armasm is a 2 pass assembler that outputs
object code from the assembly language source code. This means that
it reads the source code twice. Each read of the source code is
called a pass.
This is because assembly language source code often contains forward references. A forward reference occurs when a label is used as an operand, for example as a branch target, earlier in the code than the definition of the label. The assembler cannot know the address of the forward reference label until it reads the definition of the label. During each pass, the assembler performs different functions.
During the first pass, the assembler:
Checks the syntax of the instruction or directive. It faults if there is an error in the syntax, for example if a label is specified on a directive that does not accept one.
Determines the size of the instruction and data being assembled and reserves space.
Determines offset of labels within sections.
Creates a symbol table containing label definitions and their memory addresses.
During the second pass, the assembler:
Faults if an undefined reference is specified in an instruction operand or directive.
Encodes the instructions using the label offsets from pass 1, where applicable.
Generates debug information if requested.
Outputs the object file.
Memory addresses of labels are determined and finalized in
the first pass. Therefore, the assembly code must not change during
the second pass. All instructions must be seen in both passes. Therefore
you must not define a symbol after a
:DEF: test for
the symbol. The assembler faults if it sees code in pass 2 that
was not seen in pass 1. Example 1 shows
num EQU 42 is not seen in pass 1 but is seen in
Assembling the code in Example 1 generates the error:
A1903E: Line not seen in first pass; cannot be assembled.
Example 2 shows
MOV r1,r2 is seen in pass 1 but not in pass 2.
Assembling the code in Example 2 generates the error:
A1909E: Line not seen in second pass; cannot be assembled.