CRYSTAL-BUILD(1)

📦
version 1.22.2 · 3 versions
, in arch rolling

💖 NAME 💖

okay bestie, welcome to crystal-build(1) — straight out of the Crystal Compiler Command Line Reference Guide — and it's literally all about how to compile a Crystal program. 🦋✨

✨ SYNOPSIS ✨

run it like this, no cap:

crystal build [options] [programfile] [--] [arguments]

🎀 OPTIONS 🎀

here's the full menu, every single flag, zero skips 💅

  • --cross-compile — generate an object file for cross compilation and print the command to build the executable. The object file should be copied to the target system and the printed command should be executed there. This flag mainly exists for porting the compiler to new platforms; where possible, run the compiler on the target platform directly. 🫶
  • -d, --debug — generate the output with symbolic debug symbols. These are read when debugging the built program with tools like lldb, gdb, valgrind etc. and provide mappings to the original source code for those tools. 🔍
  • --no-debug — generate the output without any symbolic debug symbols. simple as that ✨
  • -D FLAG, --define FLAG — define a compile-time flag. This is useful to conditionally define types, methods, or commands based on flags available at compile time. The default flags are from the target triple given with --target-triple or the hosts default, if none is given. 🎀
  • --emit [asm|llvm-bc|llvm-ir|obj] — comma separated list of types of output for the compiler to emit. You can use this to see the generated LLVM IR, LLVM bitcode, assembly, and object files. 👀
  • --x86-asm-syntax [att|intel] — select the assembly dialect for --emit=asm. The default is att, which stands for the AT&T syntax supported by tools like the GNU Assembler. intel selects the Intel syntax which is preferred for Windows tools. 💻
  • --frame-pointers [auto|always|non-leaf] — control the preservation of frame pointers. The default value, --frame-pointers=auto, will preserve frame pointers on debug builds and try to omit them on release builds (certain platforms require them to stay enabled). --frame-pointers=always will always preserve them, and non-leaf will only force their preservation on non-leaf functions. 🧚
  • -f text|json, --format text|json — format of output. Defaults to text. The json format can be used to get a more parser-friendly output. 🍓
  • --error-trace — show full error trace. Disabled by default, as the full trace usually makes error messages less readable and may not always deliver relevant information. 😭
  • --ll — dump LLVM assembly file to output directory. 🔥
  • --link-flags FLAGS — pass additional flags to the linker. Though you can specify those flags on the source code, this is useful for passing environment specific information directly to the linker, like non-standard library paths or names. For more information on specifying linker flags on source, you can read the "C bindings" section of the documentation available on the official web site. 🔗
  • --mcpu CPU — specify a specific CPU to generate code for. This will pass a -mcpu flag to LLVM, and is only intended to be used for cross-compilation. For a list of available CPUs, pass --mcpu help when building any Crystal source code. Passing --mcpu native will pass the host CPU name to tune performance for the host. 👑
  • --mattr CPU — override or control specific attributes of the target, such as whether SIMD operations are enabled or not. The default set of attributes is set by the current CPU. This will pass a -mattr flag to LLVM, and is only intended to be used for cross-compilation. For a list of available attributes, invoke this exact thing:
llvm-as < /dev/null | llc -march=xyz -mattr=help.`
  • --mcmodel default|kernel|tiny|small|medium|large — specifies a specific code model to generate code for. This will pass a --code-model flag to LLVM. 🌈
  • --no-color — disable colored output. (sad but sometimes necessary 🙈)
  • --no-codegen — don’t do code generation, just parse the file. 💀
  • -o FILE, --output FILE — specify output path. If a directory, the filename is derived from the first source file (default: current directory). 📁
  • --prelude — specify prelude to use. The default one initializes the garbage collector. You can also use --prelude=empty to use no preludes. This can be useful for checking code generation for a specific source code file. 🧹
  • -O LEVEL — optimization mode: 0 (default), 1, 2, 3. See OPTIMIZATIONS for details. ⚡
  • --release — compile in release mode. Equivalent to -O3 --single-module. 🚀
  • -s, --stats — print statistics about the different compiler stages for the current build. Output time and used memory for each compiler process. 📊
  • -p, --progress — print statistics about the progress for the current build. 📈
  • -t, --time — print statistics about the execution time. ⏱️
  • --single-module — generate a single LLVM module. By default, one LLVM module is created for each type in a program. --release implies this option. 🧩
  • --threads NUM — maximum number of threads to use for code generation. The default is 8 threads. 🧵
  • --target TRIPLE — enable target triple; intended to use for cross-compilation. See llvm documentation for more information about target triple. 🎯
  • --verbose — display the commands executed by the system. 🔊
  • --static — create a statically linked executable. 🧱
  • --stdin-filename FILENAME — source file name to be read from STDIN. ⌨️

that's the whole page, bestie — sourced from crystal 1.22.2, dated 2026-10-30. mwah 💋🫶✨

📦 where this page comes from