Binary Inspection with objdump and nm

Your Task

nm lists an object file's symbols, and the letter next to each one tells you where it lives. Where a variable lands is decided entirely by its declaration: initialized or not, static or not, and whether it is really const. Write the classifier. Read file-scope C declarations (one per line) and print what nm -S would report for the object file, plus the section totals.

Input

One declaration per line. Blank lines and // comment lines are skipped. Supported forms:

cCopy
Loading…

• TYPE: char (1 byte), short (2), int (4), long (8), float (4), double (8), or void (only behind * or for functions). unsigned/signed alone means int. A pointer is 8 bytes.
• Arrays: [N] has N elements (1..100000). [] takes its size from the initializer: string length + 1 for char arrays, otherwise the element count.
• INIT: a number (C suffixes such as UL allowed), 'c' (\0, \n escapes), a "string", &name, or { e1, e2, ... } of those.

Classification

1. A function definition goes to .text (T). A prototype, or an extern declaration without an initializer, is undefined (U) unless the same name is defined elsewhere in the input, before or after.
2. The object is top-level const if the const applies to the object itself: const int, char *const p, and arrays of const elements. const char *p is not: it is a writable pointer.
3. Top-level const, with an initializer that contains an address (a string literal used as a pointer, or &name): .data.rel.ro (D; the loader must patch it). Other top-level const objects: .rodata (R), even when zero or uninitialized.
4. Otherwise: .data (D) if there is an initializer and any part of it is non-zero, or .bss (B) if it is all zero or missing.
5. static makes the letter lowercase (t r d b).
6. Every string literal used as an address goes to .rodata.str1.1, and identical literals are stored once. A string that initializes a char array is not a literal: it is copied into the array.

Output

cCopy
Loading…

• Section lines appear only when non-empty, in the order above. Byte totals are sums of object sizes (alignment padding is not included).
• Errors are printed as they occur, and the line is otherwise ignored:
  • line N: error: redefinition of NAME (defining a name twice);
  • line N: cannot parse: TEXT (anything outside the forms above, a void object, [] without an initializer, or a string longer than its array).

Example

Input:

cCopy
Loading…

Output:

cCopy
Loading…

Requirements

1. Decide the section from the declaration alone, as the compiler does. Every letter and size in the tests was checked against clang -c for x86-64 Linux followed by nm -S.
2. Merge a prototype or extern declaration with a later definition of the same name.

Hidden tests cover zero-valued initializers of every kind, const zero and uninitialized const objects, .data.rel.ro, merged string literals, &name initializers, redefinitions, prototypes followed by definitions, and unsupported declarations.