Extract Function Names from DWARF Debug Info

Topic Overview

DWARF Debug Format

DWARF is the standard debugging data format used by GCC and Clang on Unix-like systems. It stores information about types, variables, functions, and source line numbers inside sections like .debug_info and .debug_line.

Structure and Key Tags

DWARF organizes debug data as a tree of Debugging Information Entries (DIEs):

• DW_TAG_compile_unit: Root of a compilation unit
• DW_TAG_subprogram: A function or subroutine
• DW_TAG_variable: A variable
• DW_AT_name: Name of the entity
• DW_AT_low_pc / DW_AT_high_pc: Function address range

For example, scanning .debug_info for DW_TAG_subprogram entries with DW_AT_name reveals every function name even when the symbol table has been stripped.

For This Exercise

You will implement a simplified DWARF scanner that extracts function names from a .debug_info section. Full DWARF parsing is complex, but this exercise teaches you what GDB uses to map instruction addresses back to source lines.

Line Number Information

Beyond function names, .debug_line maps instruction addresses to source file and line number. GDB uses this mapping when you break main.c:42. The abbreviation table in each compilation unit compresses repeated DIE patterns, which is why raw .debug_info bytes look opaque until you decode the abbrev entries. Even a simplified scanner teaches you what debuggers rely on under the hood.

Your Task

Scan a simplified .debug_info stream the way a debugger does, and list the functions it describes. It keeps DWARF's core mechanics: tags, ULEB128 variable-length integers, names resolved through a string table, and DW_AT_high_pc in offset form.

Every record (a DIE) has the same layout:

Tag 0x11 is DW_TAG_compile_unit, 0x2E is DW_TAG_subprogram, and any other tag (such as 0x34 DW_TAG_variable) is skipped. A name offset past the end of the string table resolves to ?.

Input

1. The string-table length S (1 to 65535), then S bytes in hex. Offset 0 is a NUL byte.
2. The DIE count D (0 to 512).
3. The D records, as one stream of hex bytes.

Output

cCopy
Loading…

then one line per DIE:

• compile unit: CU: main.c
• subprogram, with the C format "FUNC %-12s low=0x%llX high=0x%llX size=%llu\n": FUNC factorial low=0x401180 high=0x4011C0 size=64
• other tags: SKIP tag=0x34 name_off=45, the tag as 2 uppercase hex digits and the offset in decimal

Then Summary: 1 CUs, 3 subprograms, 197 bytes of code (the sum of the subprogram sizes) and, if there was any compile unit, Compile unit: <name of the last one>.