Format:Breakpad symbol file
| Magic | MODULE |
|---|---|
| Extensions | .sym |
| License | BSD-3-Clause (tooling) |
| Website | https://chromium.googlesource.com/breakpad/breakpad/+/master/docs/symbol_files.md |
| On this wiki | Facts · Code · Notes · Topics |
Google Breakpad's text symbol format (.sym): ASCII records MODULE, FILE, FUNC, line records, PUBLIC, STACK CFI and STACK WIN, produced by dump_syms from PDB, DWARF or Mach-O debug info and published by some projects (Mozilla, Chromium) as a way to get function names for a build.
Layout
A Breakpad symbol file is ASCII text, one record per line, fields separated by single spaces (the last field may contain spaces). The first field names the record type, except for line records, which start with a hexadecimal address. Hex numbers have no 0x prefix and use lower case. Source: docs/symbol_files.md in google/breakpad.
MODULE <os> <arch> <id> <name>: first line. os is Linux, mac or windows; arch is x86, x86_64, ppc, ppc64 or unknown (newer dump_syms output may use other values such as arm64; Unverified); id is the debug id (below); name is the base name of the binary (PE: the PDB file name), may contain spaces.FILE <number> <path>: numbers a source file.INLINE_ORIGIN <number> <name>: names an inlined function (newer addition).FUNC [m] <address> <size> <parameter_size> <name>: a function, address relative to the module load address, all hex;mmarks identical-code-folded functions with several names; name is the demangled source-language name.INLINE <nest_level> <call_line> <call_file> <origin> [<address> <size>]+: inlined ranges inside the preceding FUNC.<address> <size> <line> <filenum>: line record for the preceding FUNC (no keyword).PUBLIC [m] <address> <parameter_size> <name>: linker symbol with no size and no line data; the processor prefers FUNC where both cover an address.STACK CFI INIT <address> <size> <reg>: <expr> ...andSTACK CFI <address> <reg>: <expr> ...: DWARF-style call frame information as Breakpad postfix expressions, for example.cfa: $esp 4 + $eip: .cfa 4 - ^.STACK WIN <type> <rva> <code_size> <prologue_size> <epilogue_size> <parameter_size> <saved_register_size> <local_size> <max_stack_size> <has_program_string> <program_or_allocates_base_pointer>: x86 Windows frame data from PDB FPO records; type 4 (FrameData) and 0 (FPO) are used.
So a stripped binary can be given names by writing out these records: FUNC and PUBLIC lines are directly usable as a function name list with RVAs (see Format:PDB (Program Database) for the source of the Windows ones).
Debug id (MODULE id)
The id is uppercase hex with no dashes and ties a binary to its .sym file; the usual store layout is <name>/<id>/<name>.sym (docs/linux_starter_guide.md shows symbols/test/<id>/test.sym).
- PE / PDB: the 16-byte GUID of the CodeView RSDS record written as uppercase hex followed by the PDB age in hex without padding (dump_syms calls GenerateDebugIdentifier(age, guid) in pdb_source_line_writer.cc and pe_util.cc). Old NB10 PDBs use the 32-bit signature instead of a GUID. Portable-PDB (deterministic) PE files with debug-directory minor version 0x504D use the timestamp in place of the age (symbolic, pe.rs).
- ELF: the GNU build id note (first 16 bytes, zero-padded or truncated) when present, otherwise an XOR of the first page (up to 4096 bytes) of
.textinto a 16-byte buffer. On little-endian files the first three GUID fields (4+2+2 bytes) are byte-swapped, so the id is NOT the plain hex of the build id: build id01020304 0506 0708 ...becomes04030201 0605 0807 ..., followed by a0age digit. Source: src/common/linux/file_id.cc (ConvertIdentifierToUUIDString) and symbolic ElfObject::compute_debug_id. - Mach-O: the LC_UUID load command (Breakpad has an MD5-of-text fallback in macho_id.cc, Unverified in detail), plus the
0age digit.
The trailing digit on ELF and Mach-O ids is the age field, always 0 there (Unverified from the reading above; observed in the example D3096ED481217FD4C16B29CD9BC208BA0 in the docs, 33 hex digits).
Using .sym files for reversing
Unverified leads: some projects publish .sym files so anyone can symbolicate crashes. Mozilla's symbol server (symbols.mozilla.org) serves Firefox, Thunderbird and other builds by <debug file>/<debug id>/<debug file>.sym (Unverified: URL scheme from memory, check before relying). Chromium and Sentry, Backtrace and BugSplat accounts keep symbols private or per-customer. To use a .sym for an unknown build: compute the debug id of your binary (rules above), look it up, then turn FUNC/PUBLIC lines into function names at image base + RVA. The names are demangled source names, not linker names.
See also Format:Minidump (the crash dump that carries the module list with these ids), Library:Breakpad, Library:Crashpad.
Code & repositories
- https://github.com/google/breakpad @
94c818ea27482d8a8d32b0c8de50cbca0270db9a· license: BSD-3-Clause (declared by the contributor for the files used; GitHub reports MIT (read from LICENSE; GitHub's API was busy) for the repository)
Breakpad client, dump_syms, minidump_stackwalk and symbol_files.md
Tested on: source read, not run: source read, not run · checked 2026-10-18 - https://github.com/getsentry/symbolic @
bbbd78cd714e282c900d1d19aa75da6b93962e98· license: MIT (declared by the contributor; GitHub's detector did not recognize a license file)
Rust debug-id and Breakpad symbol parsing rules (ELF, PE, Mach-O)
Tested on: source read, not run: source read, not run · checked 2026-10-18