y→Z copy per word); the asm calls rx knowing exactly
 //          which registers it touches and walks Z live across the whole page.
 //
-// Everything else - bring-up, activation, password gate, emergency erase,
+// Everything else — bring-up, activation, password gate, emergency erase,
 // dispatch, every SPM/EEPROM/flash primitive, every geometry/baud/info
-// constant - is C++ on libavr, and the two asm routines splice into the same
+// constant — is C++ on libavr, and the two asm routines splice into the same
 // global-register protocol the C++ uses (g_addr in Y, g_cnt in r16, g_window
 // in r7, g_receiving in r6), so calls cross the boundary with no marshalling.
 //
@@ -42,16 +41,10 @@ namespace hw = avr::hw;
 namespace tsb {
 namespace {
 
-// The loader is purely polled - it never enables interrupts - so every SPM and
+// The loader is purely polled — it never enables interrupts — so every SPM and
 // EEPROM lock folds to nothing under this posture.
 constexpr auto off = avr::irq::guard_policy::unused;
 
-// Strict request/response: every SPM operation is waited out before the next
-// byte moves, so no flash operation is ever in flight at an EEPROM access -
-// the write procedure's step 2 has nothing to guard, the omission the
-// datasheet grants (DS40002061B section 8.6.3).
-constexpr auto no_spm = ee::spm_interlock::omitted;
-
 constexpr std::uint8_t confirm = '!';
 constexpr std::uint8_t request = '?';
 constexpr std::uint8_t knock = '@';
@@ -64,34 +57,28 @@ constexpr std::uint16_t boot_bytes = 512;
 constexpr std::uint16_t app_end = spm::flash_bytes - boot_bytes - page;
 constexpr std::uint16_t eeprom_end = avr::hw::db.mem.eeprom_size - 1;
 
-// Lockout-proof floor for the activation window: the oracle's F_CPU/1MHz, so
-// it follows the clock rather than restating it (rule 41).
-constexpr auto act_min = static_cast<std::uint8_t>((16_MHz).hz / 1'000'000);
+// Lockout-proof floor for the activation window (the oracle's F_CPU/1MHz).
+constexpr std::uint8_t act_min = 16;
 // Post-activation window: the host gets seconds, not milliseconds, mid-session.
 constexpr std::uint8_t comm_window = 200;
 
 constexpr std::uint16_t build_date = 26 * 512 + 7 * 32 + 20;
 
 // Fixed 115200 8N1; the library solves UBRR + U2X from clock and baud.
-constexpr auto baud = avr::uart::solve_baud(16_MHz, 115200_Bd, 8, avr::uart::parity::none);
-
-// One bit time on the wire: the turn-around a shared-line peer needs to stop
-// driving before this one starts. Derived from the solved rate, so it follows
-// the link rather than a count measured against one.
-constexpr auto guard_cycles = static_cast<std::uint32_t>((16_MHz).hz / baud.actual);
+constexpr auto baud = avr::uart::detail::solve_baud(16_MHz, 115200_Bd);
 
 // The 16-byte device-info block, streamed out on activation.
 // clang-format off
-[[gnu::progmem]] constexpr auto info = std::to_array<std::uint8_t>({
+[[gnu::progmem]] constexpr std::uint8_t info[16] = {
     'T', 'S', 'B',
     build_date & 0xFF, build_date >> 8,
     0xF3,             // status: native-UART fixed-baud lineage
-    avr::hw::db.signature[0], avr::hw::db.signature[1], avr::hw::db.signature[2],
+    0x1E, 0x95, 0x0F, // ATmega328P signature
     page / 2,         // page size in words
     (app_end / 2) & 0xFF, (app_end / 2) >> 8,
     eeprom_end & 0xFF, eeprom_end >> 8,
     0xAA, 0xAA,
-});
+};
 // clang-format on
 
 register std::uint16_t g_addr asm("r28");
@@ -107,7 +94,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 // Bounded byte receive (asm 1 of 2): release the one-wire line on a direction
 // change, poll RXC0 under the oracle's nested X-register countdown seeded from
 // g_window (floored against lockout), byte or 0-on-silence in r24. Z survives
-// - the property the store's word loop rides on.
+// — the property the store's word loop rides on.
 [[gnu::noinline, gnu::noclone]] std::uint8_t rx()
 {
 	std::uint8_t byte;
@@ -128,7 +115,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	             "	brne	3b			\n\t"
 	             "	sbiw	r26, 1			\n\t"
 	             "	brcc	2b			\n\t"
-	             "	clr	%[b]			\n\t" // silence -> 0, which no compare accepts
+	             "	clr	%[b]			\n\t" // silence → 0, which no compare accepts
 	             "	rjmp	5f			\n\t"
 	             "4:	lds	%[b], %[udr0]		\n\t"
 	             "5:					\n\t"
@@ -142,13 +129,14 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 // One-wire transmit: take the line (TXEN0 alone) on a direction change with a
 // turn-around guard, put the byte out, hold the line until the whole frame is
 // out (TXC0, not UDRE0), W1C TXC0 by storing the sampled status back (keeps
-// U2X0). Plain C++ - it compiles *smaller* than the oracle's routine.
+// U2X0). Plain C++ — it compiles *smaller* than the oracle's routine.
 [[gnu::noinline, gnu::noclone]] void tx(std::uint8_t byte)
 {
 	if (g_receiving) {
 		g_receiving = 0;
 		hw::ucsr0b::write(hw::ucsr0b::txen0(1));
-		avr::delay::cycles<guard_cycles>();
+		for (std::uint8_t guard = 46; guard; --guard)
+			;
 	}
 	hw::udr0::write(byte);
 	std::uint8_t status;
@@ -165,7 +153,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	return rx();
 }
 
-// One flash byte <- [g_addr++] (the advance right before ret - the
+// One flash byte ← [g_addr++] (the advance right before ret — the
 // global-register rule, see tsb_tricks.cpp).
 [[gnu::noinline, gnu::noclone]] std::uint8_t sflash()
 {
@@ -174,18 +162,18 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	return byte;
 }
 
-// One EEPROM byte <- [g_addr++].
+// One EEPROM byte ← [g_addr++].
 [[gnu::noinline, gnu::noclone]] std::uint8_t eerd()
 {
-	std::uint8_t byte = ee::read<no_spm>(g_addr);
+	std::uint8_t byte = ee::read(g_addr);
 	++g_addr;
 	return byte;
 }
 
-// One EEPROM byte -> [g_addr++].
+// One EEPROM byte → [g_addr++].
 [[gnu::noinline, gnu::noclone]] void eewr(std::uint8_t byte)
 {
-	ee::write<off, no_spm>(g_addr, byte);
+	ee::write<off>(g_addr, byte);
 	++g_addr;
 }
 
@@ -197,7 +185,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	} while (--g_cnt);
 }
 
-// Wait out a running SPM op, then re-open the RWW section - after every page
+// Wait out a running SPM op, then re-open the RWW section — after every page
 // op and before handing over, as the oracle does.
 [[gnu::noinline, gnu::noclone]] void settle()
 {
@@ -213,12 +201,12 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 	tsb_app();
 }
 
-// Step g_addr one page down and erase that page (the decrement lives here  -
+// Step g_addr one page down and erase that page (the decrement lives here —
 // the global-register rule).
 [[gnu::noinline, gnu::noclone]] void erase_below()
 {
 	g_addr -= page;
-	spm::command<off>(spm::op::erase, g_addr);
+	spm::erase_page<off>(g_addr);
 	settle();
 }
 
@@ -233,7 +221,7 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 }
 
 // Stream one host page into the erased flash page at g_addr (asm 2 of 2): the
-// word pair stages in r0:r1 straight from rx (whose register set is known  -
+// word pair stages in r0:r1 straight from rx (whose register set is known —
 // the cross-call liveness C++ cannot express), Z walks the page and PGWRT
 // programs it. g_addr is left at the next page base.
 [[gnu::noinline, gnu::noclone]] void store_flash()
@@ -268,15 +256,14 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 {
 	// A watchdog reset hands straight back to the application, as the
 	// reference loader does, rather than re-entering the bootloader.
-	if (hw::mcusr::wdrf.test()) {
+	if (hw::mcusr::wdrf.test())
 		appjump();
-	}
 
 	// Lean bring-up from reset state: UCSR0C already reads 8N1, UBRR0H reads
-	// 0, and rx()/tx() raise RXEN0/TXEN0 on first use - only the divisor low
+	// 0, and rx()/tx() raise RXEN0/TXEN0 on first use — only the divisor low
 	// byte and U2X0 need a store. The library still does the datasheet work.
 	static_assert(baud.u2x && baud.ubrr < 256, "lean bring-up writes UBRR0L only, with U2X0");
-	hw::ubrr0::write(static_cast<std::uint8_t>(baud.ubrr));
+	hw::reg<"UBRR0">::write(static_cast<std::uint8_t>(baud.ubrr));
 	hw::ucsr0a::write(hw::ucsr0a::u2x0(1));
 	// General-purpose registers are undefined at power-on (no crt zeroes them);
 	// the direction latch must start "not receiving" so the first rx() enables
@@ -284,15 +271,13 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 	// same reason.
 	g_receiving = 0;
 
-	// Activation: 3x'@', each inside the config page's timeout window (rx
-	// floors it so a corrupt page cannot lock the loader out); anything else  -
-	// including silence - hands over.
+	// Activation: 3×'@', each inside the config page's timeout window (rx
+	// floors it so a corrupt page cannot lock the loader out); anything else —
+	// including silence — hands over.
 	g_window = avr::flash_load(flash_ptr(app_end + 2));
-	for (std::uint8_t k = 3; k; --k) {
-		if (rx() != knock) {
+	for (std::uint8_t k = 3; k; --k)
+		if (rx() != knock)
 			appjump();
-		}
-	}
 	g_window = comm_window;
 
 	// Password gate (config page from app_end+3, 0xff-terminated; a blank
@@ -306,19 +291,17 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 		std::uint8_t expected = avr::flash_load(flash_ptr(g_addr)) & mask;
 		++g_addr;
 		if (expected == 0xff) {
-			g_addr = reinterpret_cast<std::uint16_t>(info.data());
+			g_addr = reinterpret_cast<std::uint16_t>(&info[0]);
 			g_cnt = sizeof(info);
 			sendf();
 			break;
 		}
 		std::uint8_t got = rx();
 		if (got == 0) {
-			if (mask == 0) {
+			if (mask == 0)
 				continue;
-			}
-			if (rcnf() != confirm || rcnf() != confirm) {
+			if (rcnf() != confirm || rcnf() != confirm)
 				appjump();
-			}
 			erase_application(); // leaves g_addr = 0 for the EEPROM walk
 			do {
 				eewr(0xff);
@@ -327,9 +310,8 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 			erase_below();
 			break;
 		}
-		if (got != expected) {
+		if (got != expected)
 			mask = 0;
-		}
 	}
 
 	for (;;) {
@@ -338,27 +320,23 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 		switch (rx()) {
 		case 'f': // read application flash, one page per host '!'
 			for (;;) {
-				if (rx() != confirm) {
+				if (rx() != confirm)
 					break;
-				}
 				g_cnt = page;
 				sendf();
-				if (g_addr >= app_end) {
+				if (g_addr >= app_end)
 					break;
-				}
 			}
 			break;
 		case 'F':                // erase the application, then take pages behind '?'
 			erase_application(); // leaves g_addr = 0, the write start
-			while (rcnf() == confirm) {
+			while (rcnf() == confirm)
 				store_flash();
-			}
 			break;
 		case 'e': // read EEPROM, one page per host '!', until the host stops
 			for (;;) {
-				if (rx() != confirm) {
+				if (rx() != confirm)
 					break;
-				}
 				g_cnt = page;
 				do {
 					tx(eerd());
@@ -380,9 +358,8 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 			sendf();
 			break;
 		case 'C': // replace the config page, then echo it back to verify
-			if (rcnf() != confirm) {
+			if (rcnf() != confirm)
 				break;
-			}
 			g_addr = app_end + page;
 			erase_below(); // leaves g_addr = app_end, the store target
 			store_flash();
@@ -396,6 +373,14 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 } // namespace
 } // namespace tsb
 
-// Reset lands at the boot section base (BOOTRST): the entry stub in .vectors
-// is laid first and does the one line of crt a crt-less image needs.
-template struct avr::startup::entry<tsb::run, avr::startup::stack::hardware>;
+// Reset lands here: BOOTRST vectors to the boot section base and .vectors is
+// laid first, so this is the first instruction executed. No crt ran, so set
+// the stack pointer before anything is called.
+extern "C" [[gnu::naked, gnu::used, gnu::section(".vectors")]] void __boot_entry()
+{
+	SP = RAMEND;
+	// The one line of crt this loader needs: compiled code assumes
+	// __zero_reg__ (r1) is 0, and power-on registers are undefined.
+	asm volatile("clr __zero_reg__");
+	tsb::run();
+}

tsb/tsb_policy.cpp

@@ -1,329 +0,0 @@
-// TinySafeBoot on libavr - the policy floor: pureboot's rules, measured.
-//
-// The full TinySafeBoot feature set - watchdog bail, one-wire half-duplex,
-// config-page activation timeout, password gate, emergency erase, and
-// config/flash/EEPROM read-write - under philosophy #5 exactly as pureboot
-// obeys it: no assembly, no register variables; code, attributes, and flags
-// only. Every lesson pureboot's development produced is applied - the
-// library's half-duplex serial and startup entry, lean bring-up from reset
-// state, one merged send loop over both memories, oracle-shaped loop bounds,
-// locals threaded through noinline primitives, pureboot's codegen flags  -
-// and the result sits below the idiomatic tier and above the 512 B boot
-// section the tricks/asm tiers reach with the banned mechanisms
-// (oracle/README.md holds all four). This tier exists to keep that gap an
-// artifact
-// rather than a claim: the gap to 512 is the rent of policy-clean C++  -
-// helpers that hold a cursor across rx()/tx() pay push/pop and argument
-// threading where a global-register protocol pays nothing, and both
-// control-flow merges tried (a parametrized paged session, a merged store
-// loop) measured larger than the split cases they replaced. TSB's wire fixes
-// the per-command loop shapes on the device, so pureboot 5's one-transfer-
-// loop collapse has no purchase here.
-//
-// The wire protocol is strict request/response, which is what makes the
-// shared line safe: the device drives it only between a received command and
-// its reply, and releases it (the library's half-duplex choreography)
-// whenever it waits.
-
-#include <libavr/libavr.hpp>
-
-using namespace avr::literals;
-namespace spm = avr::spm;
-namespace ee = avr::eeprom;
-
-using dev = avr::device<{.clock = 16_MHz}>;
-// One-wire: RX and TX share the line, exactly as the native-UART TSB expects.
-// 115200 at 16 MHz lands +2.1 % off, past the receiver-tolerance table the
-// solver holds rates to - the oracle's own deployment has run there for a
-// decade, so the override states that it is meant.
-using serial_t = dev::uart0<{
-    .baud = 115200_Bd,
-    .allow_baud_error = true,
-    .half_duplex = true,
-}>;
-inline constexpr serial_t serial{};
-
-namespace tsb {
-namespace {
-
-// The loader is purely polled - it never enables interrupts - so every SPM and
-// EEPROM lock folds to nothing under this posture.
-constexpr auto off = avr::irq::guard_policy::unused;
-
-// Strict request/response: every SPM operation is waited out before the next
-// byte moves, so no flash operation is ever in flight at an EEPROM access -
-// the write procedure's step 2 has nothing to guard, the omission the
-// datasheet grants (DS40002061B section 8.6.3).
-constexpr auto no_spm = ee::spm_interlock::omitted;
-
-// The handshake bytes, identical across every TSB host.
-constexpr std::uint8_t confirm = '!';
-constexpr std::uint8_t request = '?';
-constexpr std::uint8_t knock = '@';
-
-// Boot geometry for the 1 KB boot section (BOOTSZ=10); the page size and the
-// flash/EEPROM extents are the chip database's to know. app_end is the config
-// page (TSB's LASTPAGE), one page below the boot section.
-constexpr std::uint16_t page = spm::page_bytes;
-constexpr std::uint16_t boot_bytes = 1024;
-constexpr std::uint16_t app_end = spm::flash_bytes - boot_bytes - page;
-constexpr std::uint16_t eeprom_end = avr::hw::db.mem.eeprom_size - 1;
-
-// Lockout-proof floor for the activation window: the oracle's F_CPU/1MHz, so
-// it follows the clock rather than restating it (rule 41).
-constexpr auto act_min = static_cast<std::uint8_t>(dev::clock.hz / 1'000'000);
-// Post-activation window: the host gets seconds, not milliseconds, mid-session.
-constexpr std::uint8_t comm_window = 200;
-
-// Firmware version stamp: YY*512 + MM*32 + DD, the encoding the host decodes.
-constexpr std::uint16_t build_date = 26 * 512 + 7 * 32 + 27;
-
-// The 16-byte device-info block, streamed out on activation.
-// clang-format off
-[[gnu::progmem]] constexpr auto info = std::to_array<std::uint8_t>({
-    'T', 'S', 'B',
-    build_date & 0xFF, build_date >> 8,
-    0xF3,             // status: native-UART fixed-baud lineage
-    avr::hw::db.signature[0], avr::hw::db.signature[1], avr::hw::db.signature[2],
-    page / 2,         // page size in words
-    (app_end / 2) & 0xFF, (app_end / 2) >> 8, // app-flash boundary, words
-    eeprom_end & 0xFF, eeprom_end >> 8,
-    0xAA, 0xAA,       // ATmega processor-type marker (bytes 14 == 15)
-});
-// clang-format on
-
-// The receive window, pre-floored where it is set. In .noinit: there is no
-// crt to clear a .bss image, and run() stores it before the first receive.
-[[gnu::section(".noinit")]] std::uint8_t window;
-
-const std::uint8_t *flash_ptr(std::uint16_t addr)
-{
-	return reinterpret_cast<const std::uint8_t *>(addr);
-}
-
-// Bounded byte receive: poll under nested countdowns, 0 on silence. The 0
-// then falls through every compare - not a knock, not a confirm, not a
-// command - so a silent host unwinds the loader to the application from
-// anywhere, and a mid-session cable pull cannot wedge it. The line release on
-// a direction change is the serial backend's.
-[[gnu::noinline]] std::uint8_t rx()
-{
-	std::uint16_t outer = static_cast<std::uint16_t>(window) << 8;
-	do {
-		std::uint8_t fine = 0;
-		do {
-			if (auto byte = serial.read()) {
-				return *byte;
-			}
-		} while (--fine);
-	} while (--outer);
-	return 0;
-}
-
-// One-wire transmit: the backend takes the line with a turn-around guard and
-// holds it until the whole frame is out.
-[[gnu::noinline]] void tx(std::uint8_t byte)
-{
-	serial.write(byte);
-}
-
-// '?', then hand back the host's reply for the callers' one-byte compare.
-[[gnu::noinline]] std::uint8_t rcnf()
-{
-	tx(request);
-	return rx();
-}
-
-// The one send loop: the info block, the config page, application flash and
-// EEPROM pages all stream through here.
-[[gnu::noinline]] void send_block(bool eep, std::uint16_t at, std::uint8_t count)
-{
-	do {
-		tx(eep ? ee::read<no_spm>(at) : avr::flash_load(flash_ptr(at)));
-		++at;
-	} while (--count);
-}
-
-// One EEPROM byte in - shared by the emergency wipe and the 'E' stream.
-[[gnu::noinline]] void eeput(std::uint16_t at, std::uint8_t value)
-{
-	ee::write<off, no_spm>(at, value);
-}
-
-// Wait out a running SPM op, then re-open the RWW section - after every page
-// op and before handing over, as the oracle does.
-[[gnu::noinline]] void settle()
-{
-	spm::wait();
-	spm::rww_enable<off>();
-}
-
-// One host page straight into the erased flash page at `at` - through the SPM
-// word buffer (low byte then high), no SRAM staging - then committed. `at`
-// names a page base, so the cursor's low byte reaching the boundary ends the
-// walk.
-[[gnu::noinline]] void store_flash_page(std::uint16_t at)
-{
-	const auto open = spm::page::begin<spm::from::boot_section, off>(at);
-	do {
-		std::uint8_t low = rx();
-		std::uint8_t high = rx();
-		spm::fill<off>(open, at, std::bit_cast<std::uint16_t>(std::array{low, high}));
-		at += 2;
-	} while (static_cast<std::uint8_t>(at) & (page - 1));
-	spm::command<off>(spm::op::write, at - page);
-	settle();
-}
-
-extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --defsym=tsb_app=0
-
-[[noreturn]] void appjump()
-{
-	settle();
-	tsb_app();
-}
-
-// Step one page down and erase it - the erase shared by the whole-app walk,
-// the config rewrite and the emergency wipe; hands the stepped address back.
-[[gnu::noinline]] std::uint16_t erase_below(std::uint16_t at)
-{
-	at -= page;
-	spm::command<off>(spm::op::erase, at);
-	settle();
-	return at;
-}
-
-// Erase the whole application, top-down like the oracle: the loop bound is a
-// compare with zero, and the returned 0 is the address every caller wants
-// next.
-[[gnu::noinline]] std::uint16_t erase_application()
-{
-	std::uint16_t at = app_end;
-	do {
-		at = erase_below(at);
-	} while (at != 0);
-	return at;
-}
-
-[[noreturn]] void run()
-{
-	// A watchdog reset hands straight back to the application, as the
-	// reference loader does, rather than re-entering the bootloader.
-	if (avr::hw::mcusr::wdrf.test()) {
-		appjump();
-	}
-
-	// Lean bring-up from reset state: UCSR0C already reads 8N1, UBRR0H reads
-	// 0, and the half-duplex write()/read() raise TXEN0/RXEN0 on first use  -
-	// only the divisor low byte and U2X0 need a store. The solver still does
-	// the datasheet work; the asserts pin the reset-state assumptions.
-	{
-		constexpr auto sol = avr::uart::solve_baud(dev::clock, 115200_Bd, 8, avr::uart::parity::none);
-		static_assert(sol.u2x && sol.ubrr < 256, "lean bring-up writes UBRR0L only, with U2X0");
-		avr::hw::ubrr0::write(static_cast<std::uint8_t>(sol.ubrr));
-		avr::hw::ucsr0a::write(avr::hw::ucsr0a::u2x0(1));
-	}
-
-	// Activation: 3x'@', each inside the config page's timeout window
-	// (floored so a corrupt page cannot lock the loader out); anything else  -
-	// including silence - hands over.
-	window = avr::flash_load(flash_ptr(app_end + 2)) | act_min;
-	for (std::uint8_t k = 3; k; --k) {
-		if (rx() != knock) {
-			appjump();
-		}
-	}
-	window = comm_window;
-
-	// Password gate (config page from app_end+3, 0xff-terminated; a blank
-	// page is no password). A wrong byte blanks the comparison and drains the
-	// line forever, so a wrong password can never fall through; a 0 requests
-	// emergency erase behind two confirms. On pass the info block goes out;
-	// the emergency path skips it and drops into the command loop.
-	std::uint16_t at = app_end + 3;
-	std::uint8_t mask = 0xff;
-	for (;;) {
-		std::uint8_t expected = avr::flash_load(flash_ptr(at)) & mask;
-		++at;
-		if (expected == 0xff) {
-			send_block(false, reinterpret_cast<std::uint16_t>(info.data()), info.size());
-			break;
-		}
-		std::uint8_t got = rx();
-		if (got == 0) {
-			if (mask == 0) {
-				continue;
-			}
-			if (rcnf() != confirm || rcnf() != confirm) {
-				appjump();
-			}
-			std::uint16_t a = erase_application();
-			do {
-				eeput(a, 0xff);
-			} while (++a <= eeprom_end);
-			erase_below(app_end + page);
-			break;
-		}
-		if (got != expected) {
-			mask = 0;
-		}
-	}
-
-	for (;;) {
-		tx(confirm); // Mainloop ready
-		const std::uint8_t command = rx();
-		switch (command) {
-		case 'f': // read application flash, one page per host '!'
-			for (std::uint16_t a = 0; a < app_end; a += page) {
-				if (rx() != confirm) {
-					break;
-				}
-				send_block(false, a, page);
-			}
-			break;
-		case 'e': // read EEPROM, one page per host '!', until the host stops
-			for (std::uint16_t a = 0;; a += page) {
-				if (rx() != confirm) {
-					break;
-				}
-				send_block(true, a, page);
-			}
-			break;
-		case 'F': { // erase the application, then take pages behind '?'
-			std::uint16_t a = erase_application();
-			for (; rcnf() == confirm; a += page) {
-				store_flash_page(a);
-			}
-			break;
-		}
-		case 'E': // take EEPROM pages behind '?', each write host-paced
-			for (std::uint16_t a = 0; rcnf() == confirm;) {
-				std::uint8_t count = page;
-				do {
-					eeput(a, rx());
-					++a;
-				} while (--count);
-			}
-			break;
-		case 'c': // read the config page
-		read_config:
-			send_block(false, app_end, page);
-			break;
-		case 'C': // replace the config page, then echo it back to verify
-			if (rcnf() != confirm) {
-				break;
-			}
-			store_flash_page(erase_below(app_end + page));
-			goto read_config;
-		default: // 'q' or any other byte runs the application
-			appjump();
-		}
-	}
-}
-
-} // namespace
-} // namespace tsb
-
-// Reset lands at the boot section base (BOOTRST): the entry stub in .vectors
-// is laid first and does the one line of crt a crt-less image needs.
-template struct avr::startup::entry<tsb::run, avr::startup::stack::hardware>;

tsb/tsb_pure.cpp

@@ -1,10 +1,10 @@
-// TinySafeBoot on libavr - tier 1: pure, idiomatic C++.
+// TinySafeBoot on libavr — tier 1: pure, idiomatic C++.
 //
 // A serial flash bootloader for the ATmega328P boot section, reimplementing the
 // TinySafeBoot native-UART fixed-baud protocol on libavr with the full feature
 // set of the hand-written oracle: a watchdog-reset bail, one-wire half-duplex,
 // a config-page activation timeout, the password gate, emergency erase, and
-// config/flash/EEPROM read-write. This variant is written for clarity  -
+// config/flash/EEPROM read-write. This variant is written for clarity —
 // well-factored functions, no compiler-specific size hacks, no inline assembly.
 // The one-wire wiring, the flash-resident info block and every SPM/EEPROM lock
 // are libavr's to handle; the only attribute is the naked reset entry that
@@ -20,29 +20,16 @@ namespace ee = avr::eeprom;
 
 using dev = avr::device<{.clock = 16_MHz}>;
 // One-wire: RX and TX share the line, exactly as the native-UART TSB expects.
-// 115200 at 16 MHz lands +2.1 % off, past the receiver-tolerance table the
-// solver holds rates to - the oracle's own deployment has run there for a
-// decade, so the override states that it is meant.
-using serial_t = dev::uart0<{
-    .baud = 115200_Bd,
-    .allow_baud_error = true,
-    .half_duplex = true,
-}>;
+using serial_t = dev::uart0<{.baud = 115200_Bd, .max_baud_error = 3_pct, .half_duplex = true}>;
 inline constexpr serial_t serial{};
 
 namespace tsb {
 namespace {
 
-// The loader is purely polled - it never enables interrupts - so every SPM and
+// The loader is purely polled — it never enables interrupts — so every SPM and
 // EEPROM lock folds to nothing under this posture.
 constexpr auto off = avr::irq::guard_policy::unused;
 
-// Strict request/response: every SPM operation is waited out before the next
-// byte moves, so no flash operation is ever in flight at an EEPROM access -
-// the write procedure's step 2 has nothing to guard, the omission the
-// datasheet grants (DS40002061B section 8.6.3).
-constexpr auto no_spm = ee::spm_interlock::omitted;
-
 // The handshake bytes, identical across every TSB host.
 constexpr std::uint8_t confirm = '!';
 constexpr std::uint8_t request = '?';
@@ -63,48 +50,26 @@ constexpr std::uint16_t build_date = 26 * 512 + 7 * 32 + 20;
 // The 16-byte device-info block the host reads on activation. A flash_table
 // keeps it in progmem with no .data image (there is no crt to copy one).
 // clang-format off
-inline constexpr auto info_data = std::to_array<std::uint8_t>({
+inline constexpr std::array<std::uint8_t, 16> info_data = {
     'T', 'S', 'B',
     build_date & 0xFF, build_date >> 8,
     0xF3,             // status byte (native-UART fixed-baud lineage)
-    avr::hw::db.signature[0], avr::hw::db.signature[1], avr::hw::db.signature[2],
+    0x1E, 0x95, 0x0F, // ATmega328P signature
     page / 2,         // page size in words
     (app_end / 2) & 0xFF, (app_end / 2) >> 8, // app-flash boundary, words
     eeprom_end & 0xFF, eeprom_end >> 8,
     0xAA, 0xAA, // ATmega processor-type marker (bytes 14 == 15)
-});
+};
 // clang-format on
 using info = avr::flash_table<info_data>;
 
-// The lockout-proof floor for the receive window: the oracle's F_CPU/1MHz, so
-// it follows the clock rather than restating it.
-constexpr auto act_min = static_cast<std::uint8_t>(dev::clock.hz / 1'000'000);
-
-// The receive window, pre-floored where it is set. In .noinit: there is no crt
-// to clear a .bss image, and run() stores it before the first receive.
-[[gnu::section(".noinit")]] std::uint8_t window;
-
-// Bounded byte read over the one-wire line - read() releases the line to the
-// receiver - answering 0 on silence. That 0 falls through every compare below:
-// not a knock, not a confirm, not a command, so a silent host unwinds the
-// loader to the application from anywhere and a mid-session cable pull cannot
-// wedge it. The oracle lists that timeout among its own fixes, and a blocking
-// read is how a tier loses it.
+// Blocking byte read/write over the one-wire line: read() releases the line to
+// the receiver, write() takes it and holds it until the frame is out.
 std::uint8_t rx()
 {
-	std::uint16_t outer = static_cast<std::uint16_t>(window) << 8;
-	do {
-		std::uint8_t fine = 0;
-		do {
-			if (auto byte = serial.read()) {
-				return *byte;
-			}
-		} while (--fine);
-	} while (--outer);
-	return 0;
+	return serial.read_blocking();
 }
 
-// write() takes the line and holds it until the frame is out.
 void tx(std::uint8_t byte)
 {
 	serial.write(byte);
@@ -118,16 +83,14 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 // Stream `count` bytes to the host, from flash (LPM) or from EEPROM.
 void send_flash(std::uint16_t addr, std::uint8_t count)
 {
-	while (count--) {
+	while (count--)
 		tx(avr::flash_load(flash_ptr(addr++)));
-	}
 }
 
 void send_eeprom(std::uint16_t addr, std::uint8_t count)
 {
-	while (count--) {
-		tx(ee::read<no_spm>(addr++));
-	}
+	while (count--)
+		tx(ee::read(addr++));
 }
 
 // Prompt the host with '?' and report whether it answered '!'.
@@ -138,33 +101,32 @@ bool request_confirm()
 }
 
 // Stream one page from the host straight into the already-erased flash page at
-// `addr`, filling the SPM word buffer low byte then high - no SRAM staging, so
+// `addr`, filling the SPM word buffer low byte then high — no SRAM staging, so
 // receiving and programming are the same loop.
 void store_flash_page(std::uint16_t addr)
 {
-	const auto open = spm::page::begin<spm::from::boot_section, off>(addr);
 	for (std::uint16_t i = 0; i < page; i += 2) {
 		std::uint8_t lo = rx();
 		std::uint8_t hi = rx();
-		spm::fill<off>(open, addr + i, static_cast<std::uint16_t>(lo | (hi << 8)));
+		spm::fill<off>(addr + i, static_cast<std::uint16_t>(lo | (hi << 8)));
 	}
-	spm::write_page<spm::from::boot_section, off>(addr); // blocking: waits the write out
+	spm::write_page<off>(addr);
+	spm::wait();
 }
 
 // Stream one page from the host straight into EEPROM, byte by byte.
 void store_eeprom_page(std::uint16_t addr)
 {
-	for (std::uint16_t i = 0; i < page; ++i) {
-		ee::write<off, no_spm>(addr + i, rx());
-	}
+	for (std::uint16_t i = 0; i < page; ++i)
+		ee::write<off>(addr + i, rx());
 }
 
-// Erase one flash page, waited out by the blocking spelling - the erase step
-// shared by the whole-app erase, the config-page rewrite and the emergency
-// wipe.
+// Erase one flash page and wait it out — the erase step shared by the whole-app
+// erase, the config-page rewrite and the emergency wipe.
 void erase_page(std::uint16_t addr)
 {
-	spm::erase_page<spm::from::boot_section, off>(addr);
+	spm::erase_page<off>(addr);
+	spm::wait();
 }
 
 // Erase the whole application, one page at a time, top-down as the reference
@@ -195,9 +157,8 @@ extern "C" [[noreturn]] void tsb_app();
 void read_flash()
 {
 	for (std::uint16_t a = 0; a < app_end; a += page) {
-		if (rx() != confirm) {
+		if (rx() != confirm)
 			return;
-		}
 		send_flash(a, page);
 	}
 }
@@ -206,9 +167,8 @@ void read_flash()
 void read_eeprom()
 {
 	for (std::uint16_t a = 0;; a += page) {
-		if (rx() != confirm) {
+		if (rx() != confirm)
 			return;
-		}
 		send_eeprom(a, page);
 	}
 }
@@ -218,25 +178,22 @@ void read_eeprom()
 void write_flash()
 {
 	erase_application();
-	for (std::uint16_t a = 0; request_confirm(); a += page) {
+	for (std::uint16_t a = 0; request_confirm(); a += page)
 		store_flash_page(a);
-	}
 }
 
 // 'E': take pages the host offers behind '?' into EEPROM.
 void write_eeprom()
 {
-	for (std::uint16_t a = 0; request_confirm(); a += page) {
+	for (std::uint16_t a = 0; request_confirm(); a += page)
 		store_eeprom_page(a);
-	}
 }
 
 // 'C': replace the config page, then echo it back for the host to verify.
 void write_config()
 {
-	if (!request_confirm()) {
+	if (!request_confirm())
 		return;
-	}
 	erase_page(app_end);
 	store_flash_page(app_end);
 	spm::rww_enable<off>();
@@ -249,9 +206,8 @@ void write_config()
 void emergency_erase()
 {
 	erase_application();
-	for (std::uint16_t a = 0; a <= eeprom_end; ++a) {
-		ee::write<off, no_spm>(a, 0xff);
-	}
+	for (std::uint16_t a = 0; a <= eeprom_end; ++a)
+		ee::write<off>(a, 0xff);
 	erase_page(app_end);
 	spm::rww_enable<off>();
 }
@@ -266,18 +222,14 @@ gate password_gate()
 {
 	for (const std::uint8_t *pw = flash_ptr(app_end + 3);; ++pw) {
 		std::uint8_t expected = avr::flash_load(pw);
-		if (expected == 0xff) {
+		if (expected == 0xff)
 			return gate::pass;
-		}
 		std::uint8_t got = rx();
-		if (got == 0) {
+		if (got == 0)
 			return gate::emergency;
-		}
-		if (got != expected) {
-			for (;;) {
+		if (got != expected)
+			for (;;)
 				rx();
-			}
-		}
 	}
 }
 
@@ -285,25 +237,21 @@ gate password_gate()
 {
 	// A watchdog reset hands straight back to the application, as the reference
 	// loader does, rather than re-entering the bootloader.
-	if (avr::hw::mcusr::wdrf.test()) {
+	if (avr::hw::mcusr::wdrf.test())
 		appjump();
-	}
 
 	avr::init<serial_t>();
 
 	// Activation: the host knocks three '@' inside a window whose length is the
-	// config page's timeout byte, floored so a corrupt page can never lock the
-	// loader out. An idle port times out and boots the application; the same
-	// window then bounds every receive of the session.
-	window = avr::flash_load(flash_ptr(app_end + 2)) | act_min;
-	__uint24 idle = static_cast<__uint24>(window) << 16;
+	// config page's timeout byte (floored so a corrupt page can never lock the
+	// loader out). An idle port times out and boots the application.
+	__uint24 idle = static_cast<__uint24>(avr::flash_load(flash_ptr(app_end + 2)) | 16) << 16;
 	std::uint8_t knocks = 0;
 	while (knocks < 3) {
-		if (auto byte = serial.read()) {
+		if (auto byte = serial.read())
 			knocks = *byte == knock ? knocks + 1 : 0;
-		} else if (--idle == 0) {
+		else if (--idle == 0)
 			appjump();
-		}
 	}
 
 	switch (password_gate()) {
@@ -311,9 +259,8 @@ gate password_gate()
 		send_flash(reinterpret_cast<std::uint16_t>(info::storage.data()), info::size());
 		break;
 	case gate::emergency:
-		if (!request_confirm() || !request_confirm()) {
+		if (!request_confirm() || !request_confirm())
 			appjump();
-		}
 		emergency_erase();
 		break;
 	}
@@ -348,6 +295,14 @@ gate password_gate()
 } // namespace
 } // namespace tsb
 
-// Reset lands at the boot section base (BOOTRST): the entry stub in .vectors
-// is laid first and does the one line of crt a crt-less image needs.
-template struct avr::startup::entry<tsb::run, avr::startup::stack::hardware>;
+// Reset lands here: BOOTRST vectors to the boot section base and .vectors is
+// laid first, so this is the first instruction executed. No crt ran, so set the
+// stack pointer before anything is called.
+extern "C" [[gnu::naked, gnu::used, gnu::section(".vectors")]] void __boot_entry()
+{
+	SP = RAMEND;
+	// The one line of crt this loader needs: compiled code assumes
+	// __zero_reg__ (r1) is 0, and power-on registers are undefined.
+	asm volatile("clr __zero_reg__");
+	tsb::run();
+}

tsb/tsb_tricks.cpp

@@ -1,16 +1,16 @@
-// TinySafeBoot on libavr - tier 2: C++ with compiler trickery, no assembly.
+// TinySafeBoot on libavr — tier 2: C++ with compiler trickery, no assembly.
 //
-// The full TinySafeBoot feature set - watchdog bail, one-wire half-duplex,
+// The full TinySafeBoot feature set — watchdog bail, one-wire half-duplex,
 // config-page activation timeout, password gate, emergency erase, and
-// config/flash/EEPROM read-write - in pure C++, a little over the 512-byte boot
-// section the hand-written oracle fits (oracle/README.md holds what each tier
-// measures). The structure mirrors the oracle's: a handful of tiny noinline
+// config/flash/EEPROM read-write — in pure C++, 526 bytes: 14 over the 512-byte
+// boot section the hand-written oracle fits, from 168 over at this tier's first
+// floor. The structure mirrors the oracle's: a handful of tiny noinline
 // primitives sharing one whole-loader register allocation, expressed as global
 // register variables so no helper ever saves, spills, or reloads any of it.
 //
 // The register protocol (all call-saved, so calls preserve them by ABI):
-//   Y (r28:r29)  g_addr      the walked flash/EEPROM address - adiw-able
-//   r16          g_cnt       byte countdown of the running block - ldi-able
+//   Y (r28:r29)  g_addr      the walked flash/EEPROM address — adiw-able
+//   r16          g_cnt       byte countdown of the running block — ldi-able
 //   r7           g_window    rx timeout, roughly 30 ms units at 16 MHz
 //   r6           g_receiving one-wire direction latch, cleared at bring-up
 //                            (power-on registers are undefined)
@@ -18,10 +18,10 @@
 // GCC 16.1 miscompiles stores into global register variables: an update whose
 // remaining uses all hide inside callees is deleted whenever a CALL follows it
 // before any jump/ret (the backend's liveness walk lumps fixed registers with
-// call-clobbered ones - minimal repro in libavr's
-// test/upstream/gcc-avr-globalreg-repro.cpp). Every
+// call-clobbered ones — minimal repro in libavr's
+// local/scratch/probes/gcc-avr-globalreg-repro.cpp, lessons.md entry). Every
 // g_* update below therefore sits where a *local* read or a jump/ret follows
-// it - the helpers advance g_addr immediately before returning, and rx()
+// it — the helpers advance g_addr immediately before returning, and rx()
 // re-floors the window on every call instead of storing the floored value
 // once. The layout is load-bearing; do not "simplify" it.
 //
@@ -41,16 +41,10 @@ namespace hw = avr::hw;
 namespace tsb {
 namespace {
 
-// The loader is purely polled - it never enables interrupts - so every SPM and
+// The loader is purely polled — it never enables interrupts — so every SPM and
 // EEPROM lock folds to nothing under this posture.
 constexpr auto off = avr::irq::guard_policy::unused;
 
-// Strict request/response: every SPM operation is waited out before the next
-// byte moves, so no flash operation is ever in flight at an EEPROM access -
-// the write procedure's step 2 has nothing to guard, the omission the
-// datasheet grants (DS40002061B section 8.6.3).
-constexpr auto no_spm = ee::spm_interlock::omitted;
-
 constexpr std::uint8_t confirm = '!';
 constexpr std::uint8_t request = '?';
 constexpr std::uint8_t knock = '@';
@@ -63,34 +57,28 @@ constexpr std::uint16_t boot_bytes = 1024;
 constexpr std::uint16_t app_end = spm::flash_bytes - boot_bytes - page;
 constexpr std::uint16_t eeprom_end = avr::hw::db.mem.eeprom_size - 1;
 
-// Lockout-proof floor for the activation window: the oracle's F_CPU/1MHz, so
-// it follows the clock rather than restating it (rule 41).
-constexpr auto act_min = static_cast<std::uint8_t>((16_MHz).hz / 1'000'000);
+// Lockout-proof floor for the activation window (the oracle's F_CPU/1MHz).
+constexpr std::uint8_t act_min = 16;
 // Post-activation window: the host gets seconds, not milliseconds, mid-session.
 constexpr std::uint8_t comm_window = 200;
 
 constexpr std::uint16_t build_date = 26 * 512 + 7 * 32 + 20;
 
 // Fixed 115200 8N1; the library solves UBRR + U2X from clock and baud.
-constexpr auto baud = avr::uart::solve_baud(16_MHz, 115200_Bd, 8, avr::uart::parity::none);
-
-// One bit time on the wire: the turn-around a shared-line peer needs to stop
-// driving before this one starts. Derived from the solved rate, so it follows
-// the link rather than a count measured against one.
-constexpr auto guard_cycles = static_cast<std::uint32_t>((16_MHz).hz / baud.actual);
+constexpr auto baud = avr::uart::detail::solve_baud(16_MHz, 115200_Bd);
 
 // The 16-byte device-info block, streamed out on activation.
 // clang-format off
-[[gnu::progmem]] constexpr auto info = std::to_array<std::uint8_t>({
+[[gnu::progmem]] constexpr std::uint8_t info[16] = {
     'T', 'S', 'B',
     build_date & 0xFF, build_date >> 8,
     0xF3,             // status: native-UART fixed-baud lineage
-    avr::hw::db.signature[0], avr::hw::db.signature[1], avr::hw::db.signature[2],
+    0x1E, 0x95, 0x0F, // ATmega328P signature
     page / 2,         // page size in words
     (app_end / 2) & 0xFF, (app_end / 2) >> 8,
     eeprom_end & 0xFF, eeprom_end >> 8,
     0xAA, 0xAA,
-});
+};
 // clang-format on
 
 register std::uint16_t g_addr asm("r28");
@@ -105,8 +93,8 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 
 // Bounded byte receive, the oracle's shape: release the one-wire line on a
 // direction change, poll RXC0 under nested countdowns, 0 on silence. The 0
-// then falls through every compare - not a knock, not a confirm, not a
-// command - so a silent host unwinds the loader to the application from
+// then falls through every compare — not a knock, not a confirm, not a
+// command — so a silent host unwinds the loader to the application from
 // anywhere, and a mid-session cable pull cannot wedge it.
 [[gnu::noinline, gnu::noclone]] std::uint8_t rx()
 {
@@ -114,25 +102,24 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 		g_receiving = 1;
 		hw::ucsr0b::write(hw::ucsr0b::rxen0(1)); // RXEN0 alone: release and listen
 	}
-	// act_min ORs in here, per call, not once into g_window at setup - the
+	// act_min ORs in here, per call, not once into g_window at setup — the
 	// one placement the global-register-store miscompile cannot delete.
 	std::uint16_t outer = static_cast<std::uint16_t>(g_window | act_min) << 8;
 	do {
 		std::uint8_t fine = 0;
 		do {
 			auto status = hw::ucsr0a::read();
-			if (status & hw::ucsr0a::rxc0(1).value) {
+			if (status & hw::ucsr0a::rxc0(1).value)
 				return hw::udr0::read();
-			}
 		} while (--fine);
 	} while (--outer);
 	return 0;
 }
 
-// One-wire transmit: take the line (TXEN0 alone - the receiver must be off
+// One-wire transmit: take the line (TXEN0 alone — the receiver must be off
 // while driving) on a direction change, with a turn-around guard so a shorted
 // peer can switch first; then hold the line until the whole frame is out
-// (TXC0, not UDRE0 - the stop bit must be on the wire before a caller may
+// (TXC0, not UDRE0 — the stop bit must be on the wire before a caller may
 // release the line), and W1C TXC0 by storing the sampled status back, which
 // keeps U2X0.
 [[gnu::noinline, gnu::noclone]] void tx(std::uint8_t byte)
@@ -140,7 +127,8 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	if (g_receiving) {
 		g_receiving = 0;
 		hw::ucsr0b::write(hw::ucsr0b::txen0(1));
-		avr::delay::cycles<guard_cycles>();
+		for (std::uint8_t guard = 46; guard; --guard)
+			;
 	}
 	hw::udr0::write(byte);
 	std::uint8_t status;
@@ -157,7 +145,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	return rx();
 }
 
-// One flash byte <- [g_addr++] (the advance right before ret - see header).
+// One flash byte ← [g_addr++] (the advance right before ret — see header).
 [[gnu::noinline, gnu::noclone]] std::uint8_t sflash()
 {
 	std::uint8_t byte = avr::flash_load(flash_ptr(g_addr));
@@ -165,18 +153,18 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	return byte;
 }
 
-// One EEPROM byte <- [g_addr++].
+// One EEPROM byte ← [g_addr++].
 [[gnu::noinline, gnu::noclone]] std::uint8_t eerd()
 {
-	std::uint8_t byte = ee::read<no_spm>(g_addr);
+	std::uint8_t byte = ee::read(g_addr);
 	++g_addr;
 	return byte;
 }
 
-// One EEPROM byte -> [g_addr++].
+// One EEPROM byte → [g_addr++].
 [[gnu::noinline, gnu::noclone]] void eewr(std::uint8_t byte)
 {
-	ee::write<off, no_spm>(g_addr, byte);
+	ee::write<off>(g_addr, byte);
 	++g_addr;
 }
 
@@ -188,7 +176,7 @@ const std::uint8_t *flash_ptr(std::uint16_t addr)
 	} while (--g_cnt);
 }
 
-// Wait out a running SPM op, then re-open the RWW section - after every page
+// Wait out a running SPM op, then re-open the RWW section — after every page
 // op and before handing over, as the oracle does.
 [[gnu::noinline, gnu::noclone]] void settle()
 {
@@ -210,12 +198,12 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 [[gnu::noinline, gnu::noclone]] void erase_below()
 {
 	g_addr -= page;
-	spm::command<off>(spm::op::erase, g_addr);
+	spm::erase_page<off>(g_addr);
 	settle();
 }
 
 // Erase the whole application, top-down like the oracle: the loop bound is a
-// compare with zero, and g_addr = 0 - the value every caller wants next - is
+// compare with zero, and g_addr = 0 — the value every caller wants next — is
 // handed back for free.
 [[gnu::noinline, gnu::noclone]] void erase_application()
 {
@@ -226,19 +214,18 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 }
 
 // Stream one host page into the erased flash page at g_addr (SPM word buffer,
-// low byte then high) - no SRAM staging, receive and program are one loop.
+// low byte then high) — no SRAM staging, receive and program are one loop.
 // g_addr is left at the next page base.
 [[gnu::noinline, gnu::noclone]] void store_flash()
 {
-	const auto open = spm::page::begin<spm::from::boot_section, off>(g_addr);
 	g_cnt = page / 2;
 	do {
 		std::uint16_t word = rx();
 		word |= static_cast<std::uint16_t>(rx()) << 8;
-		spm::fill<off>(open, g_addr, word);
+		spm::fill<off>(g_addr, word);
 		g_addr += 2;
 	} while (--g_cnt);
-	spm::command<off>(spm::op::write, g_addr - page);
+	spm::write_page<off>(g_addr - page);
 	settle();
 }
 
@@ -246,15 +233,14 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 {
 	// A watchdog reset hands straight back to the application, as the
 	// reference loader does, rather than re-entering the bootloader.
-	if (hw::mcusr::wdrf.test()) {
+	if (hw::mcusr::wdrf.test())
 		appjump();
-	}
 
 	// Lean bring-up from reset state: UCSR0C already reads 8N1, UBRR0H reads
-	// 0, and rx()/tx() raise RXEN0/TXEN0 on first use - only the divisor low
+	// 0, and rx()/tx() raise RXEN0/TXEN0 on first use — only the divisor low
 	// byte and U2X0 need a store. The library still does the datasheet work.
 	static_assert(baud.u2x && baud.ubrr < 256, "lean bring-up writes UBRR0L only, with U2X0");
-	hw::ubrr0::write(static_cast<std::uint8_t>(baud.ubrr));
+	hw::reg<"UBRR0">::write(static_cast<std::uint8_t>(baud.ubrr));
 	hw::ucsr0a::write(hw::ucsr0a::u2x0(1));
 	// General-purpose registers are undefined at power-on (no crt zeroes them);
 	// the direction latch must start "not receiving" so the first rx() enables
@@ -262,15 +248,13 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 	// same reason.
 	g_receiving = 0;
 
-	// Activation: 3x'@', each inside the config page's timeout window (rx
-	// floors it so a corrupt page cannot lock the loader out); anything else  -
-	// including silence - hands over.
+	// Activation: 3×'@', each inside the config page's timeout window (rx
+	// floors it so a corrupt page cannot lock the loader out); anything else —
+	// including silence — hands over.
 	g_window = avr::flash_load(flash_ptr(app_end + 2));
-	for (std::uint8_t k = 3; k; --k) {
-		if (rx() != knock) {
+	for (std::uint8_t k = 3; k; --k)
+		if (rx() != knock)
 			appjump();
-		}
-	}
 	g_window = comm_window;
 
 	// Password gate (config page from app_end+3, 0xff-terminated; a blank
@@ -284,19 +268,17 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 		std::uint8_t expected = avr::flash_load(flash_ptr(g_addr)) & mask;
 		++g_addr;
 		if (expected == 0xff) {
-			g_addr = reinterpret_cast<std::uint16_t>(info.data());
+			g_addr = reinterpret_cast<std::uint16_t>(&info[0]);
 			g_cnt = sizeof(info);
 			sendf();
 			break;
 		}
 		std::uint8_t got = rx();
 		if (got == 0) {
-			if (mask == 0) {
+			if (mask == 0)
 				continue;
-			}
-			if (rcnf() != confirm || rcnf() != confirm) {
+			if (rcnf() != confirm || rcnf() != confirm)
 				appjump();
-			}
 			erase_application(); // leaves g_addr = 0 for the EEPROM walk
 			do {
 				eewr(0xff);
@@ -305,9 +287,8 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 			erase_below();
 			break;
 		}
-		if (got != expected) {
+		if (got != expected)
 			mask = 0;
-		}
 	}
 
 	for (;;) {
@@ -316,27 +297,23 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 		switch (rx()) {
 		case 'f': // read application flash, one page per host '!'
 			for (;;) {
-				if (rx() != confirm) {
+				if (rx() != confirm)
 					break;
-				}
 				g_cnt = page;
 				sendf();
-				if (g_addr >= app_end) {
+				if (g_addr >= app_end)
 					break;
-				}
 			}
 			break;
 		case 'F':                // erase the application, then take pages behind '?'
 			erase_application(); // leaves g_addr = 0, the write start
-			while (rcnf() == confirm) {
+			while (rcnf() == confirm)
 				store_flash();
-			}
 			break;
 		case 'e': // read EEPROM, one page per host '!', until the host stops
 			for (;;) {
-				if (rx() != confirm) {
+				if (rx() != confirm)
 					break;
-				}
 				g_cnt = page;
 				do {
 					tx(eerd());
@@ -358,9 +335,8 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 			sendf();
 			break;
 		case 'C': // replace the config page, then echo it back to verify
-			if (rcnf() != confirm) {
+			if (rcnf() != confirm)
 				break;
-			}
 			g_addr = app_end + page;
 			erase_below(); // leaves g_addr = app_end, the store target
 			store_flash();
@@ -374,6 +350,14 @@ extern "C" [[noreturn]] void tsb_app(); // the application's reset vector: --def
 } // namespace
 } // namespace tsb
 
-// Reset lands at the boot section base (BOOTRST): the entry stub in .vectors
-// is laid first and does the one line of crt a crt-less image needs.
-template struct avr::startup::entry<tsb::run, avr::startup::stack::hardware>;
+// Reset lands here: BOOTRST vectors to the boot section base and .vectors is
+// laid first, so this is the first instruction executed. No crt ran, so set
+// the stack pointer before anything is called.
+extern "C" [[gnu::naked, gnu::used, gnu::section(".vectors")]] void __boot_entry()
+{
+	SP = RAMEND;
+	// The one line of crt this loader needs: compiled code assumes
+	// __zero_reg__ (r1) is 0, and power-on registers are undefined.
+	asm volatile("clr __zero_reg__");
+	tsb::run();
+}