Brink

Brink is a domain specific language for linking and composing of an output file. Brink simplifies construction of complex files by managing sizes, offsets and ordering in a readable declarative style. Brink tries to be especially useful when creating FLASH, ROM or other non-volatile memory images.

Quick Start

Install Prebuilt Binaries for Linux

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/steveking-gh/brink/releases/download/7.0.2/brink-installer.sh | sh

Install Prebuilt Binaries for Windows

Start a command prompt and execute the following:

powershell -ExecutionPolicy Bypass -c "irm https://github.com/steveking-gh/brink/releases/download/7.0.2/brink-installer.ps1 | iex"

Build From Source

Step 1: Install Rust

Brink is written in rust, which works on all major operating systems. Installing rust is simple and documented in the Rust Getting Started guide.

Step 2: Clone Brink

From a command prompt, clone Brink and change directory to your clone. For example:

git clone https://github.com/steveking-gh/brink.git
cd brink

Step 3: Build and Run Self-Tests

cargo test --release --all

All tests should pass, 0 tests should fail.

Step 4: Install Brink

The previous build step created the Brink binary as ./target/release/brink. You can install the Brink binary anywhere on your system. As a convenience, cargo provides a per-user installation as $HOME/.cargo/bin/brink.

cargo install --path ./

What Can Brink Do?

Brink can assemble any number of input files into a unified output.

Brink can calculate relative or absolute offsets, allowing the output to contain pointer tables, cross-references and so on.

Brink can add pad bytes to force parts of the file to be a certain size.

Brink can add pad bytes to force parts of the file to start at an aligned boundary or at an absolute location.

Brink can write your own strings and data defined within your Brink source file.

Brink provides full featured assert and print statement support to help with debugging complex output files.

Hello World

For a source file called hello.brink:

/*
 * A section defines part of an output.
 */
section foo {
    // Print a quoted string to the console
    print "Hello World!\n";
}

// An output statement outputs the section to a file
output foo;

Running Brink on the file produces the expected message:

$ brink hello.brink
Hello World!
$

Brink also produced an empty file called output.bin. This file is the default output when you don't specify some other name on the command line with the -o option. Why is the file empty? Because nothing in our program produced output file content -- we just printed the console message.

Let's fix that. We can replace the print command with the wrs command, which is shorthand for 'write string':

/*
 * A section defines part of an output.
 */
section foo {
    // Write a quoted string to the output
    wrs "Hello World!\n";
}

// An output statement outputs the section to a file
output foo;

Now, running the command again:

$ brink hello.brink
$

Produces output.bin containing the string Hello World!\n.

Basic Structure of a Brink Program

A Brink source file consists of one or more section definitions and exactly one output statement. The output statement specifies the top-level section that defines the output file. Starting from this top section, Brink recursively evaluates each nested section and command to produce the output file. For example, we can define a section with a write-string (wrs) command:

section foo {        // Start a new section named 'foo'
    wrs "I'm foo";   // wrs writes a string into the section.
}

output foo;          // Final output

Produces a default output named output.bin.

$ cat output.bin
I'm foo

Using the wr command, sections can embed other sections:

section foo {
    wrs "I'm foo\n";
}

section bar {
    wrs "I'm bar\n";
    wr foo;           // nested section
}

output bar;

Produces output.bin:

$ cat output.bin
I'm bar
I'm foo

Users can extend Brink with custom data processing using Brink extensions. Users write the output of their extension call into a section with a wr.

section foo {
    wrs "I'm foo\n";
}

section bar {
    wrs "I'm bar\n";
    wr foo;           // nested section
}

section final {
    wr bar;
    wr my_stuff::crc(bar);  // Write a 4 byte CRC hash for section 'bar'.
}

assert(sizeof(final) == 20);
output final;

Assert and Print

To aid in debug, Brink supports assert and print statements in your programs.

Assert expressions automate error checking. This example verifies our expectation that section 'bar' is 13 bytes long.

section bar {
    wrs "Hello World!\n";
    assert sizeof(bar) == 13;
}
output bar;

You can print this length information to the console during generation of your output:

section bar {
    print "Output size is ", sizeof(bar), " bytes\n";
    wrs "Hello World!\n";
    assert sizeof(bar) == 13;
}
output bar;

Prints the console message:

Output size is 13 bytes

Addresses and Offsets

Unlike the GNU linker 'ld' concept of a location counter, Brink uses scoped addresses and scoped offsets to track locations. Addresses and offsets are 64-bit unsigned values that mark the position of the next byte of output. Brink allows users to reference and manipulate these values, adding pad bytes as necessary.

Importantly, addresses and offsets are scoped to their enclosing section. When entering a nested (child) section, Brink saves the outer (parent) section's inflight address and offset values. When exiting a child section, Brink restores and updates the parent's address and offset values. From the perspective of the parent section, a child section is a wr with the parent's addresses and offsets updated per the size of the child.

For the specific case of the address and address offset, a child section inherits these values by default from the parent section. If the child section does not use set_addr, then the address and address offset simply continue growing in step with the parent.

The only global (non-scoped) offset is the file_offset. Starting from 0, this value monotonically increases to the end of the output file.

The following table provides a summary of the addresses and offsets used in Brink.

Variable	Section Entry	Section Exit	`set_addr`	`pad_sec_offset`	`pad_addr_offset`	`pad_file_offset`
Address	No Change	Restore & Update	Set	Pad Forward	Pad Forward	Pad Forward
Address Offset	No Change	Restore & Update	Set to 0	Pad Forward	Pad Forward	Pad Forward
Section Offset	Set to 0	Restore & Update	No change	Pad Forward	Pad Forward	Pad Forward
File Offset	No Change	No Change	No Change	Pad Forward	Pad Forward	Pad Forward

The following diagram shows several address and offset concepts. Users specify the starting logical address of the output section D using a region. Alternatively, users can change the address within D using set_addr at the top of the output section.

In this example, section D is the top level binary output and includes three other sections A, B and C. The starting address of section D is 0x8000 because the user placed D in FLASH region. Section C is special and defines its own starting address. A boot loader can find section C using the pointer at the beginning of the file, then copy section C to its proper starting address of 0xF000. Notice that addr(D) used in the context of section D returns 0x8C00, not the starting address value 0xF000 nested within section C.

Brink Disallows Address Overwrites

By address, Brink tracks all bytes written to the output. Brink reports an error if a program's offset or address manipulations cause more than one write to the same address.

Brink Disallows Negative Offset Changes

Brink enforces that set offset commands must specify an offset change greater or equal to 0. Brink emits pad bytes into the output for any offset change greater than 0.

Brink Disallows Address and Offset Overflows

Brink emits an error if an address or offset change causes 64-bit unsigned overflow. In other words, programs cannot use unsigned overflow wrapping back to 0.

Order of Execution

As a mental model, user's can think of program execution as occurring in output order. Output order means the sequence of operations that produce bytes in-order starting with the initial byte of the output file. In other words, an operation producing the first byte of the output will execute before an operation producing the second byte.

Within a section definition, output order and source code order are the same. However, outside of a section definition, output order and program order may differ. For example, source code may define whole sections in a different order than instantiated into in the output.

Output Creation Phases

This section provides an overview of Brink's internal output creation phases.

const Evaluation Phase: First, Brink evaluates all const expressions. This phase includes evaluation of all if/else statements and the dependent const-time operations such as include statements in the taken path.
Layout Phase: Next, Brink iteratively evaluates all expressions that affect output size and layout. For example, Brink evaluates align expressions and extension size() calls during this phase. Brink skips data generation, since knowing the size of operations suffices to determine the precise output structure. This phase completes when successive layout iterations produce identical results.
Generate Phase 1: Next, Brink begins populating data values into the output. In this first generation phase, Brink first evaluates wr statements that do NOT call extensions. Brink evaluates wr calls in output order.
Generate Phase 2: Next, Brink evaluates wr statement that call an extension. Like before, brink evaluates extension calls in output order. Brink executes all extension calls serially on the engine thread.
Validation Phase: Finally, Brink evaluates assert statements, including those that call extensions. Note that Brink may take an early exit in any phase if an assert statement will unambiguously fail.

Command Line Options Reference

brink [OPTIONS] <input>

The required input file contains the brink source code to compile and build the output file. Brink source files typically have a .brink file extension.

Option	Description
`-D<name>[=value]`	Defines a `const` value from the command line. See Command-Line Const Defines below.
`--list-extensions`	List all available extensions compiled into brink as controlled by Cargo feature flags.
`--max-output-size=<size>`	Reject the output if its size exceeds `<size>` bytes before writing data. Accepts a plain integer or a K/M/G suffix (e.g. `64M`, `512K`, `1G`). Default is `256M`.
`--map-csv`	Writes a CSV format map file `<stem>.map.csv` to the current directory. For example: `firmware.brink` → `firmware.map.csv`.
`--map-csv=<file>`	Writes a CSV map file to the specified file.
`--map-csv=-`	Writes a CSV map file to stdout.
`--map-c99`	Writes a C99 header file `<stem>.map.h` to the current directory. For example: `firmware.brink` → `firmware.map.h`.
`--map-c99=<file>`	Writes a C99 header to the specified file.
`--map-c99=-`	Writes a C99 header to stdout.
`--map-json`	Writes a JSON format map file `<stem>.map.json` to the current directory. For example: `firmware.brink` → `firmware.map.json`.
`--map-json=<file>`	Writes a JSON map to the specified file.
`--map-json=-`	Writes a JSON map to stdout.
`--map-rs`	Writes a Rust module file `<stem>.map.rs` to the current directory. For example: `firmware.brink` → `firmware.map.rs`.
`--map-rs=<file>`	Writes a Rust module map to the specified file.
`--map-rs=-`	Writes a Rust module map to stdout.
`--noprint`	Suppress `print` statement output from the source program.
`-o <file>`	Output file name. Defaults to `output.bin`.
`-q`, `--quiet`	Suppress all console output, including errors. Overrides `-v`. Useful for fuzz testing.
`-v`	Increase verbosity. Repeat up to four times (`-v -v -v -v`).

When the user does not specify a path, Brink writes map file(s) and the output to the current working directory.

Command-Line Const Defines

The -D option injects a const definition into the program from the command line. This option is modelled after the GCC -D preprocessor syntax. You can specify -D multiple times, once per each definition. For example:

brink -DBASE=0x8000 -DCOUNT=16 firmware.brink

The name must be a valid Brink identifier. The value is optional; without a value, Brink sets the const to 1, with type Integer, following the GCC boolean-flag convention.

-D overrides any same-named const definition in the source.

Map output lists all const definitions including -D consts.

Value Type Inference

Brink knows or infers the type from the value string using the same rules as source code for type inference.

Example	Value	Type	Description
`-DFLAG`	1	`Integer`	Defaults to true (1).
`-DCOUNT=16`	16	`Integer`	Plain decimal → `Integer`
`-DBASE=0x1000`	0x1000	`U64`	Hex/binary without suffix → implicit `U64`
`-DBASE=0x1000u`	0x1000	`U64`	`u` suffix → explicit `U64`
`-DOFFSET=0x40i`	0x40	`I64`	`i` suffix → explicit `I64`
`-DDELTA=-4`	-4	`I64`	Negative decimal → implicit `I64`

Example

Define a base address and section count at the command line:

brink -DBASE=0x0800_0000 firmware.brink -o firmware.bin

The source can reference BASE as an ordinary const:

section entry { wr8 0x01; }
section top   { set_addr BASE; wr entry; }
output top;

Brink Language Reference

Comments

Brink supports C language line and block comments.

Whitespace

Brink supports lenient C language style whitespace rules.

Semicolon Termination

Like C language, statements must be terminated with a trailing semicolon character.

Types

Brink supports the following data types:

U64: 64-bit unsigned values
I64: 64-bit signed values
Integer: 64-bit integers with flexible sign treatment
String: UTF-8 string in double quotes

Brink reports an error for under/overflow on arithmetic operations on U64, I64 and Integer types as described in Arithmetic Operators.

Identifiers

An identifier begins with a letter (A–Z, a–z) or an underscore (_), followed by zero or more letters, digits (0–9), or underscores. Identifiers are case-sensitive.

Reserved Identifiers

Brink reserves certain identifiers and rejects their use as section names, const names, or label names at compile time.

Brink also reserves two identifier prefixes. Any user defined identifier beginning with a reserved prefix triggers an error.

Reserved Prefix	Reason
`wr` + digit	Numeric write instructions (`wr8`, `wr16`, `wr32`, and future width variants)
`__`	Leading double underscore names refer to builtin identifiers.

Brink also reserves the following exact keywords:

Reserved Keyword	Reason / possible future use
`import`	Module inclusion
`if`	Conditional section inclusion
`else`	Conditional section inclusion
`true`	Boolean literal
`false`	Boolean literal
`extern`	External section references
`let`	Variable declarations
`fill`	Fill / pad byte ranges

Keyword reservation is case-sensitive. Fill and FILL are valid identifiers; fill is not.

Literals

Number Literals

Brink supports number literals in decimal, hex (0x) and binary (0b) forms. After the first digit, you can use '_' within number literals to help with readability.

assert 42 == 42;
assert -42 == -42;
assert 0x42 == 0x42;
assert 0x42 == 66;
assert 0x4_2 == 66;
assert 0x42 == 6_6;

assert 0b0 == 0;
assert 0b01000010 == 0x42;
assert 0b0100_0010 == 0x42;
assert 0b101000010 == 0x142;
assert 0b0000_0000_0100_0010 == 0x42;

The following table summarizes how Brink determines the type of number literals.

Example	Type	Description
4	Integer	Simple decimal numbers are `Integer` type with flexible signedness
4u	U64	Explicitly `U64`
4i	I64	Explicitly `I64`
-4	I64	Negative numbers are `I64`
0x4	U64	Hex numbers are `U64` by default
0x4i	I64	Explicitly `I64` hex number
0b100	U64	Binary numbers are `U64` by default

Brink does not support negative hex or binary literals.

For convenience, the compiler casts the flexible Integer type to U64 or I64 as needed.

assert 42u == 42;  // U64 operates with Integer
assert 42i == 42;  // I64 operates with Integer

Otherwise the types used in an expression must match. For example:

assert 42u == 42i; // mix unsigned and signed

Produces an error message:

[EXEC_13] Error: Input operand types do not match.  Left is 'U64', right is 'I64'
   ╭─[tests/integers_5.brink:2:12]
   │
 2 │     assert 42u == 42i; // mix unsigned and signed
   ·            ^^^    ^^^
───╯

Users can explicitly cast a number literal or expression to the required signedness using the built-in to_u64 to to_i64 functions. For example:

assert -42 != to_i64(42);  // comparing signed to unsigned

The to_u64 and to_i64 functions DO NOT report an error if the runtime value under/overflows the destination type.

assert 0xFFFF_FFFF_FFFF_FFFF == to_u64(-1); // OK
assert to_i64(0xFFFF_FFFF_FFFF_FFFF) == -1; // OK

Number Magnitude

Decimal number literals accept an optional K/M/G magnitude suffix (case sensitive) before the type suffix.

Suffix	Multiplier	Example	Value
`K`	1024	`64K`	65536
`M`	1024 × 1024	`1M`	1048576
`G`	1024 × 1024 × 1024	`2G`	2147483648

Magnitude and type suffixes combine: 4Ku is 4096 as a U64, -1Ki is -1024 as an I64.

True and False

Brink considers a zero value false and all non-zero values true.

Quoted Strings

Brink allows utf-8 quoted strings with the following escape characters:

Escape Character	UTF-8 Value	Name
\0	0x00	Null
\t	0x09	Horizontal Tab
\n	0x0A	Linefeed
\"	0x22	Quotation Mark

Newlines are Linux style, so "A\n" is a two byte string on all platforms.

Arithmetic Operators

Brink supports the following arithmetic operators with same relative precedence as the Rust language. Where applicable, Brink checks for arithmetic under/overflow.

Precedence	Operator	Under/Overflow Check?	Description
Highest	( )	n/a	Paren grouping
	* /	yes	Multiply and divide
	+ -	yes	Add and subtract
	&	n/a	Bitwise-AND
	\|	n/a	Bitwise-OR
	<< >>	no	Bitwise shift up and down
	== !=	n/a	Equal and non-equal
	>= <=	n/a	Greater-than-or-equal and less-than-or-equal
	&&	n/a	Logical-AND
Lowest	\|\|	n/a	Logical-OR

addr

addr( [identifier] ) -> U64

When called with an identifier, returns the address of the identifier as a U64. When called without an identifier, returns the current address. See Addresses and Offsets for more information.

The following table shows the scoping rules for addr. To summarize, Brink tracks exactly one address value per name. An addr(<name>) command retrieves that one value regardless of the scope of the caller.

Command Form	Scope used to determine address
`addr()`	Scope of current section
`addr(<section name>)`	Scope of parent section that contains the child section
`addr(<output section name>)`	Scope of the `output` section
`addr(<label name>)`	Scope of the section that contains the label

Example:

const BASE = 0x1000u;

section fiz {
    assert addr() == BASE + 6;
    wrs "fiz";
    assert addr() == BASE + 9;
    assert addr(foo) == BASE;
}

section bar {
    assert addr() == BASE + 3;
    wrs "bar";
    assert addr() == BASE + 6;
    wr fiz;
    assert addr() == BASE + 9;
}

// top level section
section foo {
    set_addr BASE;
    assert addr() == BASE;
    wrs "foo";
    assert addr() == BASE + 3;
    assert addr(fiz) == BASE + 6;
    wr bar;
    assert addr() == BASE + 9;
    assert addr(bar) == BASE + 3;
}

output foo;

addr_offset

addr_offset( [identifier] ) -> U64

Returns the offset from the output or most recent set_addr anchor as a U64. When called without an identifier, returns the current address offset. When called with an identifier, returns the address offset at the start of the named section or label.

The offset resets to zero on each set_addr call.

The following table shows the scoping rules for addr_offset. To summarize, Brink tracks exactly one address offset value per name. An addr_offset(<name>) command retrieves that one value regardless of the scope of the caller.

Command Form	Scope used to determine address
`addr_offset()`	Scope of current section
`addr_offset(<section name>)`	Scope of parent section that contains the child section
`addr_offset(<output section name>)`	Scope of the `output` section
`addr_offset(<label name>)`	Scope of the section that contains the label

Example:

const BASE = 0x1000u;

section fiz {
    assert addr_offset() == 6;
    wrs "fiz";
    assert addr_offset() == 9;
    assert addr_offset(foo) == 0;
}

section bar {
    assert addr_offset() == 3;
    wrs "bar";
    assert addr_offset() == 6;
    wr fiz;
    assert addr_offset() == 9;
}

// top level section
section foo {
    set_addr BASE;
    assert addr_offset() == 0;
    wrs "foo";
    assert addr_offset() == 3;
    assert addr_offset(fiz) == 6;
    wr bar;
    assert addr_offset() == 9;
    assert addr_offset(bar) == 3;
}

output foo;

align

align <expression> [, <pad byte value>];

The align statement writes pad bytes into the current section until the absolute location counter reaches the specified alignment. Align writes 0 as the default pad byte value, but the user may optionally specify a different value.

Example:

section foo {
    wrs "Hello";
    align 32;
    assert sizeof(foo) == 32;
    assert addr() == 32;
}

output foo;

assert

assert <expression>;

The assert statement reports an error if the specified expression does not evaluate to a true (non-zero) value. Assert expressions provide a means of error checking and do not affect the output file.

Example:

section foo {
    assert 1;   // OK, non-zero is true
    assert -1;  // OK, non-zero is true
    assert 1 + 1 == 2;
}

output foo;

const

const <identifier> = <expr>;

A const expression creates an immutable user defined identifier for a value. The value can consist of a number or string literal, or an expression composed of other constants and literals. Const identifier names have global scope and must be globally unique. Const identifiers cannot conflict with any other global identifiers such as section names.

Example:

const RAM_BASE = 0x8000_0000u;  // User defined unsigned constant.

section foo {
    set_addr RAM_BASE;
    wr64 RAM_BASE;
    print "RAM base address is ", RAM_BASE, "\n";
}

output foo;

Const expressions support the full set of arithmetic, bitwise and comparison operators. Comparison operators evaluate to 1 (true) or 0 (false) and are useful for expressing relationships between constants:

const FLASH_BASE = 0x0800_0000;
const FLASH_SIZE = 0x0008_0000;
const RAM_BASE   = 0x2000_0000;

// Verify flash and RAM regions do not overlap
const NO_OVERLAP = (FLASH_BASE + FLASH_SIZE) <= RAM_BASE;
assert NO_OVERLAP;

A const value expression cannot depend on addresses, sizes, offsets or any other dynamic aspect of the output file. Brink resolves all const values before starting layout of the output. For example:

const RAM_BASE = 0x8000_0000;         // OK, just a 64b unsigned literal.
const RAM_SIZE = 32768;               // OK, just a 64b integer literal.
const RAM_END = RAM_BASE + RAM_SIZE;  // OK, const composed of other consts.

section foo {
    wrs "Hello\n";
}

const RAM_USED = sizeof(foo);         // ERROR!  Const cannot depend on section properties.

output foo;

Deferred Assignment

const variables support deferred assignment. This allows the user to declare a const variable, then assign a value to the variable exactly once in later code. For example:

const IO_START;
...
IO_START = 0xF000_0000_0000_0000;

Deferred assignment is primarily useful in if/else statements, which allow users to conditionally determine the value to assign.

To provide errors and warnings, Brink tracks the defined/undefined and used/unused state of each variable.

if/else

if <expression> { ... } else { ... }

Allows conditional execution of other statements. As described in Output Creation, Brink evaluates all if/else statements before starting layout of the output. Therefore, an if/else expression must only depend on const variables and literal values. In other words, if/else statements must not depend on dynamic addresses, sizes, offsets or any other layout dependent aspect of the output file.

Users must pre-declare const variables before conditionally assigning values to them. For example:

// Assume the user specified -DMEM_CONFIG="BIG" on the command line.

// Pre-declare variables prior to conditional assignment in an if/else.
// Brink strictly tracks variable definitions to prevent use of
// uninitialized variables.
const FLASH_SIZE;
const RAM_SIZE;

print "Memory configuration is ", MEM_CONFIG, "\n";
if MEM_CONFIG == "BIG" {
    FLASH_SIZE = 0x8_0000;
    RAM_SIZE = 0x80_0000;
    include "big_config.brink";
} else {
    if MEM_CONFIG == "MEDIUM" {
        FLASH_SIZE = 0x4_0000;
        RAM_SIZE = 0x40_0000;
        include "medium_config.brink";
    } else {
        if MEM_CONFIG == "SMALL" {
            FLASH_SIZE = 0x2_0000;
            RAM_SIZE = 0x20_0000;
            include "small_config.brink";
        } else {
            print "Invalid configuration. MEM_CONFIG must be BIG, MEDIUM, or SMALL.\n";
            assert(0);  // Halt execution
        }
    }
}

If the taken path in an if/else statement does not assign a value to a predeclared const variable, then Brink reports an error if any later program statement uses that variable.

For compactness, user's may omit braces around an else/if block. For example:

if MEM_CONFIG == "BIG" { include "big_config.brink"; }
else if MEM_CONFIG == "MEDIUM" { include "medium_config.brink"; }
else if MEM_CONFIG == "SMALL" { include "small_config.brink"; }
else { assert(0); }

include

include "<file>";

Includes another Brink source file. Brink processes the included file as if it were part of the current file. For example, the included file can define sections, labels, constants and nested include files.

An included file may contain an output statement. Brink will enforce that the entire program after include file resolution contains only one output statement. See the output statement for more information.

The default path for an included file is the directory of the source file that contains the include statement. For example, if main.brink is in /home/user/project/ and contains include "sections.brink", then Brink will look for /home/user/project/sections.brink.

Include files starting with a / are absolute paths. Likewise, Brink supports relative paths such as ../.

All paths use Linux style forward slashes.

Example:

// file: main.brink
include "../constants.brink";
include "sections.brink";

output main_rom;

// file: ../constants.brink
const RAM_BASE = 0x8000_0000u;

// file: sections.brink
section main_rom {
    set_addr 0x1000;
    wrs "Hello\n";
}

Labels

<identifier>:

Labels assign an identifier to a specific location in the output file. Programs can then refer to the location of the label by name. Labels names have global scope and label names must be globally unique. Multiple different labels can refer to the same location.

Labels have the form <label identifier>: and can prefix most statement types.

For example:

section foo {
    set_addr 0x1000;
    // assign the label 'lab1' to the current location
    lab1: wrs "Wow!";
    // assign the label 'lab2' to the current location
    lab2:
    assert addr(lab1) == 0x1000;
    assert addr(lab2) == 0x1004;
    assert addr(lab3) == 0x1004;
    // yet another label, same location as 'lab2'
    lab3:
}

output foo;

output

output <section identifier>;

An output statement specifies the top section to write to the output file. Use set_addr inside the section to control the absolute starting address, or place the top section in region with a start address.

A Brink program must have exactly one output statement.

An include file may contain an output statement. Brink will enforce that the entire program after include file resolution contains only one output statement.

print

print <expression> [, <expression>, ...];

The print statement evaluates the comma separated list of expressions and prints them to the console. For expressions, print displays unsigned values in hex and signed values in decimal. If needed, the to_u64 and to_i64 functions can control the output style.

Brink executes a given print statement for each instance found in the output file. In other words, a print statement in a section written multiple times will execute multiple times in the order found.

Example:

const BASE = 0x1000;

section bar {
    print "Section 'bar' starts at ", addr(), "\n";
    wrs "bar";
}

// top level section
section foo {
    set_addr BASE;
    print "Output spans address range ", BASE, "-", BASE + sizeof(foo),
          " (", to_i64(sizeof(foo)), " bytes)\n";
    wrs "foo";
    wr bar;
    wr bar;
    wr bar;
}

output foo;

Will result in the following console output:

Output spans address range 0x1000-0x100C (12 bytes)
Section 'bar' starts at 0x1003
Section 'bar' starts at 0x1006
Section 'bar' starts at 0x1009

region

region <identifier> { ... }

A region defines the name and static properties of an address range. Regions provide a way to decouple memory placement and top-down layout control from the section content being placed. Unlike sections, regions are stateless and do not track dynamic information during layout.

Users place exactly one section in a region. We refer to this section as the bound section of the region. The bound section is a normal section with the following extra behaviors:

The region sets the starting address of the bound section.
The region caps the size of the bound section.

For example:

// Define the properties of the FLASH memory region
region FLASH {
    addr = 0xF000_0000;
    size = 1M;
}

// Define the properties of the EEPROM memory region
region EEPROM {
    addr = 0xFF00_0000;
    size = 64K;
}

// Flash sections
section boot { ... }
section flash_code { ... }
section flash_data { ... }

// EEPROM sections
section eeprom_data1 { ... }
section eeprom_data2 { ... }

// FLASH_TOP is the bound section in the FLASH region
section FLASH_TOP in FLASH {
    // Starts at address 0xF000_0000
    wr boot;
    wr flash_code;
    wr flash_data;
}

section EEPROM_TOP in EEPROM {
    assert addr() == 0xFF00_0000;
    wr runtime_code;
    wr runtime_data;
}

// The output file contains the image for FLASH and EEPROM regions.
// This section is not a bound section of a region and behaves
// like any other section.
section FIRMWARE_UPDATE_FILE {
    wr file_offset(FLASH_TOP);    // Offset to the new FLASH image
    wr file_offset(EEPROM_TOP);   // Offset to the EEPROM image
    wr FLASH_TOP;                 // FLASH image
    wr EEPROM_TOP;                // EEPROM image
}

output FIRMWARE_UPDATE_FILE;  // Write the output

Region Properties

Regions support the following properties:

addr Starting address (required)
size Size in bytes (required)

Region Property `addr`

The addr property defines the region's absolute starting address. The region's bound section starts at this address. Users can query the addr property of a region with addr(<region name>).

Region Property `size`

Specifies the size of the region in bytes. Brink reports an error if the size of the bound section exceeds this value.

Users can query the size property of a region with sizeof(<region name>).

The size value accepts a K/M/G magnitude suffix.

Region Boundary Enforcement

Regions provide automatic size and boundary checking for all operations in the region. In practical terms this means:

Write commands in a region cannot extend outside the region
Address manipulation in a region cannot result in an address outside the region
Offset manipulations in a region cannot result in a offset outside the region.

For example, the following wr32 command would extend outside the region by one byte, resulting in an error:

region LITTLE_ROM {
    addr = 0;
    size = 7;
}

section data in LITTLE_ROM {
    // occupies bytes 0-3
    wr32 0x12345678;
    // Occupies bytes 4-7, which extends 1 byte outside the region
    wr32 0x87654321;  // ERROR!
}

Of course, region enforcement occurs not just in the region's bound section, but in any reachable section. For example:

region LITTLE_ROM {
    addr = 0;
    size = 7;
}

section nested_stuff {
    pad_sec_offset 6;  // pad to last byte of region
    wr more_nested;
}

section more_nested {
    wr std::crc32c(more_nested);  // 4 bytes of output
}

section data in LITTLE_ROM {
    wr nested_stuff;  // ERROR!  Data written outside of region
}

The set_addr command and any offset manipulation commands are also constrained to fit in the region. For example:

region FLASH {
    addr = 0xF000_0000;
    size = 1M;
}

section foo in FLASH {
    assert addr() == 0xF000_0000;  // Start of FLASH region
    set_addr 0xF000_1000;          // OK, inside the region
    wrs "Inside region!";
    set_addr 0xA000_0000;          // ERROR, outside the region
    wrs "Outside region!";
}

output foo;

Nested Regions

Users can freely nest sections in different regions into each other. However, Brink allows write operations only in the address range intersection permitted by all the parent regions. For example:

region READ_ONLY {
    addr = 0xF000_0000;
    size = 0x1000_0000;
}

region FLASH {
    addr = 0xF100_0000;
    size = 64K;
}

section flash_data in FLASH {
    assert addr() == 0xF100_0000;
    ...
}

section ro_data in READ_ONLY {
    assert addr() == 0xF000_0000;
    // OK, region FLASH is a subset of READ_ONLY.
    // Note that the FLASH region anchors the starting address
    // at 0xF100_0000.  This creates a logical (unpadded) address gap
    // in the ro_data section between 0xF000_0000 and 0xF100_0000.
    wr flash_data;
}

output ro_data;

Partially Overlapping Nested Regions

For completeness, the region of a nested section need not be a proper subset of the parent region. Brink still enforces the constraints of all parent sections as follows:

Any address written by the child section must lie in the intersection of all parent regions.
The starting address of a nested section must fit the address range allowed by all parent regions.

Sections in Regions are Single Use

Placing a section in a region forces the starting address of the section to the region's addr value. Writing this section more than once results in an address address conflict with the previous instance of the section.

sec_offset

sec_offset( [identifier] ) -> U64

When called with an identifier, returns the unsigned 64-bit offset of the identifier from the start of the section that contains the identifier. When called without an identifier, returns the offset from the start of the current section.

Example:

section fiz {
    assert sec_offset() == 0;
    wrs "fiz";
    assert sec_offset() == 3;
}

section bar {
    assert sec_offset() == 0;
    wrs "bar";
    assert sec_offset() == 3;
    wr fiz;
    assert sec_offset() == 6;
    assert sec_offset(fiz) == 3;
}

// top level section
section foo {
    assert sec_offset() == 0;
    wrs "foo";
    assert sec_offset() == 3;
    wr bar;
    assert sec_offset() == 9;
}

output foo;

When a section offset specifies an identifier, the identifier must be in the scope of the current section. For example:

section fiz {
    wrs "fiz";
}

section bar {
    wr fiz;
    assert sec_offset(fiz) == 0; // OK fiz in scope in section bar
}

section foo {
    wr bar;
    assert sec_offset(bar) == 0; // OK, bar is local in this section
    assert sec_offset(fiz) == 0; // ERROR, fiz is out of scope in section foo
}

output foo;

section

section <name> [in <region>] { ... }

A section is a named, reusable block of content. Sections are the primary building block of a Brink program. Each section defines a sequence of bytes, built up from write statements and padding operations such as align. Sections may also contain labels, assertions, print statements and so on. Sections may write other sections into themselves so long as the nesting does not create a cycle.

Section names must be valid identifiers, must be globally unique, and must not conflict with const names, label names, region name, or reserved identifiers.

Sections have their own section-relative location counter which resets to zero at the start of each section. Sections can read and advance the section location counter with sec_offset() and pad_sec_offset() statements respectively.

The root section named in the output statement is the only section Brink writes to the output file. Other sections can be directly or indirectly included via wr statements from the output section. Unreachable sections produce a warning.

Sections In Regions

To help guide layout, users can place a exactly one section in a region with in <region name> after the section name. We call a section placed in a region as the bound section of the region.

Example:

section magic {
    wrs "FIRM";           // 4-byte magic number
    wr8 0x01;             // version
    assert sec_offset() == 5;    // Section location counter should be 5
}

section body {
    wr8 0xAA, 16;         // 16 bytes of payload
}

section image {
    wr magic;
    align 256;            // Body should start on 256 byte boundary
    wr body;
    assert sizeof(image) == 272;  // 256 + 16
}

output image;

set_addr

set_addr <expression>;

The set_addr command forces the current address to the specified value and resets the current addr_offset to zero. These changes happen within the scope of the containing section. Child sections inherit the parent section's addr and addr_offset values.

Using set_addr does not change the value of the section offset nor file offset. A set_addr command does not add pad bytes to the output.

The set_addr command may move the address forward or backwards. However, Brink tracks every output byte by address and reports an error if a program tries to write to the same address more than once.

Example:

section foo {
    wr8 1;
    wr8 2;
    wr8 3;
    wr8 4;
    wr8 5;
    set_addr 16;
    assert addr() == 16;
    assert addr_offset() == 0;   // set_addr resets addr_offset
    assert file_offset() == 5;  // set_addr does not pad
    assert sec_offset() == 5;
    wr8 0xAA, 3;
    assert addr_offset() == 3;
    assert file_offset() == 8;
    assert sec_offset() == 8;
    pad_sec_offset 24, 0xFF;     // Adds 24 - 8 = 16 pad bytes
    assert addr() == 35;         // 19 + 16 = 35
    assert addr_offset() == 19;  // 3 + 16 = 19
    assert file_offset() == 24;  // 8 + 16 = 24
    assert sec_offset() == 24;   // 8 + 16 = 24
}

output foo;

When used in a section in a region, Brink reports an error if the set_addr command sets the address outside of a region.

pad_addr_offset

pad_addr_offset <expression> [, <pad byte value>];

Pads the output until addr_offset reaches the specified value. Users may specify an optional pad byte value or use the default value of 0.

If the specified value is less than the current addr_offset, Brink reports an error.

pad_addr_offset is most useful after a set_addr call, because set_addr resets addr_offset to zero. This lets users pad to a size relative to their chosen address anchor without knowing what the surrounding section's sec_offset happens to be.

Example:

const BASE = 0x1000;

section header {
    wrs "FIRM";           // 4-byte magic number
    wr8 0x01;             // version byte
}                         // addr_offset == 5 on exit

section body {
    set_addr BASE;
    wr header;
    // Relocate body to its target load address.
    // addr_offset resets to 0.
    set_addr 0xF000;
    wr8 0xAA, 3;          // 3 bytes of payload
    // Pad to 0x20 bytes from the 0xF000 anchor.
    pad_addr_offset 0x20;
    assert addr() == 0xF020;
    assert addr_offset() == 0x20;
    assert sec_offset() == 0x25;  // 5 (header) + 3 (payload) + 29 (pad) = 0x25
}

output body;

pad_file_offset

pad_file_offset <expression> [, <pad byte value>];

The pad_file_offset command pads the output file until the file offset reaches the specified value. Users may specify an optional pad byte value or use the default value of 0.

If the specified offset is less the current offset, Brink reports an error.

pad_file_offset is most useful when a section is written inside a parent section, because sec_offset resets to zero at the start of each child section while file_offset continues from the parent's position. This lets a child section pad to an absolute file position regardless of where the parent places it.

Example:

// A firmware container: an 8-byte header at file offset 0, followed by a
// payload that must start at file offset 512 for bootloader compatibility.

section header {
    wrs "FIRM";       // 4-byte magic
    wr32 0x00000001;  // version
}

section payload {
    // firmware writes header first (8 bytes), so payload opens at
    // file_offset 8.  Pad to the protocol-required file position 512.
    pad_file_offset 512, 0xFF;
    assert file_offset() == 512; // absolute position in the output file
    assert sec_offset() == 504;  // sec_offset starts from 0 inside payload
    wrs "PAYLOAD";               // 7 bytes of payload data
    assert file_offset() == 519;
    assert sec_offset() == 511;
}

section firmware {
    wr header;
    wr payload;
}

output firmware;

pad_sec_offset

pad_sec_offset <expression> [, <pad byte value>];

The pad_sec_offset command pads the current section until the section offset reaches the specified value. Users may specify an optional pad byte value or use the default value of 0.

If the specified offset is less the current offset, Brink reports an error.

Example:

section foo {
    wr8 1;
    wr8 2;
    wr8 3;
    wr8 4;
    wr8 5;
    pad_sec_offset 16;
    assert addr() == 16;
    assert file_offset() == 16;
    assert sec_offset() == 16;
    wr8 0xAA, 3;
    pad_sec_offset 24, 0xFF;
    assert addr() == 24;
    assert file_offset() == 24;
    assert sec_offset() == 24;
    pad_sec_offset 24, 0xEE; // should do Nothing
    wr8 0xAA, 3;
    pad_sec_offset 27, 0x33; // should do nothing
    pad_sec_offset 28, 0x77; // should pad to 28
    assert sizeof(foo) == 28;
}

output foo;

sizeof

sizeof( <identifier> ) -> U64

Returns the size in bytes of the specified identifier.

Example:

section empty_one {}
section foo {
    wrs "Wow!";
    wr empty_one;
    assert sizeof(empty_one) == 0;
    assert sizeof(foo) == 4;
}

output foo;

When called with an extension identifier, sizeof returns the size of the extension's output. For example:

print "CRC size=", sizeof(std::crc32c);  // returns "CRC size=4"

When called with a region identifier, sizeof returns the fixed size of the region regardless of whether the user's program writes any data in the region.

region FLASH { ...; size = 8K; ... }
...
print "FLASH size=", sizeof(FLASH);  // returns "FLASH size=8192"

When called with a section identifier, sizeof returns the size of the section in the file. Therefore, this size does not take into account operations that do not write data nor pad bytes. For example, address jumps, e.g. by using set_addr do not change the sizeof() result for a section.

section foo {
    set_addr 0;
    wrs "Hello\n";
    // Address jumps by 0x1000, but no data nor pads written, so
    // no effect on sizeof(foo).
    set_addr 0x1000;
    wrs "World\n";
    assert sizeof(foo) == 12;
}

to_i64

to_i64( <expression> ) -> I64

Converts the specified expression to the I64 type without regard to under/overflow.

Example:

section foo {
    assert to_i64(0xFFFF_FFFF_FFFF_FFFF) == -1;
    assert to_i64(42u) == 42;
    assert to_i64(42u) == 42i;
    assert to_i64(42) == 42i;
}

output foo;

to_u64

to_u64( <expression> ) -> U64

Converts the specified expression to the U64 type without regard to under/overflow.

Example:

section foo {
    assert 0xFFFF_FFFF_FFFF_FFFF == to_u64(-1);
    assert to_u64(42i) == 42;
    assert to_u64(42i) == 42u;
    assert to_u64(42) == 42u;
}

output foo;

wr

The wr command has two forms. The first form writes the contents of another section into the current section. The second wr form invokes an extension and writes the output into the current section.

wr section

wr <section identifier>;

Brink adds the specified in section to the current section at the current section offset.

wr extension

wr <namespace>::<extension_name>(<arg1>, <arg2>, ...);

Evaluates the specified extension call and writes the result to the output. The extension's .size() method specifies the size of the result. See Brink Extensions for more information.

Example

Using wr, you can build complex outputs by composing smaller, modular sections together.

Example:

section header {
    wrs "FILE";   // Write a string.
    wr8 0x01;     // Write a byte.
}

section data {
    wrs "DATA";
    wr8 0xFF, 4;
}

// Compose the top-level section
section my_firmware {
    wr header;
    wr data;
    // Use an extension to append a CRC to a section.
    // Extensions can refer to their containing section.
    wr std::crc32c(my_firmware);
}

output my_firmware;

wr8 to wr64

wr8 <expression> [, <expression>]; wr16 <expression> [, <expression>]; wr24 <expression> [, <expression>]; wr32 <expression> [, <expression>]; wr40 <expression> [, <expression>]; wr48 <expression> [, <expression>]; wr56 <expression> [, <expression>]; wr64 <expression> [, <expression>];

Evaluates the first expression and writes the result as a little-endian binary value to the output file. The optional second expression specifies the repetition count.

[!IMPORTANT] Brink silently truncates the upper bits of the expression to fit the specified width.

Example:

// Test expressions in wrx; addr(foo) == 10 as set by region TOP
region TOP { addr = 10; size = 64K; }
section foo in TOP {
    wr8  (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 0 + 10 + 36  = 49
    wr16 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 1 + 10 + 36  = 50 00
    wr24 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 3 + 10 + 36  = 52 00 00
    wr32 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 6 + 10 + 36  = 55 00 00 00
    wr40 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 10 + 10 + 36 = 59 00 00 00 00
    wr48 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 15 + 10 + 36 = 64 00 00 00 00 00
    wr56 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 21 + 10 + 36 = 70 00 00 00 00 00 00
    wr64 (1 + 2) + file_offset() + addr(foo) + sizeof(foo); // 3 + 28 + 10 + 36 = 77 00 00 00 00 00 00 00
    assert sizeof(foo) == 36;
}

output foo;

Another example using the optional repetition expression.

section foo {
    wr32 0x12345678, 10; // write 0x12345678 10 times to the output file.
    wr8 0, addr() % 4096; // write zero enough times to align to 4KB boundary.
}

`wrf "<quoted file path>";`

Write the file at the specified path into the output file. Brink treats all input files as binary files. Paths can be relative to the current directory or absolute.

For example, given the file test_source_1.txt containing:

Hello!

The following program simply copies these 6 UTF-8 characters to the output file.

section foo {
    wrf "test_source_1.txt"; // Hello!
    assert sizeof(foo) == 6;
}

output foo;

`wrs <expression> [, <expression>, ...];`

Evaluates the comma separated list of expressions and writes the resulting string to the output file. Wrs accepts the same expressions and operates similarly to the print statement. For more information, see print.

The wrs statement does not implicitly write a terminating 0 byte after the string. Users creating null terminated (C style) strings in an output file should add an explicit \0.

wrs "my null terminated string\0";

Built-in Variables

Brink pre-defines built-in identifiers that begin with __ (double underscore). They can appear in any expression context that accepts the corresponding type. As shown in the table below, some builtins cannot be used in const expressions because their values depend on dynamic layout values.

Variable	Type	OK in `const`?	Description
`__OUTPUT_SIZE`	`U64`	No	Total output size in bytes. Equivalent to `sizeof(<output-section>)`.
`__OUTPUT_ADDR`	`U64`	No	Address of output section at SectionStart. Equivalent to `addr(<output-section>)`.
`__BRINK_VERSION_STRING`	`String`	Yes	Brink version as a string, e.g. `"4.3.2"`.
`__BRINK_VERSION_MAJOR`	`U64`	Yes	Major version component, e.g. "4" in "4.3.2"
`__BRINK_VERSION_MINOR`	`U64`	Yes	Minor version component, e.g. "3" in "4.3.2"
`__BRINK_VERSION_PATCH`	`U64`	Yes	Patch version component, e.g. "2" in "4.3.2"

`__OUTPUT_SIZE`

Returns the total size of the output file in bytes.

Example — write a 4-byte header field containing the total output size:

section payload {
    wrs "Hello";
}

section hdr {
    wr32 __OUTPUT_SIZE;  // filled with total image size at link time
}

section image {
    wr hdr;
    wr payload;
    assert __OUTPUT_SIZE == sizeof(image);  // equivalent forms
}

output image;

`__OUTPUT_ADDR`

Returns the absolute starting address of the output section. Equivalent to addr(<output-section>). Without placing the section in region, __OUTPUT_ADDR is zero regardless of set_addr command internal to the output section. This occurs because set_addr is a scoped operation. A set_addr within a section affects address calculations for subsequent writes internal to the section, not the logical start of the section itself.

If user places the output section in a region, then __OUTPUT_ADDR is the starting address of the region.

Example — embed the output base address in a table:

region TOP { addr = 0x0800_0000; size = 64K; }

section vtable {
    wr32 __OUTPUT_ADDR;  // base address of the output image
}

section code {
    wrs "code";
}

section image in TOP {
    wr vtable;
    wr code;
    assert __OUTPUT_ADDR == addr(image);  // equivalent expressions
}

output image;

`__BRINK_VERSION_STRING`

Returns the Brink tool version as a string (e.g. "4.0.0"). The value is fixed at compile time and may be used in const expressions, wrs, and print.

Example — stamp the tool version into a firmware header:

section hdr {
    wrs __BRINK_VERSION_STRING;
}

section image {
    wr hdr;
    wrs "payload";
}

output image;

`__BRINK_VERSION_MAJOR`, `__BRINK_VERSION_MINOR`, `__BRINK_VERSION_PATCH`

Return the individual numeric components of the Brink version as U64 values. All three are fixed at compile time and may be used in const expressions and arithmetic.

Example — pack the version into a 3-byte field and assert the tool is new enough:

const MIN_MAJOR = 4u;

section hdr {
    assert __BRINK_VERSION_MAJOR >= MIN_MAJOR;
    wr8 __BRINK_VERSION_MAJOR;
    wr8 __BRINK_VERSION_MINOR;
    wr8 __BRINK_VERSION_PATCH;
}

section image {
    wr hdr;
    wrs "payload";
}

output image;

Brink Extensions

Brink supports compile time extensions to simplify the addition of new functionality. This extension capability enables user defined hashing, compression, validation and other binary data processing tasks. The following sections describe how extensions work and how to create them.

The command line option --list-extensions outputs the names of all available extensions as enabled by Cargo feature flags.

Extensions Are A Compile-Time Feature

Extensions build and link to Brink at compile time as controlled by Cargo feature flags. Because Rust does not guarantee a stable ABI between versions, Brink requires compile time construction to eliminate ABI incompatibilities and enable the use of safe Rust. The following bullets provide an overview of how extensions work:

Extensions interact with Brink through the BrinkExtension trait.
Extensions can read directly from the output buffer for a specified section via zero-copy and safe-memory slices (&[u8]).
In addition to output buffer access, extensions can have their own input parameters like a normal function call.
Extensions are identified by a name in a namespace. Brink reserves the namespaces std and brink.
Extensions report their fixed length binary footprint by implementing the .size() trait method. Brink calls each extension's .size() method exactly once during output layout calculations and caches the result. Brink always passes a mutable output slice (&mut [u8]) of the reported size to the extension's .generate() method.
Extensions register themselves at compile time in Brink's internal extension registry.
The BrinkExtension trait interface allows extensions to return logging and error diagnostics integrated with Brink's own diagnostic output. See []

Invoking Extensions

Users invoke extensions using function-style syntax. Users creating their own extension can take any number of parameters of any Brink support type:

turbo::boost("Big", 1, -42, 0x12345678);

Fixed-size write commands like wr32 are invalid for extensions. If the designer needs to pad the extension's output to a specific size, they must follow the wr command with a pad_sec_offset or align statement.

Passing Section Data to Extensions

Users can pass the data in a section to an extension by passing the section name as a parameter. Extensions take section data as an immutable zero-copy slice parameter of Rust type &[u8]. Section data passed to the extension at the time of the call includes all data generated by non-extension write commands. Furthermore, the data includes the output of extensions executed before the current extension.

As an example, consider the std::crc32c extension. This extension generates a CRC hash over the data provided by the specified section. The extension produces a 4-byte output.

section foo_binary {
    wrf "foo.bin"; // Write the file foo.bin in this section.
}

section bar {
    wr foo_binary;
    // Write the CRC hash of everything in foo_binary
    wr std::crc32c(foo_binary);
}

Users can also pass the section containing the extensions own output to the extension. The extension receives a slice of the full size of the section, including the size of the extension's own output. On input, the slice contains zero bytes at the location of the extension's future output. For example:

section foo_binary {
    wrf "foo.bin";
    // Warning, the CRC input data is the full length of the foo_binary
    // section and includes 4 trailing zero bytes in place of the
    // extension's output.
    wr std::crc32c(foo_binary);
}

section bar {
    wr foo_binary;
}

Named Parameters

To help eliminate bugs, Brink extensions support named parameters. Extensions define their parameter names when registered. In the example below, we call the extension custom::my_extension passing it the required parameters data_section and code_section. The compiler passes the values by name in the order expected by the extension, regardless of the order given at the call site.

//
// extension example
//
section my_data {
    wrf "cool_data.bin";
};

section my_code {
    wrf "cool_code.bin";
};

section stuff {
    wr my_data;
    wr my_code;
    // Use named arguments to avoid positional and semantic bugs!
    custom::my_extension(data_section=my_data, code_section=my_code);
};

Size of Extension Output

Users can query the size of an extension's output using the sizeof command. For example:

assert sizeof(std::crc32c) == 4;

Creating and Registering a New Extension

Extensions register through the extensions crate (extensions/src/lib.rs). process.rs calls extensions::register_all once at startup; adding an extension requires no changes outside extensions/.

Step 1 — Create the extension crate

Place new extensions under std/ for proposed standard library extensions, or under a workspace path matching your namespace for third-party extensions. Implement the BrinkExtension trait from the brink_extension crate.

// my_extension/src/lib.rs
use brink_extension::BrinkExtension;
use extension_registry::ExtensionRegistry;

pub struct MyExtension;

impl BrinkExtension for MyExtension {
    fn name(&self) -> &str { "my_ns::my_ext" }
    fn size(&self) -> usize { 4 }
    fn execute(&self, _args: &[u64], img: &[u8], out: &mut [u8]) -> Result<(), String> {
        // write 4 bytes into out
        Ok(())
    }
}

pub fn register(registry: &mut ExtensionRegistry) {
    registry.register_ranged(Box::new(MyExtension));
}

Step 2 — Add the crate to the workspace

In the root Cargo.toml, add the crate path to [workspace] members.

Step 3 — Wire into `extensions/`

In extensions/Cargo.toml, add the new crate as a dependency:

my_extension = { path = "../my_extension" }

In extensions/src/lib.rs, call its register function inside register_all:

pub fn register_all(registry: &mut ExtensionRegistry) {
    std_crc32c::register(registry);
    my_extension::register(registry);  // add this line
}

Step 4 — Add tests

Create a tests/ directory in your extension crate with .brink scripts and an integration.rs test file. Use CARGO_MANIFEST_DIR to locate .brink files relative to the workspace root — see std/crc32c/tests/integration.rs for a complete example.

Run the extension's tests with:

cargo test -p my_extension

Brink Development

This section provides notes for developers interested in contributing to Brink.

Unit Testing

Brink relies on 100's of unit tests to catch bugs. You can run these with:

cargo test --all

Fuzz Testing

Brink supports fuzz tests for several of its internal libraries. Fuzz testing starts from a corpus of random inputs and then further randomizes those inputs to try to cause crashes and hangs. At the time of writing, fuzz testing requires the nightly build. See fuzz_help.md in the source repo for more information.

Checking Test Code Coverage

If you're using Windows as a development platform, then this worked for me to install the llvm-cov tool. I have the free version of Microsoft Visual Studio installed.

rustup component add llvm-tools
cargo install cargo-llvm-cov --locked

To generate an ASCII table of coverage stats to the terminal:

cargo llvm-cov --all-features --workspace

To update the coverage table in this README from Windows, run .\update_coverage.ps1.

Filename                                     Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover    Branches   Missed Branches     Cover
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
ast\ast.rs                                      2318               452    80.50%          70                 8    88.57%        1265               208    83.56%           0                 0         -
ast\lexer.rs                                     376                12    96.81%          16                 0   100.00%         234                 9    96.15%           0                 0         -
brink_extension\lib.rs                             3                 0   100.00%           1                 0   100.00%           3                 0   100.00%           0                 0         -
const_eval\const_eval.rs                        1081               174    83.90%          26                 5    80.77%         701               150    78.60%           0                 0         -
diags\diags.rs                                   209                25    88.04%          10                 1    90.00%         106                19    82.08%           0                 0         -
engine\engine.rs                                2292               483    78.93%          66                 6    90.91%        1480               271    81.69%           0                 0         -
extension_registry\extension_registry.rs         248                 9    96.37%          17                 3    82.35%         121                 9    92.56%           0                 0         -
extension_registry\test_mocks.rs                 259                34    86.87%          38                 6    84.21%         204                33    83.82%           0                 0         -
extensions\src\lib.rs                              8                 0   100.00%           1                 0   100.00%           5                 0   100.00%           0                 0         -
ir\ir.rs                                         226                28    87.61%          22                 1    95.45%         172                18    89.53%           0                 0         -
irdb\irdb.rs                                     818                97    88.14%          17                 1    94.12%         482                60    87.55%           0                 0         -
layoutdb\layoutdb.rs                             772               160    79.27%          18                 0   100.00%         434                72    83.41%           0                 0         -
linearizer\linearizer.rs                         568                87    84.68%          19                 1    94.74%         332                38    88.55%           0                 0         -
map\map.rs                                       860                13    98.49%          58                 0   100.00%         579                 9    98.45%           0                 0         -
process\process.rs                               360                25    93.06%          22                 5    77.27%         195                 9    95.38%           0                 0         -
prune\prune.rs                                   150                 9    94.00%          12                 2    83.33%          95                 7    92.63%           0                 0         -
src\main.rs                                      123                 6    95.12%           8                 1    87.50%          84                 5    94.05%           0                 0         -
std\crc32c\src\crc32c.rs                          31                 2    93.55%           5                 0   100.00%          26                 3    88.46%           0                 0         -
std\md5\src\md5.rs                                31                 2    93.55%           5                 0   100.00%          26                 3    88.46%           0                 0         -
std\sha256\src\sha256.rs                          31                 2    93.55%           5                 0   100.00%          26                 3    88.46%           0                 0         -
symtable\symtable.rs                             114                 5    95.61%          14                 2    85.71%          80                 5    93.75%           0                 0         -
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                                          10878              1625    85.06%         450                42    90.67%        6650               931    86.00%           0                 0         -

Brink Source Code Overview

File	Stage	Summary
ast/ast.rs	Stage 1	Hand-rolled lexer -> token stream -> arena AST -> AstDb validation
const_eval/const_eval.rs	Stage 2	Lowers const and region AST statements to LinIR, returns SymbolTable and RegionBindings
prune/prune.rs	Stage 3	Eliminates if/else nodes from the AST; promotes sections from the taken branch
layoutdb/layoutdb.rs	Stage 4	AST flattening into linear IR and operand vectors; values are still strings
irdb/irdb.rs	Stage 5	String to typed value conversion, operand and file validation
layout_phase/layout_phase.rs	Stage 6	Iterative address resolution and section footprint calculation
validation_phase/validation_phase.rs	Stage 7	Evaluates all assert instructions after layout and before binary output
exec_phase/exec_phase.rs	Stage 8	Writes inline data, padding, file contents, and extension output to binary
symtable/symtable.rs	Shared types	SymbolTable tracking every compile-time const from declaration through use
linearizer/linearizer.rs	Shared types	LinIR and LinOperand types; shared lowering infrastructure for stages 2 and 4
ir/ir.rs	Shared types	IRKind, ParameterValue, IROperand, IR — the data flowing between stages 4–8
locationdb/locationdb.rs	Shared types	LocationDb and Location produced by stage 6 and consumed by stages 7 and 8
map_phase/map_phase.rs	Map output	Builds MapDb from LocationDb and IRDb; renders map to CSV, JSON, C99, and RS
process/process.rs	Orchestrator	Orchestration of all stages, parses `-D` defines, opens the output file
diags/diags.rs	Cross-cutting	Ariadne-backed diagnostic output channel used by every stage
extensions/src/lib.rs	Extensions	Single registration point for all extensions
brink_extension/lib.rs	Extensions	Public API for extension authors
extension_registry/extension_registry.rs	Extensions	Runtime extension registry and dispatch wrapper
std/crc32c/src/lib.rs	std extension	CRC-32C (Castagnoli) hash over caller-specified output region
std/sha256/src/lib.rs	std extension	SHA256 hash over caller-specified output region

Rebuilding the vscode Syntax Highlighting Extension

Rebuilding the extension require Node.js. After you install Node.js, you may need to restart your command prompt.

Building the extension requires vsce. One time, you'll need to use npm to install vsce

npm install -g @vscode/vsce

Now you're ready to rebuild the extension.

cd vscode-brink
vsce package

To install the extension into vscode locally:

code --install-extension vscode-brink-0.1.0.vsix

Name		Name	Last commit message	Last commit date
Latest commit History 499 Commits
.cargo		.cargo
.claude		.claude
.github/workflows		.github/workflows
.vscode		.vscode
argvaldb		argvaldb
ast		ast
astdb		astdb
brink_extension		brink_extension
const_eval		const_eval
depth_guard		depth_guard
diags		diags
docs		docs
exec_phase		exec_phase
extension_registry		extension_registry
extensions		extensions
images		images
ir		ir
irdb		irdb
layout_phase		layout_phase
layoutdb		layoutdb
linearizer		linearizer
locationdb		locationdb
map		map
map_phase		map_phase
mapdb		mapdb
process		process
prune		prune
regiondb		regiondb
src		src
std		std
symtable		symtable
tests		tests
validation_phase		validation_phase
vscode-brink		vscode-brink
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
Cargo.lock		Cargo.lock
Cargo.toml		Cargo.toml
LICENSE.txt		LICENSE.txt
README.md		README.md
ast.dot		ast.dot
build.rs		build.rs
dist-workspace.toml		dist-workspace.toml
fuzz_help.md		fuzz_help.md
update_coverage.ps1		update_coverage.ps1
update_coverage.py		update_coverage.py

Folders and files

Latest commit

History

Repository files navigation

Brink

Quick Start

Install Prebuilt Binaries for Linux

Install Prebuilt Binaries for Windows

Build From Source

Step 1: Install Rust

Step 2: Clone Brink

Step 3: Build and Run Self-Tests

Step 4: Install Brink

What Can Brink Do?

Hello World

Basic Structure of a Brink Program

Assert and Print

Addresses and Offsets

Brink Disallows Address Overwrites

Brink Disallows Negative Offset Changes

Brink Disallows Address and Offset Overflows

Order of Execution

Output Creation Phases

Command Line Options Reference

Command-Line Const Defines

Value Type Inference

Example

Brink Language Reference

Comments

Whitespace

Semicolon Termination

Types

Identifiers

Reserved Identifiers

Literals

Number Literals

Number Magnitude

True and False

Quoted Strings

Arithmetic Operators

addr

addr_offset

align

assert

const

Deferred Assignment

if/else

include

Labels

output

print

region

Region Properties

Region Property addr

Region Property size

Region Boundary Enforcement

Nested Regions

Partially Overlapping Nested Regions

Sections in Regions are Single Use

sec_offset

section

Sections In Regions

set_addr

pad_addr_offset

pad_file_offset

pad_sec_offset

sizeof

to_i64

to_u64

wr

wr section

wr extension

Example

wr8 to wr64

wrf "<quoted file path>";

wrs <expression> [, <expression>, ...];

Built-in Variables

__OUTPUT_SIZE

__OUTPUT_ADDR

__BRINK_VERSION_STRING

Region Property `addr`

Region Property `size`

`wrf "<quoted file path>";`

`wrs <expression> [, <expression>, ...];`

`__OUTPUT_SIZE`

`__OUTPUT_ADDR`

`__BRINK_VERSION_STRING`

`__BRINK_VERSION_MAJOR`, `__BRINK_VERSION_MINOR`, `__BRINK_VERSION_PATCH`

Step 3 — Wire into `extensions/`

Packages