* elfcpp.h (NT_VERSION, NT_ARCH): Define as enum constants.

[binutils.git] / gas / doc / internals.texi

\input texinfo
@c  Copyright 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
@c  2001, 2002, 2003, 2004, 2005, 2006
@c  Free Software Foundation, Inc.
@setfilename internals.info
@node Top
@top Assembler Internals
@raisesections
@cindex internals

This chapter describes the internals of the assembler.  It is incomplete, but
it may help a bit.

This chapter is not updated regularly, and it may be out of date.

@menu
* Data types::          Data types
* GAS processing::      What GAS does when it runs
* Porting GAS::         Porting GAS
* Relaxation::          Relaxation
* Broken words::        Broken words
* Internal functions::  Internal functions
* Test suite::          Test suite
@end menu

@node Data types
@section Data types
@cindex internals, data types

This section describes some fundamental GAS data types.

@menu
* Symbols::             The symbolS structure
* Expressions::         The expressionS structure
* Fixups::              The fixS structure
* Frags::               The fragS structure
@end menu

@node Symbols
@subsection Symbols
@cindex internals, symbols
@cindex symbols, internal
@cindex symbolS structure

The definition for the symbol structure, @code{symbolS}, is located in
@file{struc-symbol.h}.

In general, the fields of this structure may not be referred to directly.
Instead, you must use one of the accessor functions defined in @file{symbol.h}.
These accessor functions should work for any GAS version.

Symbol structures contain the following fields:

@table @code
@item sy_value
This is an @code{expressionS} that describes the value of the symbol.  It might
refer to one or more other symbols; if so, its true value may not be known
until @code{resolve_symbol_value} is called with @var{finalize_syms} non-zero
in @code{write_object_file}.

The expression is often simply a constant.  Before @code{resolve_symbol_value}
is called with @var{finalize_syms} set, the value is the offset from the frag
(@pxref{Frags}).  Afterward, the frag address has been added in.

@item sy_resolved
This field is non-zero if the symbol's value has been completely resolved.  It
is used during the final pass over the symbol table.

@item sy_resolving
This field is used to detect loops while resolving the symbol's value.

@item sy_used_in_reloc
This field is non-zero if the symbol is used by a relocation entry.  If a local
symbol is used in a relocation entry, it must be possible to redirect those
relocations to other symbols, or this symbol cannot be removed from the final
symbol list.

@item sy_next
@itemx sy_previous
These pointers to other @code{symbolS} structures describe a doubly
linked list.  These fields should be accessed with
the @code{symbol_next} and @code{symbol_previous} macros.

@item sy_frag
This points to the frag (@pxref{Frags}) that this symbol is attached to.

@item sy_used
Whether the symbol is used as an operand or in an expression.  Note: Not all of
the backends keep this information accurate; backends which use this bit are
responsible for setting it when a symbol is used in backend routines.

@item sy_mri_common
Whether the symbol is an MRI common symbol created by the @code{COMMON}
pseudo-op when assembling in MRI mode.

@item sy_volatile
Whether the symbol can be re-defined.

@item sy_forward_ref
Whether the symbol's value must only be evaluated upon use.

@item sy_weakrefr
Whether the symbol is a @code{weakref} alias to another symbol.

@item sy_weakrefd
Whether the symbol is or was referenced by one or more @code{weakref} aliases,
and has not had any direct references.

@item bsym
This points to the BFD @code{asymbol} that
will be used in writing the object file.

@item sy_obj
This format-specific data is of type @code{OBJ_SYMFIELD_TYPE}.  If no macro by
that name is defined in @file{obj-format.h}, this field is not defined.

@item sy_tc
This processor-specific data is of type @code{TC_SYMFIELD_TYPE}.  If no macro
by that name is defined in @file{targ-cpu.h}, this field is not defined.

@end table

Here is a description of the accessor functions.  These should be used rather
than referring to the fields of @code{symbolS} directly.

@table @code
@item S_SET_VALUE
@cindex S_SET_VALUE
Set the symbol's value.

@item S_GET_VALUE
@cindex S_GET_VALUE
Get the symbol's value.  This will cause @code{resolve_symbol_value} to be
called if necessary.

@item S_SET_SEGMENT
@cindex S_SET_SEGMENT
Set the section of the symbol.

@item S_GET_SEGMENT
@cindex S_GET_SEGMENT
Get the symbol's section.

@item S_GET_NAME
@cindex S_GET_NAME
Get the name of the symbol.

@item S_SET_NAME
@cindex S_SET_NAME
Set the name of the symbol.

@item S_IS_EXTERNAL
@cindex S_IS_EXTERNAL
Return non-zero if the symbol is externally visible.

@item S_IS_EXTERN
@cindex S_IS_EXTERN
A synonym for @code{S_IS_EXTERNAL}.  Don't use it.

@item S_IS_WEAK
@cindex S_IS_WEAK
Return non-zero if the symbol is weak, or if it is a @code{weakref} alias or
symbol that has not been strongly referenced.

@item S_IS_WEAKREFR
@cindex S_IS_WEAKREFR
Return non-zero if the symbol is a @code{weakref} alias.

@item S_IS_WEAKREFD
@cindex S_IS_WEAKREFD
Return non-zero if the symbol was aliased by a @code{weakref} alias and has not
had any strong references.

@item S_IS_VOLATILE
@cindex S_IS_VOLATILE
Return non-zero if the symbol may be re-defined. Such symbols get created by
the @code{=} operator, @code{equ}, or @code{set}.

@item S_IS_FORWARD_REF
@cindex S_IS_FORWARD_REF
Return non-zero if the symbol is a forward reference, that is its value must
only be determined upon use.

@item S_IS_COMMON
@cindex S_IS_COMMON
Return non-zero if this is a common symbol.  Common symbols are sometimes
represented as undefined symbols with a value, in which case this function will
not be reliable.

@item S_IS_DEFINED
@cindex S_IS_DEFINED
Return non-zero if this symbol is defined.  This function is not reliable when
called on a common symbol.

@item S_IS_DEBUG
@cindex S_IS_DEBUG
Return non-zero if this is a debugging symbol.

@item S_IS_LOCAL
@cindex S_IS_LOCAL
Return non-zero if this is a local assembler symbol which should not be
included in the final symbol table.  Note that this is not the opposite of
@code{S_IS_EXTERNAL}.  The @samp{-L} assembler option affects the return value
of this function.

@item S_SET_EXTERNAL
@cindex S_SET_EXTERNAL
Mark the symbol as externally visible.

@item S_CLEAR_EXTERNAL
@cindex S_CLEAR_EXTERNAL
Mark the symbol as not externally visible.

@item S_SET_WEAK
@cindex S_SET_WEAK
Mark the symbol as weak.

@item S_SET_WEAKREFR
@cindex S_SET_WEAKREFR
Mark the symbol as the referrer in a @code{weakref} directive.  The symbol it
aliases must have been set to the value expression before this point.  If the
alias has already been used, the symbol is marked as used too.

@item S_CLEAR_WEAKREFR
@cindex S_CLEAR_WEAKREFR
Clear the @code{weakref} alias status of a symbol.  This is implicitly called
whenever a symbol is defined or set to a new expression.

@item S_SET_WEAKREFD
@cindex S_SET_WEAKREFD
Mark the symbol as the referred symbol in a @code{weakref} directive.
Implicitly marks the symbol as weak, but see below.  It should only be called
if the referenced symbol has just been added to the symbol table.

@item S_SET_WEAKREFD
@cindex S_SET_WEAKREFD
Clear the @code{weakref} aliased status of a symbol.  This is implicitly called
whenever the symbol is looked up, as part of a direct reference or a
definition, but not as part of a @code{weakref} directive.

@item S_SET_VOLATILE
@cindex S_SET_VOLATILE
Indicate that the symbol may be re-defined.

@item S_CLEAR_VOLATILE
@cindex S_CLEAR_VOLATILE
Indicate that the symbol may no longer be re-defined.

@item S_SET_FORWARD_REF
@cindex S_SET_FORWARD_REF
Indicate that the symbol is a forward reference, that is its value must only
be determined upon use.

@item S_GET_TYPE
@item S_GET_DESC
@item S_GET_OTHER
@cindex S_GET_TYPE
@cindex S_GET_DESC
@cindex S_GET_OTHER
Get the @code{type}, @code{desc}, and @code{other} fields of the symbol.  These
are only defined for object file formats for which they make sense (primarily
a.out).

@item S_SET_TYPE
@item S_SET_DESC
@item S_SET_OTHER
@cindex S_SET_TYPE
@cindex S_SET_DESC
@cindex S_SET_OTHER
Set the @code{type}, @code{desc}, and @code{other} fields of the symbol.  These
are only defined for object file formats for which they make sense (primarily
a.out).

@item S_GET_SIZE
@cindex S_GET_SIZE
Get the size of a symbol.  This is only defined for object file formats for
which it makes sense (primarily ELF).

@item S_SET_SIZE
@cindex S_SET_SIZE
Set the size of a symbol.  This is only defined for object file formats for
which it makes sense (primarily ELF).

@item symbol_get_value_expression
@cindex symbol_get_value_expression
Get a pointer to an @code{expressionS} structure which represents the value of
the symbol as an expression.

@item symbol_set_value_expression
@cindex symbol_set_value_expression
Set the value of a symbol to an expression.

@item symbol_set_frag
@cindex symbol_set_frag
Set the frag where a symbol is defined.

@item symbol_get_frag
@cindex symbol_get_frag
Get the frag where a symbol is defined.

@item symbol_mark_used
@cindex symbol_mark_used
Mark a symbol as having been used in an expression.

@item symbol_clear_used
@cindex symbol_clear_used
Clear the mark indicating that a symbol was used in an expression.

@item symbol_used_p
@cindex symbol_used_p
Return whether a symbol was used in an expression.

@item symbol_mark_used_in_reloc
@cindex symbol_mark_used_in_reloc
Mark a symbol as having been used by a relocation.

@item symbol_clear_used_in_reloc
@cindex symbol_clear_used_in_reloc
Clear the mark indicating that a symbol was used in a relocation.

@item symbol_used_in_reloc_p
@cindex symbol_used_in_reloc_p
Return whether a symbol was used in a relocation.

@item symbol_mark_mri_common
@cindex symbol_mark_mri_common
Mark a symbol as an MRI common symbol.

@item symbol_clear_mri_common
@cindex symbol_clear_mri_common
Clear the mark indicating that a symbol is an MRI common symbol.

@item symbol_mri_common_p
@cindex symbol_mri_common_p
Return whether a symbol is an MRI common symbol.

@item symbol_mark_written
@cindex symbol_mark_written
Mark a symbol as having been written.

@item symbol_clear_written
@cindex symbol_clear_written
Clear the mark indicating that a symbol was written.

@item symbol_written_p
@cindex symbol_written_p
Return whether a symbol was written.

@item symbol_mark_resolved
@cindex symbol_mark_resolved
Mark a symbol as having been resolved.

@item symbol_resolved_p
@cindex symbol_resolved_p
Return whether a symbol has been resolved.

@item symbol_section_p
@cindex symbol_section_p
Return whether a symbol is a section symbol.

@item symbol_equated_p
@cindex symbol_equated_p
Return whether a symbol is equated to another symbol.

@item symbol_constant_p
@cindex symbol_constant_p
Return whether a symbol has a constant value, including being an offset within
some frag.

@item symbol_get_bfdsym
@cindex symbol_get_bfdsym
Return the BFD symbol associated with a symbol.

@item symbol_set_bfdsym
@cindex symbol_set_bfdsym
Set the BFD symbol associated with a symbol.

@item symbol_get_obj
@cindex symbol_get_obj
Return a pointer to the @code{OBJ_SYMFIELD_TYPE} field of a symbol.

@item symbol_set_obj
@cindex symbol_set_obj
Set the @code{OBJ_SYMFIELD_TYPE} field of a symbol.

@item symbol_get_tc
@cindex symbol_get_tc
Return a pointer to the @code{TC_SYMFIELD_TYPE} field of a symbol.

@item symbol_set_tc
@cindex symbol_set_tc
Set the @code{TC_SYMFIELD_TYPE} field of a symbol.

@end table

GAS attempts to store local
symbols--symbols which will not be written to the output file--using a
different structure, @code{struct local_symbol}.  This structure can only
represent symbols whose value is an offset within a frag.

Code outside of the symbol handler will always deal with @code{symbolS}
structures and use the accessor functions.  The accessor functions correctly
deal with local symbols.  @code{struct local_symbol} is much smaller than
@code{symbolS} (which also automatically creates a bfd @code{asymbol}
structure), so this saves space when assembling large files.

The first field of @code{symbolS} is @code{bsym}, the pointer to the BFD
symbol.  The first field of @code{struct local_symbol} is a pointer which is
always set to NULL.  This is how the symbol accessor functions can distinguish
local symbols from ordinary symbols.  The symbol accessor functions
automatically convert a local symbol into an ordinary symbol when necessary.

@node Expressions
@subsection Expressions
@cindex internals, expressions
@cindex expressions, internal
@cindex expressionS structure

Expressions are stored in an @code{expressionS} structure.  The structure is
defined in @file{expr.h}.

@cindex expression
The macro @code{expression} will create an @code{expressionS} structure based
on the text found at the global variable @code{input_line_pointer}.

@cindex make_expr_symbol
@cindex expr_symbol_where
A single @code{expressionS} structure can represent a single operation.
Complex expressions are formed by creating @dfn{expression symbols} and
combining them in @code{expressionS} structures.  An expression symbol is
created by calling @code{make_expr_symbol}.  An expression symbol should
naturally never appear in a symbol table, and the implementation of
@code{S_IS_LOCAL} (@pxref{Symbols}) reflects that.  The function
@code{expr_symbol_where} returns non-zero if a symbol is an expression symbol,
and also returns the file and line for the expression which caused it to be
created.

The @code{expressionS} structure has two symbol fields, a number field, an
operator field, and a field indicating whether the number is unsigned.

The operator field is of type @code{operatorT}, and describes how to interpret
the other fields; see the definition in @file{expr.h} for the possibilities.

An @code{operatorT} value of @code{O_big} indicates either a floating point
number, stored in the global variable @code{generic_floating_point_number}, or
an integer too large to store in an @code{offsetT} type, stored in the global
array @code{generic_bignum}.  This rather inflexible approach makes it
impossible to use floating point numbers or large expressions in complex
expressions.

@node Fixups
@subsection Fixups
@cindex internals, fixups
@cindex fixups
@cindex fixS structure

A @dfn{fixup} is basically anything which can not be resolved in the first
pass.  Sometimes a fixup can be resolved by the end of the assembly; if not,
the fixup becomes a relocation entry in the object file.

@cindex fix_new
@cindex fix_new_exp
A fixup is created by a call to @code{fix_new} or @code{fix_new_exp}.  Both
take a frag (@pxref{Frags}), a position within the frag, a size, an indication
of whether the fixup is PC relative, and a type.
The type is nominally a @code{bfd_reloc_code_real_type}, but several
targets use other type codes to represent fixups that can not be described as
relocations.

The @code{fixS} structure has a number of fields, several of which are obsolete
or are only used by a particular target.  The important fields are:

@table @code
@item fx_frag
The frag (@pxref{Frags}) this fixup is in.

@item fx_where
The location within the frag where the fixup occurs.

@item fx_addsy
The symbol this fixup is against.  Typically, the value of this symbol is added
into the object contents.  This may be NULL.

@item fx_subsy
The value of this symbol is subtracted from the object contents.  This is
normally NULL.

@item fx_offset
A number which is added into the fixup.

@item fx_addnumber
Some CPU backends use this field to convey information between
@code{md_apply_fix} and @code{tc_gen_reloc}.  The machine independent code does
not use it.

@item fx_next
The next fixup in the section.

@item fx_r_type
The type of the fixup.

@item fx_size
The size of the fixup.  This is mostly used for error checking.

@item fx_pcrel
Whether the fixup is PC relative.

@item fx_done
Non-zero if the fixup has been applied, and no relocation entry needs to be
generated.

@item fx_file
@itemx fx_line
The file and line where the fixup was created.

@item tc_fix_data
This has the type @code{TC_FIX_TYPE}, and is only defined if the target defines
that macro.
@end table

@node Frags
@subsection Frags
@cindex internals, frags
@cindex frags
@cindex fragS structure.

The @code{fragS} structure is defined in @file{as.h}.  Each frag represents a
portion of the final object file.  As GAS reads the source file, it creates
frags to hold the data that it reads.  At the end of the assembly the frags and
fixups are processed to produce the final contents.

@table @code
@item fr_address
The address of the frag.  This is not set until the assembler rescans the list
of all frags after the entire input file is parsed.  The function
@code{relax_segment} fills in this field.

@item fr_next
Pointer to the next frag in this (sub)section.

@item fr_fix
Fixed number of characters we know we're going to emit to the output file.  May
be zero.

@item fr_var
Variable number of characters we may output, after the initial @code{fr_fix}
characters.  May be zero.

@item fr_offset
The interpretation of this field is controlled by @code{fr_type}.  Generally,
if @code{fr_var} is non-zero, this is a repeat count: the @code{fr_var}
characters are output @code{fr_offset} times.

@item line
Holds line number info when an assembler listing was requested.

@item fr_type
Relaxation state.  This field indicates the interpretation of @code{fr_offset},
@code{fr_symbol} and the variable-length tail of the frag, as well as the
treatment it gets in various phases of processing.  It does not affect the
initial @code{fr_fix} characters; they are always supposed to be output
verbatim (fixups aside).  See below for specific values this field can have.

@item fr_subtype
Relaxation substate.  If the macro @code{md_relax_frag} isn't defined, this is
assumed to be an index into @code{TC_GENERIC_RELAX_TABLE} for the generic
relaxation code to process (@pxref{Relaxation}).  If @code{md_relax_frag} is
defined, this field is available for any use by the CPU-specific code.

@item fr_symbol
This normally indicates the symbol to use when relaxing the frag according to
@code{fr_type}.

@item fr_opcode
Points to the lowest-addressed byte of the opcode, for use in relaxation.

@item tc_frag_data
Target specific fragment data of type TC_FRAG_TYPE.
Only present if @code{TC_FRAG_TYPE} is defined.

@item fr_file
@itemx fr_line
The file and line where this frag was last modified.

@item fr_literal
Declared as a one-character array, this last field grows arbitrarily large to
hold the actual contents of the frag.
@end table

These are the possible relaxation states, provided in the enumeration type
@code{relax_stateT}, and the interpretations they represent for the other
fields:

@table @code
@item rs_align
@itemx rs_align_code
The start of the following frag should be aligned on some boundary.  In this
frag, @code{fr_offset} is the logarithm (base 2) of the alignment in bytes.
(For example, if alignment on an 8-byte boundary were desired, @code{fr_offset}
would have a value of 3.)  The variable characters indicate the fill pattern to
be used.  The @code{fr_subtype} field holds the maximum number of bytes to skip
when doing this alignment.  If more bytes are needed, the alignment is not
done.  An @code{fr_subtype} value of 0 means no maximum, which is the normal
case.  Target backends can use @code{rs_align_code} to handle certain types of
alignment differently.

@item rs_broken_word
This indicates that ``broken word'' processing should be done (@pxref{Broken
words}).  If broken word processing is not necessary on the target machine,
this enumerator value will not be defined.

@item rs_cfa
This state is used to implement exception frame optimizations.  The
@code{fr_symbol} is an expression symbol for the subtraction which may be
relaxed.  The @code{fr_opcode} field holds the frag for the preceding command
byte.  The @code{fr_offset} field holds the offset within that frag.  The
@code{fr_subtype} field is used during relaxation to hold the current size of
the frag.

@item rs_fill
The variable characters are to be repeated @code{fr_offset} times.  If
@code{fr_offset} is 0, this frag has a length of @code{fr_fix}.  Most frags
have this type.

@item rs_leb128
This state is used to implement the DWARF ``little endian base 128''
variable length number format.  The @code{fr_symbol} is always an expression
symbol, as constant expressions are emitted directly.  The @code{fr_offset}
field is used during relaxation to hold the previous size of the number so
that we can determine if the fragment changed size.

@item rs_machine_dependent
Displacement relaxation is to be done on this frag.  The target is indicated by
@code{fr_symbol} and @code{fr_offset}, and @code{fr_subtype} indicates the
particular machine-specific addressing mode desired.  @xref{Relaxation}.

@item rs_org
The start of the following frag should be pushed back to some specific offset
within the section.  (Some assemblers use the value as an absolute address; GAS
does not handle final absolute addresses, but rather requires that the linker
set them.)  The offset is given by @code{fr_symbol} and @code{fr_offset}; one
character from the variable-length tail is used as the fill character.
@end table

@cindex frchainS structure
A chain of frags is built up for each subsection.  The data structure
describing a chain is called a @code{frchainS}, and contains the following
fields:

@table @code
@item frch_root
Points to the first frag in the chain.  May be NULL if there are no frags in
this chain.
@item frch_last
Points to the last frag in the chain, or NULL if there are none.
@item frch_next
Next in the list of @code{frchainS} structures.
@item frch_seg
Indicates the section this frag chain belongs to.
@item frch_subseg
Subsection (subsegment) number of this frag chain.
@item fix_root, fix_tail
Point to first and last @code{fixS} structures associated with this subsection.
@item frch_obstack
Not currently used.  Intended to be used for frag allocation for this
subsection.  This should reduce frag generation caused by switching sections.
@item frch_frag_now
The current frag for this subsegment.
@end table

A @code{frchainS} corresponds to a subsection; each section has a list of
@code{frchainS} records associated with it.  In most cases, only one subsection
of each section is used, so the list will only be one element long, but any
processing of frag chains should be prepared to deal with multiple chains per
section.

After the input files have been completely processed, and no more frags are to
be generated, the frag chains are joined into one per section for further
processing.  After this point, it is safe to operate on one chain per section.

The assembler always has a current frag, named @code{frag_now}.  More space is
allocated for the current frag using the @code{frag_more} function; this
returns a pointer to the amount of requested space.  The function
@code{frag_room} says by how much the current frag can be extended.
Relaxing is done using variant frags allocated by @code{frag_var}
or @code{frag_variant} (@pxref{Relaxation}).

@node GAS processing
@section What GAS does when it runs
@cindex internals, overview

This is a quick look at what an assembler run looks like.

@itemize @bullet
@item
The assembler initializes itself by calling various init routines.

@item
For each source file, the @code{read_a_source_file} function reads in the file
and parses it.  The global variable @code{input_line_pointer} points to the
current text; it is guaranteed to be correct up to the end of the line, but not
farther.

@item
For each line, the assembler passes labels to the @code{colon} function, and
isolates the first word.  If it looks like a pseudo-op, the word is looked up
in the pseudo-op hash table @code{po_hash} and dispatched to a pseudo-op
routine.  Otherwise, the target dependent @code{md_assemble} routine is called
to parse the instruction.

@item
When pseudo-ops or instructions output data, they add it to a frag, calling
@code{frag_more} to get space to store it in.

@item
Pseudo-ops and instructions can also output fixups created by @code{fix_new} or
@code{fix_new_exp}.

@item
For certain targets, instructions can create variant frags which are used to
store relaxation information (@pxref{Relaxation}).

@item
When the input file is finished, the @code{write_object_file} routine is
called.  It assigns addresses to all the frags (@code{relax_segment}), resolves
all the fixups (@code{fixup_segment}), resolves all the symbol values (using
@code{resolve_symbol_value}), and finally writes out the file.
@end itemize

@node Porting GAS
@section Porting GAS
@cindex porting

Each GAS target specifies two main things: the CPU file and the object format
file.  Two main switches in the @file{configure.in} file handle this.  The
first switches on CPU type to set the shell variable @code{cpu_type}.  The
second switches on the entire target to set the shell variable @code{fmt}.

The configure script uses the value of @code{cpu_type} to select two files in
the @file{config} directory: @file{tc-@var{CPU}.c} and @file{tc-@var{CPU}.h}.
The configuration process will create a file named @file{targ-cpu.h} in the
build directory which includes @file{tc-@var{CPU}.h}.

The configure script also uses the value of @code{fmt} to select two files:
@file{obj-@var{fmt}.c} and @file{obj-@var{fmt}.h}.  The configuration process
will create a file named @file{obj-format.h} in the build directory which
includes @file{obj-@var{fmt}.h}.

You can also set the emulation in the configure script by setting the @code{em}
variable.  Normally the default value of @samp{generic} is fine.  The
configuration process will create a file named @file{targ-env.h} in the build
directory which includes @file{te-@var{em}.h}.

There is a special case for COFF. For historical reason, the GNU COFF
assembler doesn't follow the documented behavior on certain debug symbols for
the compatibility with other COFF assemblers. A port can define
@code{STRICTCOFF} in the configure script to make the GNU COFF assembler
to follow the documented behavior.

Porting GAS to a new CPU requires writing the @file{tc-@var{CPU}} files.
Porting GAS to a new object file format requires writing the
@file{obj-@var{fmt}} files.  There is sometimes some interaction between these
two files, but it is normally minimal.

The best approach is, of course, to copy existing files.  The documentation
below assumes that you are looking at existing files to see usage details.

These interfaces have grown over time, and have never been carefully thought
out or designed.  Nothing about the interfaces described here is cast in stone.
It is possible that they will change from one version of the assembler to the
next.  Also, new macros are added all the time as they are needed.

@menu
* CPU backend::                 Writing a CPU backend
* Object format backend::       Writing an object format backend
* Emulations::                  Writing emulation files
@end menu

@node CPU backend
@subsection Writing a CPU backend
@cindex CPU backend
@cindex @file{tc-@var{CPU}}

The CPU backend files are the heart of the assembler.  They are the only parts
of the assembler which actually know anything about the instruction set of the
processor.

You must define a reasonably small list of macros and functions in the CPU
backend files.  You may define a large number of additional macros in the CPU
backend files, not all of which are documented here.  You must, of course,
define macros in the @file{.h} file, which is included by every assembler
source file.  You may define the functions as macros in the @file{.h} file, or
as functions in the @file{.c} file.

@table @code
@item TC_@var{CPU}
@cindex TC_@var{CPU}
By convention, you should define this macro in the @file{.h} file.  For
example, @file{tc-m68k.h} defines @code{TC_M68K}.  You might have to use this
if it is necessary to add CPU specific code to the object format file.

@item TARGET_FORMAT
This macro is the BFD target name to use when creating the output file.  This
will normally depend upon the @code{OBJ_@var{FMT}} macro.

@item TARGET_ARCH
This macro is the BFD architecture to pass to @code{bfd_set_arch_mach}.

@item TARGET_MACH
This macro is the BFD machine number to pass to @code{bfd_set_arch_mach}.  If
it is not defined, GAS will use 0.

@item TARGET_BYTES_BIG_ENDIAN
You should define this macro to be non-zero if the target is big endian, and
zero if the target is little endian.

@item md_shortopts
@itemx md_longopts
@itemx md_longopts_size
@itemx md_parse_option
@itemx md_show_usage
@itemx md_after_parse_args
@cindex md_shortopts
@cindex md_longopts
@cindex md_longopts_size
@cindex md_parse_option
@cindex md_show_usage
@cindex md_after_parse_args
GAS uses these variables and functions during option processing.
@code{md_shortopts} is a @code{const char *} which GAS adds to the machine
independent string passed to @code{getopt}.  @code{md_longopts} is a
@code{struct option []} which GAS adds to the machine independent long options
passed to @code{getopt}; you may use @code{OPTION_MD_BASE}, defined in
@file{as.h}, as the start of a set of long option indices, if necessary.
@code{md_longopts_size} is a @code{size_t} holding the size @code{md_longopts}.

GAS will call @code{md_parse_option} whenever @code{getopt} returns an
unrecognized code, presumably indicating a special code value which appears in
@code{md_longopts}.  This function should return non-zero if it handled the
option and zero otherwise.  There is no need to print a message about an option
not being recognized.  This will be handled by the generic code.

GAS will call @code{md_show_usage} when a usage message is printed; it should
print a description of the machine specific options. @code{md_after_pase_args},
if defined, is called after all options are processed, to let the backend
override settings done by the generic option parsing.

@item md_begin
@cindex md_begin
GAS will call this function at the start of the assembly, after the command
line arguments have been parsed and all the machine independent initializations
have been completed.

@item md_cleanup
@cindex md_cleanup
If you define this macro, GAS will call it at the end of each input file.

@item md_assemble
@cindex md_assemble
GAS will call this function for each input line which does not contain a
pseudo-op.  The argument is a null terminated string.  The function should
assemble the string as an instruction with operands.  Normally
@code{md_assemble} will do this by calling @code{frag_more} and writing out
some bytes (@pxref{Frags}).  @code{md_assemble} will call @code{fix_new} to
create fixups as needed (@pxref{Fixups}).  Targets which need to do special
purpose relaxation will call @code{frag_var}.

@item md_pseudo_table
@cindex md_pseudo_table
This is a const array of type @code{pseudo_typeS}.  It is a mapping from
pseudo-op names to functions.  You should use this table to implement
pseudo-ops which are specific to the CPU.

@item tc_conditional_pseudoop
@cindex tc_conditional_pseudoop
If this macro is defined, GAS will call it with a @code{pseudo_typeS} argument.
It should return non-zero if the pseudo-op is a conditional which controls
whether code is assembled, such as @samp{.if}.  GAS knows about the normal
conditional pseudo-ops, and you should normally not have to define this macro.

@item comment_chars
@cindex comment_chars
This is a null terminated @code{const char} array of characters which start a
comment.

@item tc_comment_chars
@cindex tc_comment_chars
If this macro is defined, GAS will use it instead of @code{comment_chars}.

@item tc_symbol_chars
@cindex tc_symbol_chars
If this macro is defined, it is a pointer to a null terminated list of
characters which may appear in an operand.  GAS already assumes that all
alphanumeric characters, and @samp{$}, @samp{.}, and @samp{_} may appear in an
operand (see @samp{symbol_chars} in @file{app.c}).  This macro may be defined
to treat additional characters as appearing in an operand.  This affects the
way in which GAS removes whitespace before passing the string to
@samp{md_assemble}.

@item line_comment_chars
@cindex line_comment_chars
This is a null terminated @code{const char} array of characters which start a
comment when they appear at the start of a line.

@item line_separator_chars
@cindex line_separator_chars
This is a null terminated @code{const char} array of characters which separate
lines (null and newline are such characters by default, and need not be
listed in this array).  Note that line_separator_chars do not separate lines
if found in a comment, such as after a character in line_comment_chars or
comment_chars.

@item EXP_CHARS
@cindex EXP_CHARS
This is a null terminated @code{const char} array of characters which may be
used as the exponent character in a floating point number.  This is normally
@code{"eE"}.

@item FLT_CHARS
@cindex FLT_CHARS
This is a null terminated @code{const char} array of characters which may be
used to indicate a floating point constant.  A zero followed by one of these
characters is assumed to be followed by a floating point number; thus they
operate the way that @code{0x} is used to indicate a hexadecimal constant.
Usually this includes @samp{r} and @samp{f}.

@item LEX_AT
@cindex LEX_AT
You may define this macro to the lexical type of the @kbd{@@} character.  The
default is zero.

Lexical types are a combination of @code{LEX_NAME} and @code{LEX_BEGIN_NAME},
both defined in @file{read.h}.  @code{LEX_NAME} indicates that the character
may appear in a name.  @code{LEX_BEGIN_NAME} indicates that the character may
appear at the beginning of a name.

@item LEX_BR
@cindex LEX_BR
You may define this macro to the lexical type of the brace characters @kbd{@{},
@kbd{@}}, @kbd{[}, and @kbd{]}.  The default value is zero.

@item LEX_PCT
@cindex LEX_PCT
You may define this macro to the lexical type of the @kbd{%} character.  The
default value is zero.

@item LEX_QM
@cindex LEX_QM
You may define this macro to the lexical type of the @kbd{?} character.  The
default value it zero.

@item LEX_DOLLAR
@cindex LEX_DOLLAR
You may define this macro to the lexical type of the @kbd{$} character.  The
default value is @code{LEX_NAME | LEX_BEGIN_NAME}.

@item NUMBERS_WITH_SUFFIX
@cindex NUMBERS_WITH_SUFFIX
When this macro is defined to be non-zero, the parser allows the radix of a
constant to be indicated with a suffix.  Valid suffixes are binary (B),
octal (Q), and hexadecimal (H).  Case is not significant.

@item SINGLE_QUOTE_STRINGS
@cindex SINGLE_QUOTE_STRINGS
If you define this macro, GAS will treat single quotes as string delimiters.
Normally only double quotes are accepted as string delimiters.

@item NO_STRING_ESCAPES
@cindex NO_STRING_ESCAPES
If you define this macro, GAS will not permit escape sequences in a string.

@item ONLY_STANDARD_ESCAPES
@cindex ONLY_STANDARD_ESCAPES
If you define this macro, GAS will warn about the use of nonstandard escape
sequences in a string.

@item md_start_line_hook
@cindex md_start_line_hook
If you define this macro, GAS will call it at the start of each line.

@item LABELS_WITHOUT_COLONS
@cindex LABELS_WITHOUT_COLONS
If you define this macro, GAS will assume that any text at the start of a line
is a label, even if it does not have a colon.

@item TC_START_LABEL
@itemx TC_START_LABEL_WITHOUT_COLON
@cindex TC_START_LABEL
You may define this macro to control what GAS considers to be a label.  The
default definition is to accept any name followed by a colon character.

@item TC_START_LABEL_WITHOUT_COLON
@cindex TC_START_LABEL_WITHOUT_COLON
Same as TC_START_LABEL, but should be used instead of TC_START_LABEL when
LABELS_WITHOUT_COLONS is defined.

@item TC_FAKE_LABEL
@cindex TC_FAKE_LABEL
You may define this macro to control what GAS considers to be a fake
label.  The default fake label is FAKE_LABEL_NAME.

@item NO_PSEUDO_DOT
@cindex NO_PSEUDO_DOT
If you define this macro, GAS will not require pseudo-ops to start with a
@kbd{.} character.

@item TC_EQUAL_IN_INSN
@cindex TC_EQUAL_IN_INSN
If you define this macro, it should return nonzero if the instruction is
permitted to contain an @kbd{=} character.  GAS will call it with two
arguments, the character before the @kbd{=} character, and the value of
the string preceding the equal sign. GAS uses this macro to decide if a
@kbd{=} is an assignment or an instruction.

@item TC_EOL_IN_INSN
@cindex TC_EOL_IN_INSN
If you define this macro, it should return nonzero if the current input line
pointer should be treated as the end of a line.

@item TC_CASE_SENSITIVE
@cindex TC_CASE_SENSITIVE
Define this macro if instruction mnemonics and pseudos are case sensitive.
The default is to have it undefined giving case insensitive names.

@item md_parse_name
@cindex md_parse_name
If this macro is defined, GAS will call it for any symbol found in an
expression.  You can define this to handle special symbols in a special way.
If a symbol always has a certain value, you should normally enter it in the
symbol table, perhaps using @code{reg_section}.

@item md_undefined_symbol
@cindex md_undefined_symbol
GAS will call this function when a symbol table lookup fails, before it
creates a new symbol.  Typically this would be used to supply symbols whose
name or value changes dynamically, possibly in a context sensitive way.
Predefined symbols with fixed values, such as register names or condition
codes, are typically entered directly into the symbol table when @code{md_begin}
is called.  One argument is passed, a @code{char *} for the symbol.

@item md_operand
@cindex md_operand
GAS will call this function with one argument, an @code{expressionS}
pointer, for any expression that can not be recognized.  When the function
is called, @code{input_line_pointer} will point to the start of the
expression.

@item md_register_arithmetic
@cindex md_register_arithmetic
If this macro is defined and evaluates to zero then GAS will not fold
expressions that add or subtract a constant to/from a register to give
another register.  For example GAS's default behaviour is to fold the
expression "r8 + 1" into "r9", which is probably not the result
intended by the programmer.  The default is to allow such folding,
since this maintains backwards compatibility with earlier releases of
GAS.

@item tc_unrecognized_line
@cindex tc_unrecognized_line
If you define this macro, GAS will call it when it finds a line that it can not
parse.

@item md_do_align
@cindex md_do_align
You may define this macro to handle an alignment directive.  GAS will call it
when the directive is seen in the input file.  For example, the i386 backend
uses this to generate efficient nop instructions of varying lengths, depending
upon the number of bytes that the alignment will skip.

@item HANDLE_ALIGN
@cindex HANDLE_ALIGN
You may define this macro to do special handling for an alignment directive.
GAS will call it at the end of the assembly.

@item TC_IMPLICIT_LCOMM_ALIGNMENT (@var{size}, @var{p2var})
@cindex TC_IMPLICIT_LCOMM_ALIGNMENT
An @code{.lcomm} directive with no explicit alignment parameter will use this
macro to set @var{p2var} to the alignment that a request for @var{size} bytes
will have.  The alignment is expressed as a power of two.  If no alignment
should take place, the macro definition should do nothing.  Some targets define
a @code{.bss} directive that is also affected by this macro.  The default
definition will set @var{p2var} to the truncated power of two of sizes up to
eight bytes.

@item md_flush_pending_output
@cindex md_flush_pending_output
If you define this macro, GAS will call it each time it skips any space because of a
space filling or alignment or data allocation pseudo-op.

@item TC_PARSE_CONS_EXPRESSION
@cindex TC_PARSE_CONS_EXPRESSION
You may define this macro to parse an expression used in a data allocation
pseudo-op such as @code{.word}.  You can use this to recognize relocation
directives that may appear in such directives.

@item BITFIELD_CONS_EXPRESSION
@cindex BITFIELD_CONS_EXPRESSION
If you define this macro, GAS will recognize bitfield instructions in data
allocation pseudo-ops, as used on the i960.

@item REPEAT_CONS_EXPRESSION
@cindex REPEAT_CONS_EXPRESSION
If you define this macro, GAS will recognize repeat counts in data allocation
pseudo-ops, as used on the MIPS.

@item md_cons_align
@cindex md_cons_align
You may define this macro to do any special alignment before a data allocation
pseudo-op.

@item TC_CONS_FIX_NEW
@cindex TC_CONS_FIX_NEW
You may define this macro to generate a fixup for a data allocation pseudo-op.

@item TC_ADDRESS_BYTES
@cindex TC_ADDRESS_BYTES
Define this macro to specify the number of bytes used to store an address.
Used to implement @code{dc.a}.  The target must have a reloc for this size.

@item TC_INIT_FIX_DATA (@var{fixp})
@cindex TC_INIT_FIX_DATA
A C statement to initialize the target specific fields of fixup @var{fixp}.
These fields are defined with the @code{TC_FIX_TYPE} macro.

@item TC_FIX_DATA_PRINT (@var{stream}, @var{fixp})
@cindex TC_FIX_DATA_PRINT
A C statement to output target specific debugging information for
fixup @var{fixp} to @var{stream}.  This macro is called by @code{print_fixup}.

@item TC_FRAG_INIT (@var{fragp})
@cindex TC_FRAG_INIT
A C statement to initialize the target specific fields of frag @var{fragp}.
These fields are defined with the @code{TC_FRAG_TYPE} macro.

@item md_number_to_chars
@cindex md_number_to_chars
This should just call either @code{number_to_chars_bigendian} or
@code{number_to_chars_littleendian}, whichever is appropriate.  On targets like
the MIPS which support options to change the endianness, which function to call
is a runtime decision.  On other targets, @code{md_number_to_chars} can be a
simple macro.

@item md_atof (@var{type},@var{litP},@var{sizeP})
@cindex md_atof
This function is called to convert an ASCII string into a floating point value
in format used by the CPU.  It takes three arguments.  The first is @var{type}
which is a byte describing the type of floating point number to be created.  It
is one of the characters defined in the @xref{FLT_CHARS} macro.  Possible
values are @var{'f'} or @var{'s'} for single precision, @var{'d'} or @var{'r'}
for double precision and @var{'x'} or @var{'p'} for extended precision.  Either
lower or upper case versions of these letters can be used.  Note: some targets
do not support all of these types, and some targets may also support other
types not mentioned here.

The second parameter is @var{litP} which is a pointer to a byte array where the
converted value should be stored.  The value is converted into LITTLENUMs and
is stored in the target's endian-ness order.  (@var{LITTLENUM} is defined in
gas/bignum.h).  Single precision values occupy 2 littlenums.  Double precision
values occupy 4 littlenums and extended precision values occupy either 5 or 6
littlenums, depending upon the target.

The third argument is @var{sizeP}, which is a pointer to a integer that should
be filled in with the number of chars emitted into the byte array.

The function should return NULL upon success or an error string upon failure.

@item TC_LARGEST_EXPONENT_IS_NORMAL
@cindex TC_LARGEST_EXPONENT_IS_NORMAL (@var{precision})
This macro is used only by @file{atof-ieee.c}.  It should evaluate to true
if floats of the given precision use the largest exponent for normal numbers
instead of NaNs and infinities.  @var{precision} is @samp{F_PRECISION} for
single precision, @samp{D_PRECISION} for double precision, or
@samp{X_PRECISION} for extended double precision.

The macro has a default definition which returns 0 for all cases.

@item WORKING_DOT_WORD
@itemx md_short_jump_size
@itemx md_long_jump_size
@itemx md_create_short_jump
@itemx md_create_long_jump
@itemx TC_CHECK_ADJUSTED_BROKEN_DOT_WORD
@cindex WORKING_DOT_WORD
@cindex md_short_jump_size
@cindex md_long_jump_size
@cindex md_create_short_jump
@cindex md_create_long_jump
@cindex TC_CHECK_ADJUSTED_BROKEN_DOT_WORD
If @code{WORKING_DOT_WORD} is defined, GAS will not do broken word processing
(@pxref{Broken words}).  Otherwise, you should set @code{md_short_jump_size} to
the size of a short jump (a jump that is just long enough to jump around a
number of long jumps) and @code{md_long_jump_size} to the size of a long jump
(a jump that can go anywhere in the function).  You should define
@code{md_create_short_jump} to create a short jump around a number of long
jumps, and define @code{md_create_long_jump} to create a long jump.
If defined, the macro TC_CHECK_ADJUSTED_BROKEN_DOT_WORD will be called for each
adjusted word just before the word is output.  The macro takes two arguments,
an @code{addressT} with the adjusted word and a pointer to the current
@code{struct broken_word}.

@item md_estimate_size_before_relax
@cindex md_estimate_size_before_relax
This function returns an estimate of the size of a @code{rs_machine_dependent}
frag before any relaxing is done.  It may also create any necessary
relocations.

@item md_relax_frag
@cindex md_relax_frag
This macro may be defined to relax a frag.  GAS will call this with the
segment, the frag, and the change in size of all previous frags;
@code{md_relax_frag} should return the change in size of the frag.
@xref{Relaxation}.

@item TC_GENERIC_RELAX_TABLE
@cindex TC_GENERIC_RELAX_TABLE
If you do not define @code{md_relax_frag}, you may define
@code{TC_GENERIC_RELAX_TABLE} as a table of @code{relax_typeS} structures.  The
machine independent code knows how to use such a table to relax PC relative
references.  See @file{tc-m68k.c} for an example.  @xref{Relaxation}.

@item md_prepare_relax_scan
@cindex md_prepare_relax_scan
If defined, it is a C statement that is invoked prior to scanning
the relax table.

@item LINKER_RELAXING_SHRINKS_ONLY
@cindex LINKER_RELAXING_SHRINKS_ONLY
If you define this macro, and the global variable @samp{linkrelax} is set
(because of a command line option, or unconditionally in @code{md_begin}), a
@samp{.align} directive will cause extra space to be allocated.  The linker can
then discard this space when relaxing the section.

@item TC_LINKRELAX_FIXUP (@var{segT})
@cindex TC_LINKRELAX_FIXUP
If defined, this macro allows control over whether fixups for a
given section will be processed when the @var{linkrelax} variable is
set.  The macro is given the N_TYPE bits for the section in its
@var{segT} argument.  If the macro evaluates to a non-zero value
then the fixups will be converted into relocs, otherwise they will
be passed to @var{md_apply_fix} as normal.

@item md_convert_frag
@cindex md_convert_frag
GAS will call this for each rs_machine_dependent fragment.
The instruction is completed using the data from the relaxation pass.
It may also create any necessary relocations.
@xref{Relaxation}.

@item TC_FINALIZE_SYMS_BEFORE_SIZE_SEG
@cindex TC_FINALIZE_SYMS_BEFORE_SIZE_SEG
Specifies the value to be assigned to @code{finalize_syms} before the function
@code{size_segs} is called.  Since @code{size_segs} calls @code{cvt_frag_to_fill}
which can call @code{md_convert_frag}, this constant governs whether the symbols 
accessed in @code{md_convert_frag} will be fully resolved.  In particular it
governs whether local symbols will have been resolved, and had their frag
information removed.  Depending upon the processing performed by
@code{md_convert_frag} the frag information may or may not be necessary, as may
the resolved values of the symbols.  The default value is 1.

@item TC_VALIDATE_FIX (@var{fixP}, @var{seg}, @var{skip})
@cindex TC_VALIDATE_FIX
This macro is evaluated for each fixup (when @var{linkrelax} is not set).
It may be used to change the fixup in @code{struct fix *@var{fixP}} before
the generic code sees it, or to fully process the fixup.  In the latter case,
a @code{goto @var{skip}} will bypass the generic code.

@item md_apply_fix (@var{fixP}, @var{valP}, @var{seg})
@cindex md_apply_fix
GAS will call this for each fixup that passes the @code{TC_VALIDATE_FIX} test
when @var{linkrelax} is not set.  It should store the correct value in the
object file.  @code{struct fix *@var{fixP}} is the fixup @code{md_apply_fix}
is operating on.  @code{valueT *@var{valP}} is the value to store into the
object files, or at least is the generic code's best guess.  Specifically,
*@var{valP} is the value of the fixup symbol, perhaps modified by
@code{MD_APPLY_SYM_VALUE}, plus @code{@var{fixP}->fx_offset} (symbol addend),
less @code{MD_PCREL_FROM_SECTION} for pc-relative fixups.
@code{segT @var{seg}} is the section the fix is in.
@code{fixup_segment} performs a generic overflow check on *@var{valP} after
@code{md_apply_fix} returns.  If the overflow check is relevant for the target
machine, then @code{md_apply_fix} should modify *@var{valP}, typically to the
value stored in the object file.

@item TC_FORCE_RELOCATION (@var{fix})
@cindex TC_FORCE_RELOCATION
If this macro returns non-zero, it guarantees that a relocation will be emitted
even when the value can be resolved locally, as @code{fixup_segment} tries to
reduce the number of relocations emitted.  For example, a fixup expression
against an absolute symbol will normally not require a reloc.  If undefined,
a default of @w{@code{(S_FORCE_RELOC ((@var{fix})->fx_addsy))}} is used.

@item TC_FORCE_RELOCATION_ABS (@var{fix})
@cindex TC_FORCE_RELOCATION_ABS
Like @code{TC_FORCE_RELOCATION}, but used only for fixup expressions against an
absolute symbol.  If undefined, @code{TC_FORCE_RELOCATION} will be used.

@item TC_FORCE_RELOCATION_LOCAL (@var{fix})
@cindex TC_FORCE_RELOCATION_LOCAL
Like @code{TC_FORCE_RELOCATION}, but used only for fixup expressions against a
symbol in the current section.  If undefined, fixups that are not
@code{fx_pcrel} or for which @code{TC_FORCE_RELOCATION}
returns non-zero, will emit relocs.

@item TC_FORCE_RELOCATION_SUB_SAME (@var{fix}, @var{seg})
@cindex TC_FORCE_RELOCATION_SUB_SAME
This macro controls resolution of fixup expressions involving the
difference of two symbols in the same section.  If this macro returns zero,
the subtrahend will be resolved and @code{fx_subsy} set to @code{NULL} for
@code{md_apply_fix}.  If undefined, the default of
@w{@code{! SEG_NORMAL (@var{seg}) || TC_FORCE_RELOCATION (@var{fix})}} will
be used.

@item TC_FORCE_RELOCATION_SUB_ABS (@var{fix})
@cindex TC_FORCE_RELOCATION_SUB_ABS
Like @code{TC_FORCE_RELOCATION_SUB_SAME}, but used when the subtrahend is an
absolute symbol.  If the macro is undefined a default of @code{0} is used.

@item TC_FORCE_RELOCATION_SUB_LOCAL (@var{fix})
@cindex TC_FORCE_RELOCATION_SUB_LOCAL
Like @code{TC_FORCE_RELOCATION_SUB_ABS}, but the subtrahend is a symbol in the
same section as the fixup.

@item TC_VALIDATE_FIX_SUB (@var{fix})
@cindex TC_VALIDATE_FIX_SUB
This macro is evaluated for any fixup with a @code{fx_subsy} that
@code{fixup_segment} cannot reduce to a number.  If the macro returns
@code{false} an error will be reported.

@item MD_APPLY_SYM_VALUE (@var{fix})
@cindex MD_APPLY_SYM_VALUE
This macro controls whether the symbol value becomes part of the value passed
to @code{md_apply_fix}.  If the macro is undefined, or returns non-zero, the
symbol value will be included.  For ELF, a suitable definition might simply be
@code{0}, because ELF relocations don't include the symbol value in the addend.

@item S_FORCE_RELOC (@var{sym}, @var{strict})
@cindex S_FORCE_RELOC
This function returns true for symbols
that should not be reduced to section symbols or eliminated from expressions,
because they may be overridden by the linker.  ie. for symbols that are
undefined or common, and when @var{strict} is set, weak, or global (for ELF
assemblers that support ELF shared library linking semantics).

@item EXTERN_FORCE_RELOC
@cindex EXTERN_FORCE_RELOC
This macro controls whether @code{S_FORCE_RELOC} returns true for global
symbols.  If undefined, the default is @code{true} for ELF assemblers, and
@code{false} for non-ELF.

@item tc_gen_reloc
@cindex tc_gen_reloc
GAS will call this to generate a reloc.  GAS will pass
the resulting reloc to @code{bfd_install_relocation}.  This currently works
poorly, as @code{bfd_install_relocation} often does the wrong thing, and
instances of @code{tc_gen_reloc} have been written to work around the problems,
which in turns makes it difficult to fix @code{bfd_install_relocation}.

@item RELOC_EXPANSION_POSSIBLE
@cindex RELOC_EXPANSION_POSSIBLE
If you define this macro, it means that @code{tc_gen_reloc} may return multiple
relocation entries for a single fixup.  In this case, the return value of
@code{tc_gen_reloc} is a pointer to a null terminated array.

@item MAX_RELOC_EXPANSION
@cindex MAX_RELOC_EXPANSION
You must define this if @code{RELOC_EXPANSION_POSSIBLE} is defined; it
indicates the largest number of relocs which @code{tc_gen_reloc} may return for
a single fixup.

@item tc_fix_adjustable
@cindex tc_fix_adjustable
You may define this macro to indicate whether a fixup against a locally defined
symbol should be adjusted to be against the section symbol.  It should return a
non-zero value if the adjustment is acceptable.

@item MD_PCREL_FROM_SECTION (@var{fixp}, @var{section})
@cindex MD_PCREL_FROM_SECTION
If you define this macro, it should return the position from which the PC
relative adjustment for a PC relative fixup should be made.  On many
processors, the base of a PC relative instruction is the next instruction,
so this macro would return the length of an instruction, plus the address of
the PC relative fixup.  The latter can be calculated as
@var{fixp}->fx_where + @var{fixp}->fx_frag->fr_address .

@item md_pcrel_from
@cindex md_pcrel_from
This is the default value of @code{MD_PCREL_FROM_SECTION}.  The difference is
that @code{md_pcrel_from} does not take a section argument.

@item tc_frob_label
@cindex tc_frob_label
If you define this macro, GAS will call it each time a label is defined.

@item md_section_align
@cindex md_section_align
GAS will call this function for each section at the end of the assembly, to
permit the CPU backend to adjust the alignment of a section.  The function
must take two arguments, a @code{segT} for the section and a @code{valueT}
for the size of the section, and return a @code{valueT} for the rounded
size.

@item md_macro_start
@cindex md_macro_start
If defined, GAS will call this macro when it starts to include a macro
expansion.  @code{macro_nest} indicates the current macro nesting level, which
includes the one being expanded.

@item md_macro_info
@cindex md_macro_info
If defined, GAS will call this macro after the macro expansion has been
included in the input and after parsing the macro arguments.  The single
argument is a pointer to the macro processing's internal representation of the
macro (macro_entry *), which includes expansion of the formal arguments.

@item md_macro_end
@cindex md_macro_end
Complement to md_macro_start.  If defined, it is called when finished
processing an inserted macro expansion, just before decrementing macro_nest.

@item DOUBLEBAR_PARALLEL
@cindex DOUBLEBAR_PARALLEL
Affects the preprocessor so that lines containing '||' don't have their
whitespace stripped following the double bar.  This is useful for targets that
implement parallel instructions.

@item KEEP_WHITE_AROUND_COLON
@cindex KEEP_WHITE_AROUND_COLON
Normally, whitespace is compressed and removed when, in the presence of the
colon, the adjoining tokens can be distinguished.  This option affects the
preprocessor so that whitespace around colons is preserved.  This is useful
when colons might be removed from the input after preprocessing but before
assembling, so that adjoining tokens can still be distinguished if there is
whitespace, or concatenated if there is not.

@item tc_frob_section
@cindex tc_frob_section
If you define this macro, GAS will call it for each
section at the end of the assembly.

@item tc_frob_file_before_adjust
@cindex tc_frob_file_before_adjust
If you define this macro, GAS will call it after the symbol values are
resolved, but before the fixups have been changed from local symbols to section
symbols.

@item tc_frob_symbol
@cindex tc_frob_symbol
If you define this macro, GAS will call it for each symbol.  You can indicate
that the symbol should not be included in the object file by defining this
macro to set its second argument to a non-zero value.

@item tc_frob_file
@cindex tc_frob_file
If you define this macro, GAS will call it after the symbol table has been
completed, but before the relocations have been generated.

@item tc_frob_file_after_relocs
If you define this macro, GAS will call it after the relocs have been
generated.

@item md_post_relax_hook
If you define this macro, GAS will call it after relaxing and sizing the
segments.

@item LISTING_HEADER
A string to use on the header line of a listing.  The default value is simply
@code{"GAS LISTING"}.

@item LISTING_WORD_SIZE
The number of bytes to put into a word in a listing.  This affects the way the
bytes are clumped together in the listing.  For example, a value of 2 might
print @samp{1234 5678} where a value of 1 would print @samp{12 34 56 78}.  The
default value is 4.

@item LISTING_LHS_WIDTH
The number of words of data to print on the first line of a listing for a
particular source line, where each word is @code{LISTING_WORD_SIZE} bytes.  The
default value is 1.

@item LISTING_LHS_WIDTH_SECOND
Like @code{LISTING_LHS_WIDTH}, but applying to the second and subsequent line
of the data printed for a particular source line.  The default value is 1.

@item LISTING_LHS_CONT_LINES
The maximum number of continuation lines to print in a listing for a particular
source line.  The default value is 4.

@item LISTING_RHS_WIDTH
The maximum number of characters to print from one line of the input file.  The
default value is 100.

@item TC_COFF_SECTION_DEFAULT_ATTRIBUTES
@cindex TC_COFF_SECTION_DEFAULT_ATTRIBUTES
The COFF @code{.section} directive will use the value of this macro to set
a new section's attributes when a directive has no valid flags or when the
flag is @code{w}. The default value of the macro is @code{SEC_LOAD | SEC_DATA}.

@item DWARF2_FORMAT ()
@cindex DWARF2_FORMAT
If you define this, it should return one of @code{dwarf2_format_32bit},
@code{dwarf2_format_64bit}, or @code{dwarf2_format_64bit_irix} to indicate
the size of internal DWARF section offsets and the format of the DWARF initial
length fields.  When @code{dwarf2_format_32bit} is returned, the initial
length field will be 4 bytes long and section offsets are 32 bits in size.
For @code{dwarf2_format_64bit} and @code{dwarf2_format_64bit_irix}, section
offsets are 64 bits in size, but the initial length field differs.  An 8 byte
initial length is indicated by @code{dwarf2_format_64bit_irix} and
@code{dwarf2_format_64bit} indicates a 12 byte initial length field in
which the first four bytes are 0xffffffff and the next 8 bytes are
the section's length.

If you don't define this, @code{dwarf2_format_32bit} will be used as
the default.

This define only affects @code{.debug_info} and @code{.debug_line}
sections generated by the assembler.  DWARF 2 sections generated by
other tools will be unaffected by this setting.

@item DWARF2_ADDR_SIZE (@var{bfd})
@cindex DWARF2_ADDR_SIZE
It should return the size of an address, as it should be represented in
debugging info.  If you don't define this macro, the default definition uses
the number of bits per address, as defined in @var{bfd}, divided by 8.

@item   MD_DEBUG_FORMAT_SELECTOR
@cindex MD_DEBUG_FORMAT_SELECTOR
If defined this macro is the name of a function to be called when the
@samp{--gen-debug} switch is detected on the assembler's command line.  The
prototype for the function looks like this:

@smallexample
   enum debug_info_type MD_DEBUG_FORMAT_SELECTOR (int * use_gnu_extensions)
@end smallexample

The function should return the debug format that is preferred by the CPU
backend.  This format will be used when generating assembler specific debug
information.

@item md_allow_local_subtract (@var{left}, @var{right}, @var{section})
If defined, GAS will call this macro when evaluating an expression which is the
difference of two symbols defined in the same section.  It takes three
arguments: @code{expressioS * @var{left}} which is the symbolic expression on
the left hand side of the subtraction operation, @code{expressionS *
@var{right}} which is the symbolic expression on the right hand side of the
subtraction, and @code{segT @var{section}} which is the section containing the two
symbols.  The macro should return a non-zero value if the expression should be
evaluated.  Targets which implement link time relaxation which may change the
position of the two symbols relative to each other should ensure that this
macro returns zero in situations where this can occur.

@item md_allow_eh_opt
If defined, GAS will check this macro before performing any optimizations on
the DWARF call frame debug information that is emitted.  Targets which
implement link time relaxation may need to define this macro and set it to zero
if it is possible to change the size of a function's prologue.
@end table

@node Object format backend
@subsection Writing an object format backend
@cindex object format backend
@cindex @file{obj-@var{fmt}}

As with the CPU backend, the object format backend must define a few things,
and may define some other things.  The interface to the object format backend
is generally simpler; most of the support for an object file format consists of
defining a number of pseudo-ops.

The object format @file{.h} file must include @file{targ-cpu.h}.

@table @code
@item OBJ_@var{format}
@cindex OBJ_@var{format}
By convention, you should define this macro in the @file{.h} file.  For
example, @file{obj-elf.h} defines @code{OBJ_ELF}.  You might have to use this
if it is necessary to add object file format specific code to the CPU file.

@item obj_begin
If you define this macro, GAS will call it at the start of the assembly, after
the command line arguments have been parsed and all the machine independent
initializations have been completed.

@item obj_app_file
@cindex obj_app_file
If you define this macro, GAS will invoke it when it sees a @code{.file}
pseudo-op or a @samp{#} line as used by the C preprocessor.

@item OBJ_COPY_SYMBOL_ATTRIBUTES
@cindex OBJ_COPY_SYMBOL_ATTRIBUTES
You should define this macro to copy object format specific information from
one symbol to another.  GAS will call it when one symbol is equated to
another.

@item obj_sec_sym_ok_for_reloc
@cindex obj_sec_sym_ok_for_reloc
You may define this macro to indicate that it is OK to use a section symbol in
a relocation entry.  If it is not, GAS will define a new symbol at the start
of a section.

@item EMIT_SECTION_SYMBOLS
@cindex EMIT_SECTION_SYMBOLS
You should define this macro with a zero value if you do not want to include
section symbols in the output symbol table.  The default value for this macro
is one.

@item obj_adjust_symtab
@cindex obj_adjust_symtab
If you define this macro, GAS will invoke it just before setting the symbol
table of the output BFD.  For example, the COFF support uses this macro to
generate a @code{.file} symbol if none was generated previously.

@item SEPARATE_STAB_SECTIONS
@cindex SEPARATE_STAB_SECTIONS
You may define this macro to a nonzero value to indicate that stabs should be
placed in separate sections, as in ELF.

@item INIT_STAB_SECTION
@cindex INIT_STAB_SECTION
You may define this macro to initialize the stabs section in the output file.

@item OBJ_PROCESS_STAB
@cindex OBJ_PROCESS_STAB
You may define this macro to do specific processing on a stabs entry.

@item obj_frob_section
@cindex obj_frob_section
If you define this macro, GAS will call it for each section at the end of the
assembly.

@item obj_frob_file_before_adjust
@cindex obj_frob_file_before_adjust
If you define this macro, GAS will call it after the symbol values are
resolved, but before the fixups have been changed from local symbols to section
symbols.

@item obj_frob_symbol
@cindex obj_frob_symbol
If you define this macro, GAS will call it for each symbol.  You can indicate
that the symbol should not be included in the object file by defining this
macro to set its second argument to a non-zero value.

@item obj_set_weak_hook
@cindex obj_set_weak_hook
If you define this macro, @code{S_SET_WEAK} will call it before modifying the
symbol's flags.

@item obj_clear_weak_hook
@cindex obj_clear_weak_hook
If you define this macro, @code{S_CLEAR_WEAKREFD} will call it after cleaning
the @code{weakrefd} flag, but before modifying any other flags.

@item obj_frob_file
@cindex obj_frob_file
If you define this macro, GAS will call it after the symbol table has been
completed, but before the relocations have been generated.

@item obj_frob_file_after_relocs
If you define this macro, GAS will call it after the relocs have been
generated.

@item SET_SECTION_RELOCS (@var{sec}, @var{relocs}, @var{n})
@cindex SET_SECTION_RELOCS
If you define this, it will be called after the relocations have been set for
the section @var{sec}.  The list of relocations is in @var{relocs}, and the
number of relocations is in @var{n}.
@end table

@node Emulations
@subsection Writing emulation files

Normally you do not have to write an emulation file.  You can just use
@file{te-generic.h}.

If you do write your own emulation file, it must include @file{obj-format.h}.

An emulation file will often define @code{TE_@var{EM}}; this may then be used
in other files to change the output.

@node Relaxation
@section Relaxation
@cindex relaxation

@dfn{Relaxation} is a generic term used when the size of some instruction or
data depends upon the value of some symbol or other data.

GAS knows to relax a particular type of PC relative relocation using a table.
You can also define arbitrarily complex forms of relaxation yourself.

@menu
* Relaxing with a table::       Relaxing with a table
* General relaxing::            General relaxing
@end menu

@node Relaxing with a table
@subsection Relaxing with a table

If you do not define @code{md_relax_frag}, and you do define
@code{TC_GENERIC_RELAX_TABLE}, GAS will relax @code{rs_machine_dependent} frags
based on the frag subtype and the displacement to some specified target
address.  The basic idea is that several machines have different addressing
modes for instructions that can specify different ranges of values, with
successive modes able to access wider ranges, including the entirety of the
previous range.  Smaller ranges are assumed to be more desirable (perhaps the
instruction requires one word instead of two or three); if this is not the
case, don't describe the smaller-range, inferior mode.

The @code{fr_subtype} field of a frag is an index into a CPU-specific
relaxation table.  That table entry indicates the range of values that can be
stored, the number of bytes that will have to be added to the frag to
accommodate the addressing mode, and the index of the next entry to examine if
the value to be stored is outside the range accessible by the current
addressing mode.  The @code{fr_symbol} field of the frag indicates what symbol
is to be accessed; the @code{fr_offset} field is added in.

If the @code{TC_PCREL_ADJUST} macro is defined, which currently should only happen
for the NS32k family, the @code{TC_PCREL_ADJUST} macro is called on the frag to
compute an adjustment to be made to the displacement.

The value fitted by the relaxation code is always assumed to be a displacement
from the current frag.  (More specifically, from @code{fr_fix} bytes into the
frag.)
@ignore
This seems kinda silly.  What about fitting small absolute values?  I suppose
@code{md_assemble} is supposed to take care of that, but if the operand is a
difference between symbols, it might not be able to, if the difference was not
computable yet.
@end ignore

The end of the relaxation sequence is indicated by a ``next'' value of 0.  This
means that the first entry in the table can't be used.

For some configurations, the linker can do relaxing within a section of an
object file.  If call instructions of various sizes exist, the linker can
determine which should be used in each instance, when a symbol's value is
resolved.  In order for the linker to avoid wasting space and having to insert
no-op instructions, it must be able to expand or shrink the section contents
while still preserving intra-section references and meeting alignment
requirements.

For the i960 using b.out format, no expansion is done; instead, each
@samp{.align} directive causes extra space to be allocated, enough that when
the linker is relaxing a section and removing unneeded space, it can discard
some or all of this extra padding and cause the following data to be correctly
aligned.

For the H8/300, I think the linker expands calls that can't reach, and doesn't
worry about alignment issues; the cpu probably never needs any significant
alignment beyond the instruction size.

The relaxation table type contains these fields:

@table @code
@item long rlx_forward
Forward reach, must be non-negative.
@item long rlx_backward
Backward reach, must be zero or negative.
@item rlx_length
Length in bytes of this addressing mode.
@item rlx_more
Index of the next-longer relax state, or zero if there is no next relax state.
@end table

The relaxation is done in @code{relax_segment} in @file{write.c}.  The
difference in the length fields between the original mode and the one finally
chosen by the relaxing code is taken as the size by which the current frag will
be increased in size.  For example, if the initial relaxing mode has a length
of 2 bytes, and because of the size of the displacement, it gets upgraded to a
mode with a size of 6 bytes, it is assumed that the frag will grow by 4 bytes.
(The initial two bytes should have been part of the fixed portion of the frag,
since it is already known that they will be output.)  This growth must be
effected by @code{md_convert_frag}; it should increase the @code{fr_fix} field
by the appropriate size, and fill in the appropriate bytes of the frag.
(Enough space for the maximum growth should have been allocated in the call to
frag_var as the second argument.)

If relocation records are needed, they should be emitted by
@code{md_estimate_size_before_relax}.  This function should examine the target
symbol of the supplied frag and correct the @code{fr_subtype} of the frag if
needed.  When this function is called, if the symbol has not yet been defined,
it will not become defined later; however, its value may still change if the
section it is in gets relaxed.

Usually, if the symbol is in the same section as the frag (given by the
@var{sec} argument), the narrowest likely relaxation mode is stored in
@code{fr_subtype}, and that's that.

If the symbol is undefined, or in a different section (and therefore movable
to an arbitrarily large distance), the largest available relaxation mode is
specified, @code{fix_new} is called to produce the relocation record,
@code{fr_fix} is increased to include the relocated field (remember, this
storage was allocated when @code{frag_var} was called), and @code{frag_wane} is
called to convert the frag to an @code{rs_fill} frag with no variant part.
Sometimes changing addressing modes may also require rewriting the instruction.
It can be accessed via @code{fr_opcode} or @code{fr_fix}.

If you generate frags separately for the basic insn opcode and any relaxable
operands, do not call @code{fix_new} thinking you can emit fixups for the
opcode field from the relaxable frag.  It is not guaranteed to be the same frag.
If you need to emit fixups for the opcode field from inspection of the
relaxable frag, then you need to generate a common frag for both the basic
opcode and relaxable fields, or you need to provide the frag for the opcode to
pass to @code{fix_new}.  The latter can be done for example by defining
@code{TC_FRAG_TYPE} to include a pointer to it and defining @code{TC_FRAG_INIT}
to set the pointer.

Sometimes @code{fr_var} is increased instead, and @code{frag_wane} is not
called.  I'm not sure, but I think this is to keep @code{fr_fix} referring to
an earlier byte, and @code{fr_subtype} set to @code{rs_machine_dependent} so
that @code{md_convert_frag} will get called.

@node General relaxing
@subsection General relaxing

If using a simple table is not suitable, you may implement arbitrarily complex
relaxation semantics yourself.  For example, the MIPS backend uses this to emit
different instruction sequences depending upon the size of the symbol being
accessed.

When you assemble an instruction that may need relaxation, you should allocate
a frag using @code{frag_var} or @code{frag_variant} with a type of
@code{rs_machine_dependent}.  You should store some sort of information in the
@code{fr_subtype} field so that you can figure out what to do with the frag
later.

When GAS reaches the end of the input file, it will look through the frags and
work out their final sizes.

GAS will first call @code{md_estimate_size_before_relax} on each
@code{rs_machine_dependent} frag.  This function must return an estimated size
for the frag.

GAS will then loop over the frags, calling @code{md_relax_frag} on each
@code{rs_machine_dependent} frag.  This function should return the change in
size of the frag.  GAS will keep looping over the frags until none of the frags
changes size.

@node Broken words
@section Broken words
@cindex internals, broken words
@cindex broken words

Some compilers, including GCC, will sometimes emit switch tables specifying
16-bit @code{.word} displacements to branch targets, and branch instructions
that load entries from that table to compute the target address.  If this is
done on a 32-bit machine, there is a chance (at least with really large
functions) that the displacement will not fit in 16 bits.  The assembler
handles this using a concept called @dfn{broken words}.  This idea is well
named, since there is an implied promise that the 16-bit field will in fact
hold the specified displacement.

If broken word processing is enabled, and a situation like this is encountered,
the assembler will insert a jump instruction into the instruction stream, close
enough to be reached with the 16-bit displacement.  This jump instruction will
transfer to the real desired target address.  Thus, as long as the @code{.word}
value really is used as a displacement to compute an address to jump to, the
net effect will be correct (minus a very small efficiency cost).  If
@code{.word} directives with label differences for values are used for other
purposes, however, things may not work properly.  For targets which use broken
words, the @samp{-K} option will warn when a broken word is discovered.

The broken word code is turned off by the @code{WORKING_DOT_WORD} macro.  It
isn't needed if @code{.word} emits a value large enough to contain an address
(or, more correctly, any possible difference between two addresses).

@node Internal functions
@section Internal functions

This section describes basic internal functions used by GAS.

@menu
* Warning and error messages::  Warning and error messages
* Hash tables::                 Hash tables
@end menu

@node Warning and error messages
@subsection Warning and error messages

@deftypefun  @{@} int had_warnings (void)
@deftypefunx @{@} int had_errors (void)
Returns non-zero if any warnings or errors, respectively, have been printed
during this invocation.
@end deftypefun

@deftypefun  @{@} void as_tsktsk (const char *@var{format}, ...)
@deftypefunx @{@} void as_warn (const char *@var{format}, ...)
@deftypefunx @{@} void as_bad (const char *@var{format}, ...)
@deftypefunx @{@} void as_fatal (const char *@var{format}, ...)
These functions display messages about something amiss with the input file, or
internal problems in the assembler itself.  The current file name and line
number are printed, followed by the supplied message, formatted using
@code{vfprintf}, and a final newline.

An error indicated by @code{as_bad} will result in a non-zero exit status when
the assembler has finished.  Calling @code{as_fatal} will result in immediate
termination of the assembler process.
@end deftypefun

@deftypefun @{@} void as_warn_where (char *@var{file}, unsigned int @var{line}, const char *@var{format}, ...)
@deftypefunx @{@} void as_bad_where (char *@var{file}, unsigned int @var{line}, const char *@var{format}, ...)
These variants permit specification of the file name and line number, and are
used when problems are detected when reprocessing information saved away when
processing some earlier part of the file.  For example, fixups are processed
after all input has been read, but messages about fixups should refer to the
original filename and line number that they are applicable to.
@end deftypefun

@deftypefun @{@} void sprint_value (char *@var{buf}, valueT @var{val})
This function is helpful for converting a @code{valueT} value into printable
format, in case it's wider than modes that @code{*printf} can handle.  If the
type is narrow enough, a decimal number will be produced; otherwise, it will be
in hexadecimal.  The value itself is not examined to make this determination.
@end deftypefun

@node Hash tables
@subsection Hash tables
@cindex hash tables

@deftypefun @{@} @{struct hash_control *@} hash_new (void)
Creates the hash table control structure.
@end deftypefun

@deftypefun @{@} void hash_die (struct hash_control *)
Destroy a hash table.
@end deftypefun

@deftypefun @{@} PTR hash_delete (struct hash_control *, const char *)
Deletes entry from the hash table, returns the value it had.
@end deftypefun

@deftypefun @{@} PTR hash_replace (struct hash_control *, const char *, PTR)
Updates the value for an entry already in the table, returning the old value.
If no entry was found, just returns NULL.
@end deftypefun

@deftypefun @{@} @{const char *@} hash_insert (struct hash_control *, const char *, PTR)
Inserting a value already in the table is an error.
Returns an error message or NULL.
@end deftypefun

@deftypefun @{@} @{const char *@} hash_jam (struct hash_control *, const char *, PTR)
Inserts if the value isn't already present, updates it if it is.
@end deftypefun

@node Test suite
@section Test suite
@cindex test suite

The test suite is kind of lame for most processors.  Often it only checks to
see if a couple of files can be assembled without the assembler reporting any
errors.  For more complete testing, write a test which either examines the
assembler listing, or runs @code{objdump} and examines its output.  For the
latter, the TCL procedure @code{run_dump_test} may come in handy.  It takes the
base name of a file, and looks for @file{@var{file}.d}.  This file should
contain as its initial lines a set of variable settings in @samp{#} comments,
in the form:

@example
        #@var{varname}: @var{value}
@end example

The @var{varname} may be @code{objdump}, @code{nm}, or @code{as}, in which case
it specifies the options to be passed to the specified programs.  Exactly one
of @code{objdump} or @code{nm} must be specified, as that also specifies which
program to run after the assembler has finished.  If @var{varname} is
@code{source}, it specifies the name of the source file; otherwise,
@file{@var{file}.s} is used.  If @var{varname} is @code{name}, it specifies the
name of the test to be used in the @code{pass} or @code{fail} messages.

The non-commented parts of the file are interpreted as regular expressions, one
per line.  Blank lines in the @code{objdump} or @code{nm} output are skipped,
as are blank lines in the @code{.d} file; the other lines are tested to see if
the regular expression matches the program output.  If it does not, the test
fails.

Note that this means the tests must be modified if the @code{objdump} output
style is changed.

@bye
@c Local Variables:
@c fill-column: 79
@c End: