Amiga HUNK File Format – Symbols and Debug

Introduction

Three blocks carry names and line numbers. They sit after a hunk’s contents and relocations, before its HUNK_END. None of them changes what the program does; a loader skips them. HUNK_SYMBOL names places in the hunk, HUNK_EXT (object files) lists what a unit defines and what it needs from others, and HUNK_DEBUG holds whatever a debugger wants, most often line numbers.

HUNK_SYMBOL

Pairs of a name and a value, then a zero long. A name is a count of longs followed by that many longs of text, padded with zero bytes; the value is the long after it, an offset from the start of the hunk. A name count of 0 ends the block. The viewer lists the symbols and uses them as labels in the disassembly: the row at that address carries the name, and a jump or reference to it shows the name. In the real files tested 1,655 symbols occur in the HUNK_SYMBOL blocks of executables; almost every Workbench file is stripped.

HUNK_EXT

A list of entries and a zero long. An entry starts with a long: the symbol type in the top byte and the length of the name, in longs, in the low 24 bits. The name follows, and then, by type:

The numbers are in the vasm assembler’s header file and in the amitools constants. The vasm writer always writes the count of offsets, also after a common size, and the viewer reads it the same way (the amitools reader treats a count of 0 in a reference as 1; the viewer takes the count as written). A type not in the table stops the block and is reported.

HUNK_DEBUG

A count of longs, then that many longs of data whose meaning depends on the tool. AmigaDOS needs only the count. The data usually begins with a base offset and a four-letter tag. The tags the sources name:

A LINE block is: the count of longs N; a base offset; the four letters LINE; the source file name (a count of longs, then the text); then pairs of a line number and an offset from the base, N − 3 − (name longs) longs in all. The vasm writer’s linedebug_hunk writes exactly this, and the real files agree: the loose Amiga files contain two LINE blocks (a library whose source file is lib.e, with 161 pairs, and a program whose source file is Thanks.c, with 192), next to a HEAD block and a block tagged SRC6 that the viewer names without decoding (SRC6 is on none of the lists the sources give). The viewer shows a line table (hunk, file, line, address) for every LINE block.

A HUNK_DEBUG may come before the first block of a hunk (one real program does this); the viewer keeps it with the hunk that follows.

Not described

Known, not yet covered

• Reference: the contents of the debug formats other than LINE (their layouts are on the amiga-dev wiki; only the tag is read here), what SRC6 is, and EXT_RES’s use.
• Viewer: line numbers are listed, not linked to the disassembly rows; symbol and debug data from a.out-style ZMAGIC blocks are not read.

Links