xref: /netbsd-src/external/gpl3/binutils/dist/gas/doc/internals.texi (revision cb63e24e8d6aae7ddac1859a9015f48b1d8bd90e)
1\input texinfo
2@c  Copyright (C) 1991-2024 Free Software Foundation, Inc.
3@setfilename internals.info
4@node Top
5@top Assembler Internals
6@raisesections
7@cindex internals
8
9This chapter describes the internals of the assembler.  It is incomplete, but
10it may help a bit.
11
12This chapter is not updated regularly, and it may be out of date.
13
14@menu
15* Data types::		Data types
16* GAS processing::      What GAS does when it runs
17* Porting GAS::         Porting GAS
18* Relaxation::          Relaxation
19* Broken words::        Broken words
20* Internal functions::  Internal functions
21* Test suite::          Test suite
22@end menu
23
24@node Data types
25@section Data types
26@cindex internals, data types
27
28This section describes some fundamental GAS data types.
29
30@menu
31* Symbols::             The symbolS structure
32* Expressions::         The expressionS structure
33* Fixups::		The fixS structure
34* Frags::               The fragS structure
35@end menu
36
37@node Symbols
38@subsection Symbols
39@cindex internals, symbols
40@cindex symbols, internal
41@cindex symbolS structure
42
43The definition for the symbol structure, @code{symbolS}, is located in
44@file{symbols.c}.
45
46The fields of this structure may not be referred to directly.
47Instead, you must use one of the accessor functions defined in @file{symbol.h}.
48
49Symbol structures contain the following fields:
50
51@table @code
52@item sy_value
53This is an @code{expressionS} that describes the value of the symbol.  It might
54refer to one or more other symbols; if so, its true value may not be known
55until @code{resolve_symbol_value} is called with @var{finalize_syms} non-zero
56in @code{write_object_file}.
57
58The expression is often simply a constant.  Before @code{resolve_symbol_value}
59is called with @var{finalize_syms} set, the value is the offset from the frag
60(@pxref{Frags}).  Afterward, the frag address has been added in.
61
62@item sy_resolved
63This field is non-zero if the symbol's value has been completely resolved.  It
64is used during the final pass over the symbol table.
65
66@item sy_resolving
67This field is used to detect loops while resolving the symbol's value.
68
69@item sy_used_in_reloc
70This field is non-zero if the symbol is used by a relocation entry.  If a local
71symbol is used in a relocation entry, it must be possible to redirect those
72relocations to other symbols, or this symbol cannot be removed from the final
73symbol list.
74
75@item sy_next
76@itemx sy_previous
77These pointers to other @code{symbolS} structures describe a doubly
78linked list.  These fields should be accessed with
79the @code{symbol_next} and @code{symbol_previous} macros.
80
81@item sy_frag
82This points to the frag (@pxref{Frags}) that this symbol is attached to.
83
84@item sy_used
85Whether the symbol is used as an operand or in an expression.  Note: Not all of
86the backends keep this information accurate; backends which use this bit are
87responsible for setting it when a symbol is used in backend routines.
88
89@item sy_mri_common
90Whether the symbol is an MRI common symbol created by the @code{COMMON}
91pseudo-op when assembling in MRI mode.
92
93@item sy_volatile
94Whether the symbol can be re-defined.
95
96@item sy_forward_ref
97Whether the symbol's value must only be evaluated upon use.
98
99@item sy_weakrefr
100Whether the symbol is a @code{weakref} alias to another symbol.
101
102@item sy_weakrefd
103Whether the symbol is or was referenced by one or more @code{weakref} aliases,
104and has not had any direct references.
105
106@item bsym
107This points to the BFD @code{asymbol} that
108will be used in writing the object file.
109
110@item sy_obj
111This format-specific data is of type @code{OBJ_SYMFIELD_TYPE}.  If no macro by
112that name is defined in @file{obj-format.h}, this field is not defined.
113
114@item sy_tc
115This processor-specific data is of type @code{TC_SYMFIELD_TYPE}.  If no macro
116by that name is defined in @file{targ-cpu.h}, this field is not defined.
117
118@end table
119
120Here is a description of the accessor functions.  These should be used rather
121than referring to the fields of @code{symbolS} directly.
122
123@table @code
124@item S_SET_VALUE
125@cindex S_SET_VALUE
126Set the symbol's value.
127
128@item S_GET_VALUE
129@cindex S_GET_VALUE
130Get the symbol's value.  This will cause @code{resolve_symbol_value} to be
131called if necessary.
132
133@item S_SET_SEGMENT
134@cindex S_SET_SEGMENT
135Set the section of the symbol.
136
137@item S_GET_SEGMENT
138@cindex S_GET_SEGMENT
139Get the symbol's section.
140
141@item S_GET_NAME
142@cindex S_GET_NAME
143Get the name of the symbol.
144
145@item S_SET_NAME
146@cindex S_SET_NAME
147Set the name of the symbol.
148
149@item S_IS_EXTERNAL
150@cindex S_IS_EXTERNAL
151Return non-zero if the symbol is externally visible.
152
153@item S_IS_WEAK
154@cindex S_IS_WEAK
155Return non-zero if the symbol is weak, or if it is a @code{weakref} alias or
156symbol that has not been strongly referenced.
157
158@item S_IS_WEAKREFR
159@cindex S_IS_WEAKREFR
160Return non-zero if the symbol is a @code{weakref} alias.
161
162@item S_IS_WEAKREFD
163@cindex S_IS_WEAKREFD
164Return non-zero if the symbol was aliased by a @code{weakref} alias and has not
165had any strong references.
166
167@item S_IS_VOLATILE
168@cindex S_IS_VOLATILE
169Return non-zero if the symbol may be re-defined. Such symbols get created by
170the @code{=} operator, @code{equ}, or @code{set}.
171
172@item S_IS_FORWARD_REF
173@cindex S_IS_FORWARD_REF
174Return non-zero if the symbol is a forward reference, that is its value must
175only be determined upon use.
176
177@item S_IS_COMMON
178@cindex S_IS_COMMON
179Return non-zero if this is a common symbol.  Common symbols are sometimes
180represented as undefined symbols with a value, in which case this function will
181not be reliable.
182
183@item S_IS_DEFINED
184@cindex S_IS_DEFINED
185Return non-zero if this symbol is defined.  This function is not reliable when
186called on a common symbol.
187
188@item S_IS_DEBUG
189@cindex S_IS_DEBUG
190Return non-zero if this is a debugging symbol.
191
192@item S_IS_LOCAL
193@cindex S_IS_LOCAL
194Return non-zero if this is a local assembler symbol which should not be
195included in the final symbol table.  Note that this is not the opposite of
196@code{S_IS_EXTERNAL}.  The @samp{-L} assembler option affects the return value
197of this function.
198
199@item S_SET_EXTERNAL
200@cindex S_SET_EXTERNAL
201Mark the symbol as externally visible.
202
203@item S_CLEAR_EXTERNAL
204@cindex S_CLEAR_EXTERNAL
205Mark the symbol as not externally visible.
206
207@item S_SET_WEAK
208@cindex S_SET_WEAK
209Mark the symbol as weak.
210
211@item S_SET_WEAKREFR
212@cindex S_SET_WEAKREFR
213Mark the symbol as the referrer in a @code{weakref} directive.  The symbol it
214aliases must have been set to the value expression before this point.  If the
215alias has already been used, the symbol is marked as used too.
216
217@item S_CLEAR_WEAKREFR
218@cindex S_CLEAR_WEAKREFR
219Clear the @code{weakref} alias status of a symbol.  This is implicitly called
220whenever a symbol is defined or set to a new expression.
221
222@item S_SET_WEAKREFD
223@cindex S_SET_WEAKREFD
224Mark the symbol as the referred symbol in a @code{weakref} directive.
225Implicitly marks the symbol as weak, but see below.  It should only be called
226if the referenced symbol has just been added to the symbol table.
227
228@item S_SET_WEAKREFD
229@cindex S_SET_WEAKREFD
230Clear the @code{weakref} aliased status of a symbol.  This is implicitly called
231whenever the symbol is looked up, as part of a direct reference or a
232definition, but not as part of a @code{weakref} directive.
233
234@item S_SET_VOLATILE
235@cindex S_SET_VOLATILE
236Indicate that the symbol may be re-defined.
237
238@item S_CLEAR_VOLATILE
239@cindex S_CLEAR_VOLATILE
240Indicate that the symbol may no longer be re-defined.
241
242@item S_SET_FORWARD_REF
243@cindex S_SET_FORWARD_REF
244Indicate that the symbol is a forward reference, that is its value must only
245be determined upon use.
246
247@item S_GET_TYPE
248@itemx S_GET_DESC
249@itemx S_GET_OTHER
250@cindex S_GET_TYPE
251@cindex S_GET_DESC
252@cindex S_GET_OTHER
253Get the @code{type}, @code{desc}, and @code{other} fields of the symbol.  These
254are only defined for object file formats for which they make sense (primarily
255a.out).
256
257@item S_SET_TYPE
258@itemx S_SET_DESC
259@itemx S_SET_OTHER
260@cindex S_SET_TYPE
261@cindex S_SET_DESC
262@cindex S_SET_OTHER
263Set the @code{type}, @code{desc}, and @code{other} fields of the symbol.  These
264are only defined for object file formats for which they make sense (primarily
265a.out).
266
267@item S_GET_SIZE
268@cindex S_GET_SIZE
269Get the size of a symbol.  This is only defined for object file formats for
270which it makes sense (primarily ELF).
271
272@item S_SET_SIZE
273@cindex S_SET_SIZE
274Set the size of a symbol.  This is only defined for object file formats for
275which it makes sense (primarily ELF).
276
277@item symbol_get_value_expression
278@cindex symbol_get_value_expression
279Get a pointer to an @code{expressionS} structure which represents the value of
280the symbol as an expression.
281
282@item symbol_set_value_expression
283@cindex symbol_set_value_expression
284Set the value of a symbol to an expression.
285
286@item symbol_set_frag
287@cindex symbol_set_frag
288Set the frag where a symbol is defined.
289
290@item symbol_get_frag
291@cindex symbol_get_frag
292Get the frag where a symbol is defined.
293
294@item symbol_mark_used
295@cindex symbol_mark_used
296Mark a symbol as having been used in an expression.
297
298@item symbol_clear_used
299@cindex symbol_clear_used
300Clear the mark indicating that a symbol was used in an expression.
301
302@item symbol_used_p
303@cindex symbol_used_p
304Return whether a symbol was used in an expression.
305
306@item symbol_mark_used_in_reloc
307@cindex symbol_mark_used_in_reloc
308Mark a symbol as having been used by a relocation.
309
310@item symbol_clear_used_in_reloc
311@cindex symbol_clear_used_in_reloc
312Clear the mark indicating that a symbol was used in a relocation.
313
314@item symbol_used_in_reloc_p
315@cindex symbol_used_in_reloc_p
316Return whether a symbol was used in a relocation.
317
318@item symbol_mark_mri_common
319@cindex symbol_mark_mri_common
320Mark a symbol as an MRI common symbol.
321
322@item symbol_clear_mri_common
323@cindex symbol_clear_mri_common
324Clear the mark indicating that a symbol is an MRI common symbol.
325
326@item symbol_mri_common_p
327@cindex symbol_mri_common_p
328Return whether a symbol is an MRI common symbol.
329
330@item symbol_mark_written
331@cindex symbol_mark_written
332Mark a symbol as having been written.
333
334@item symbol_clear_written
335@cindex symbol_clear_written
336Clear the mark indicating that a symbol was written.
337
338@item symbol_written_p
339@cindex symbol_written_p
340Return whether a symbol was written.
341
342@item symbol_mark_resolved
343@cindex symbol_mark_resolved
344Mark a symbol as having been resolved.
345
346@item symbol_resolved_p
347@cindex symbol_resolved_p
348Return whether a symbol has been resolved.
349
350@item symbol_section_p
351@cindex symbol_section_p
352Return whether a symbol is a section symbol.
353
354@item symbol_equated_p
355@cindex symbol_equated_p
356Return whether a symbol is equated to another symbol.
357
358@item symbol_constant_p
359@cindex symbol_constant_p
360Return whether a symbol has a constant value, including being an offset within
361some frag.
362
363@item symbol_get_bfdsym
364@cindex symbol_get_bfdsym
365Return the BFD symbol associated with a symbol.
366
367@item symbol_set_bfdsym
368@cindex symbol_set_bfdsym
369Set the BFD symbol associated with a symbol.
370
371@item symbol_get_obj
372@cindex symbol_get_obj
373Return a pointer to the @code{OBJ_SYMFIELD_TYPE} field of a symbol.
374
375@item symbol_set_obj
376@cindex symbol_set_obj
377Set the @code{OBJ_SYMFIELD_TYPE} field of a symbol.
378
379@item symbol_get_tc
380@cindex symbol_get_tc
381Return a pointer to the @code{TC_SYMFIELD_TYPE} field of a symbol.
382
383@item symbol_set_tc
384@cindex symbol_set_tc
385Set the @code{TC_SYMFIELD_TYPE} field of a symbol.
386
387@end table
388
389GAS attempts to store local
390symbols--symbols which will not be written to the output file--using a
391different structure, @code{struct local_symbol}.  This structure can only
392represent symbols whose value is an offset within a frag.
393
394Code outside of the symbol handler will always deal with @code{symbolS}
395structures and use the accessor functions.  The accessor functions correctly
396deal with local symbols.  @code{struct local_symbol} is much smaller than
397@code{symbolS} (which also automatically creates a bfd @code{asymbol}
398structure), so this saves space when assembling large files.
399
400@node Expressions
401@subsection Expressions
402@cindex internals, expressions
403@cindex expressions, internal
404@cindex expressionS structure
405
406Expressions are stored in an @code{expressionS} structure.  The structure is
407defined in @file{expr.h}.
408
409@cindex expression
410The macro @code{expression} will create an @code{expressionS} structure based
411on the text found at the global variable @code{input_line_pointer}.
412
413@cindex make_expr_symbol
414@cindex expr_symbol_where
415A single @code{expressionS} structure can represent a single operation.
416Complex expressions are formed by creating @dfn{expression symbols} and
417combining them in @code{expressionS} structures.  An expression symbol is
418created by calling @code{make_expr_symbol}.  An expression symbol should
419naturally never appear in a symbol table, and the implementation of
420@code{S_IS_LOCAL} (@pxref{Symbols}) reflects that.  The function
421@code{expr_symbol_where} returns non-zero if a symbol is an expression symbol,
422and also returns the file and line for the expression which caused it to be
423created.
424
425The @code{expressionS} structure has two symbol fields, a number field, an
426operator field, and a field indicating whether the number is unsigned.
427
428The operator field is of type @code{operatorT}, and describes how to interpret
429the other fields; see the definition in @file{expr.h} for the possibilities.
430
431An @code{operatorT} value of @code{O_big} indicates either a floating point
432number, stored in the global variable @code{generic_floating_point_number}, or
433an integer too large to store in an @code{offsetT} type, stored in the global
434array @code{generic_bignum}.  This rather inflexible approach makes it
435impossible to use floating point numbers or large expressions in complex
436expressions.
437
438@node Fixups
439@subsection Fixups
440@cindex internals, fixups
441@cindex fixups
442@cindex fixS structure
443
444A @dfn{fixup} is basically anything which can not be resolved in the first
445pass.  Sometimes a fixup can be resolved by the end of the assembly; if not,
446the fixup becomes a relocation entry in the object file.
447
448@cindex fix_new
449@cindex fix_new_exp
450A fixup is created by a call to @code{fix_new} or @code{fix_new_exp}.  Both
451take a frag (@pxref{Frags}), a position within the frag, a size, an indication
452of whether the fixup is PC relative, and a type.
453The type is nominally a @code{bfd_reloc_code_real_type}, but several
454targets use other type codes to represent fixups that can not be described as
455relocations.
456
457The @code{fixS} structure has a number of fields, several of which are obsolete
458or are only used by a particular target.  The important fields are:
459
460@table @code
461@item fx_frag
462The frag (@pxref{Frags}) this fixup is in.
463
464@item fx_where
465The location within the frag where the fixup occurs.
466
467@item fx_addsy
468The symbol this fixup is against.  Typically, the value of this symbol is added
469into the object contents.  This may be NULL.
470
471@item fx_subsy
472The value of this symbol is subtracted from the object contents.  This is
473normally NULL.
474
475@item fx_offset
476A number which is added into the fixup.
477
478@item fx_addnumber
479Some CPU backends use this field to convey information between
480@code{md_apply_fix} and @code{tc_gen_reloc}.  The machine independent code does
481not use it.
482
483@item fx_next
484The next fixup in the section.
485
486@item fx_r_type
487The type of the fixup.
488
489@item fx_size
490The size of the fixup.  This is mostly used for error checking.
491
492@item fx_pcrel
493Whether the fixup is PC relative.
494
495@item fx_done
496Non-zero if the fixup has been applied, and no relocation entry needs to be
497generated.
498
499@item fx_file
500@itemx fx_line
501The file and line where the fixup was created.
502
503@item tc_fix_data
504This has the type @code{TC_FIX_TYPE}, and is only defined if the target defines
505that macro.
506@end table
507
508@node Frags
509@subsection Frags
510@cindex internals, frags
511@cindex frags
512@cindex fragS structure.
513
514The @code{fragS} structure is defined in @file{as.h}.  Each frag represents a
515portion of the final object file.  As GAS reads the source file, it creates
516frags to hold the data that it reads.  At the end of the assembly the frags and
517fixups are processed to produce the final contents.
518
519@table @code
520@item fr_address
521The address of the frag.  This is not set until the assembler rescans the list
522of all frags after the entire input file is parsed.  The function
523@code{relax_segment} fills in this field.
524
525@item fr_next
526Pointer to the next frag in this (sub)section.
527
528@item fr_fix
529Fixed number of characters we know we're going to emit to the output file.  May
530be zero.
531
532@item fr_var
533Variable number of characters we may output, after the initial @code{fr_fix}
534characters.  May be zero.
535
536@item fr_offset
537The interpretation of this field is controlled by @code{fr_type}.  Generally,
538if @code{fr_var} is non-zero, this is a repeat count: the @code{fr_var}
539characters are output @code{fr_offset} times.
540
541@item line
542Holds line number info when an assembler listing was requested.
543
544@item fr_type
545Relaxation state.  This field indicates the interpretation of @code{fr_offset},
546@code{fr_symbol} and the variable-length tail of the frag, as well as the
547treatment it gets in various phases of processing.  It does not affect the
548initial @code{fr_fix} characters; they are always supposed to be output
549verbatim (fixups aside).  See below for specific values this field can have.
550
551@item fr_subtype
552Relaxation substate.  If the macro @code{md_relax_frag} isn't defined, this is
553assumed to be an index into @code{TC_GENERIC_RELAX_TABLE} for the generic
554relaxation code to process (@pxref{Relaxation}).  If @code{md_relax_frag} is
555defined, this field is available for any use by the CPU-specific code.
556
557@item fr_symbol
558This normally indicates the symbol to use when relaxing the frag according to
559@code{fr_type}.
560
561@item fr_opcode
562Points to the lowest-addressed byte of the opcode, for use in relaxation.
563
564@item tc_frag_data
565Target specific fragment data of type TC_FRAG_TYPE.
566Only present if @code{TC_FRAG_TYPE} is defined.
567
568@item fr_file
569@itemx fr_line
570The file and line where this frag was last modified.
571
572@item fr_literal
573Declared as a one-character array, this last field grows arbitrarily large to
574hold the actual contents of the frag.
575@end table
576
577These are the possible relaxation states, provided in the enumeration type
578@code{relax_stateT}, and the interpretations they represent for the other
579fields:
580
581@table @code
582@item rs_align
583@itemx rs_align_code
584The start of the following frag should be aligned on some boundary.  In this
585frag, @code{fr_offset} is the logarithm (base 2) of the alignment in bytes.
586(For example, if alignment on an 8-byte boundary were desired, @code{fr_offset}
587would have a value of 3.)  The variable characters indicate the fill pattern to
588be used.  The @code{fr_subtype} field holds the maximum number of bytes to skip
589when doing this alignment.  If more bytes are needed, the alignment is not
590done.  An @code{fr_subtype} value of 0 means no maximum, which is the normal
591case.  Target backends can use @code{rs_align_code} to handle certain types of
592alignment differently.
593
594@item rs_broken_word
595This indicates that ``broken word'' processing should be done (@pxref{Broken
596words}).  If broken word processing is not necessary on the target machine,
597this enumerator value will not be defined.
598
599@item rs_cfa
600This state is used to implement exception frame optimizations.  The
601@code{fr_symbol} is an expression symbol for the subtraction which may be
602relaxed.  The @code{fr_opcode} field holds the frag for the preceding command
603byte.  The @code{fr_offset} field holds the offset within that frag.  The
604@code{fr_subtype} field is used during relaxation to hold the current size of
605the frag.
606
607@item rs_fill
608The variable characters are to be repeated @code{fr_offset} times.  If
609@code{fr_offset} is 0, this frag has a length of @code{fr_fix}.  Most frags
610have this type.
611
612@item rs_leb128
613This state is used to implement the DWARF ``little endian base 128''
614variable length number format.  The @code{fr_symbol} is always an expression
615symbol, as constant expressions are emitted directly.  The @code{fr_offset}
616field is used during relaxation to hold the previous size of the number so
617that we can determine if the fragment changed size.
618
619@item rs_machine_dependent
620Displacement relaxation is to be done on this frag.  The target is indicated by
621@code{fr_symbol} and @code{fr_offset}, and @code{fr_subtype} indicates the
622particular machine-specific addressing mode desired.  @xref{Relaxation}.
623
624@item rs_org
625The start of the following frag should be pushed back to some specific offset
626within the section.  (Some assemblers use the value as an absolute address; GAS
627does not handle final absolute addresses, but rather requires that the linker
628set them.)  The offset is given by @code{fr_symbol} and @code{fr_offset}; one
629character from the variable-length tail is used as the fill character.
630@end table
631
632@cindex frchainS structure
633A chain of frags is built up for each subsection.  The data structure
634describing a chain is called a @code{frchainS}, and contains the following
635fields:
636
637@table @code
638@item frch_root
639Points to the first frag in the chain.  May be NULL if there are no frags in
640this chain.
641@item frch_last
642Points to the last frag in the chain, or NULL if there are none.
643@item frch_next
644Next in the list of @code{frchainS} structures.
645@item frch_seg
646Indicates the section this frag chain belongs to.
647@item frch_subseg
648Subsection (subsegment) number of this frag chain.
649@item fix_root, fix_tail
650Point to first and last @code{fixS} structures associated with this subsection.
651@item frch_obstack
652Not currently used.  Intended to be used for frag allocation for this
653subsection.  This should reduce frag generation caused by switching sections.
654@item frch_frag_now
655The current frag for this subsegment.
656@end table
657
658A @code{frchainS} corresponds to a subsection; each section has a list of
659@code{frchainS} records associated with it.  In most cases, only one subsection
660of each section is used, so the list will only be one element long, but any
661processing of frag chains should be prepared to deal with multiple chains per
662section.
663
664After the input files have been completely processed, and no more frags are to
665be generated, the frag chains are joined into one per section for further
666processing.  After this point, it is safe to operate on one chain per section.
667
668The assembler always has a current frag, named @code{frag_now}.  More space is
669allocated for the current frag using the @code{frag_more} function; this
670returns a pointer to the amount of requested space.  The function
671@code{frag_room} says by how much the current frag can be extended.
672Relaxing is done using variant frags allocated by @code{frag_var}
673or @code{frag_variant} (@pxref{Relaxation}).
674
675@node GAS processing
676@section What GAS does when it runs
677@cindex internals, overview
678
679This is a quick look at what an assembler run looks like.
680
681@itemize @bullet
682@item
683The assembler initializes itself by calling various init routines.
684
685@item
686For each source file, the @code{read_a_source_file} function reads in the file
687and parses it.  The global variable @code{input_line_pointer} points to the
688current text; it is guaranteed to be correct up to the end of the line, but not
689farther.
690
691@item
692For each line, the assembler passes labels to the @code{colon} function, and
693isolates the first word.  If it looks like a pseudo-op, the word is looked up
694in the pseudo-op hash table @code{po_hash} and dispatched to a pseudo-op
695routine.  Otherwise, the target dependent @code{md_assemble} routine is called
696to parse the instruction.
697
698@item
699When pseudo-ops or instructions output data, they add it to a frag, calling
700@code{frag_more} to get space to store it in.
701
702@item
703Pseudo-ops and instructions can also output fixups created by @code{fix_new} or
704@code{fix_new_exp}.
705
706@item
707For certain targets, instructions can create variant frags which are used to
708store relaxation information (@pxref{Relaxation}).
709
710@item
711When the input file is finished, the @code{write_object_file} routine is
712called.  It assigns addresses to all the frags (@code{relax_segment}), resolves
713all the fixups (@code{fixup_segment}), resolves all the symbol values (using
714@code{resolve_symbol_value}), and finally writes out the file.
715@end itemize
716
717@node Porting GAS
718@section Porting GAS
719@cindex porting
720
721Each GAS target specifies two main things: the CPU file and the object format
722file.  Two main switches in the @file{configure.ac} file handle this.  The
723first switches on CPU type to set the shell variable @code{cpu_type}.  The
724second switches on the entire target to set the shell variable @code{fmt}.
725
726The configure script uses the value of @code{cpu_type} to select two files in
727the @file{config} directory: @file{tc-@var{CPU}.c} and @file{tc-@var{CPU}.h}.
728The configuration process will create a file named @file{targ-cpu.h} in the
729build directory which includes @file{tc-@var{CPU}.h}.
730
731The configure script also uses the value of @code{fmt} to select two files:
732@file{obj-@var{fmt}.c} and @file{obj-@var{fmt}.h}.  The configuration process
733will create a file named @file{obj-format.h} in the build directory which
734includes @file{obj-@var{fmt}.h}.
735
736You can also set the emulation in the configure script by setting the @code{em}
737variable.  Normally the default value of @samp{generic} is fine.  The
738configuration process will create a file named @file{targ-env.h} in the build
739directory which includes @file{te-@var{em}.h}.
740
741There is a special case for COFF. For historical reason, the GNU COFF
742assembler doesn't follow the documented behavior on certain debug symbols for
743the compatibility with other COFF assemblers. A port can define
744@code{STRICTCOFF} in the configure script to make the GNU COFF assembler
745to follow the documented behavior.
746
747Porting GAS to a new CPU requires writing the @file{tc-@var{CPU}} files.
748Porting GAS to a new object file format requires writing the
749@file{obj-@var{fmt}} files.  There is sometimes some interaction between these
750two files, but it is normally minimal.
751
752The best approach is, of course, to copy existing files.  The documentation
753below assumes that you are looking at existing files to see usage details.
754
755These interfaces have grown over time, and have never been carefully thought
756out or designed.  Nothing about the interfaces described here is cast in stone.
757It is possible that they will change from one version of the assembler to the
758next.  Also, new macros are added all the time as they are needed.
759
760@menu
761* CPU backend::                 Writing a CPU backend
762* Object format backend::       Writing an object format backend
763* Emulations::                  Writing emulation files
764@end menu
765
766@node CPU backend
767@subsection Writing a CPU backend
768@cindex CPU backend
769@cindex @file{tc-@var{CPU}}
770
771The CPU backend files are the heart of the assembler.  They are the only parts
772of the assembler which actually know anything about the instruction set of the
773processor.
774
775You must define a reasonably small list of macros and functions in the CPU
776backend files.  You may define a large number of additional macros in the CPU
777backend files, not all of which are documented here.  You must, of course,
778define macros in the @file{.h} file, which is included by every assembler
779source file.  You may define the functions as macros in the @file{.h} file, or
780as functions in the @file{.c} file.
781
782@table @code
783@item TC_@var{CPU}
784@cindex TC_@var{CPU}
785By convention, you should define this macro in the @file{.h} file.  For
786example, @file{tc-m68k.h} defines @code{TC_M68K}.  You might have to use this
787if it is necessary to add CPU specific code to the object format file.
788
789@item TARGET_FORMAT
790This macro is the BFD target name to use when creating the output file.  This
791will normally depend upon the @code{OBJ_@var{FMT}} macro.
792
793@item TARGET_ARCH
794This macro is the BFD architecture to pass to @code{bfd_set_arch_mach}.
795
796@item TARGET_MACH
797This macro is the BFD machine number to pass to @code{bfd_set_arch_mach}.  If
798it is not defined, GAS will use 0.
799
800@item TARGET_BYTES_BIG_ENDIAN
801You should define this macro to be non-zero if the target is big endian, and
802zero if the target is little endian.
803
804@item md_shortopts
805@itemx md_longopts
806@itemx md_longopts_size
807@itemx md_parse_option
808@itemx md_show_usage
809@itemx md_after_parse_args
810@cindex md_shortopts
811@cindex md_longopts
812@cindex md_longopts_size
813@cindex md_parse_option
814@cindex md_show_usage
815@cindex md_after_parse_args
816GAS uses these variables and functions during option processing.
817@code{md_shortopts} is a @code{const char *} which GAS adds to the machine
818independent string passed to @code{getopt}.  @code{md_longopts} is a
819@code{struct option []} which GAS adds to the machine independent long options
820passed to @code{getopt}; you may use @code{OPTION_MD_BASE}, defined in
821@file{as.h}, as the start of a set of long option indices, if necessary.
822@code{md_longopts_size} is a @code{size_t} holding the size @code{md_longopts}.
823
824GAS will call @code{md_parse_option} whenever @code{getopt} returns an
825unrecognized code, presumably indicating a special code value which appears in
826@code{md_longopts}.  This function should return non-zero if it handled the
827option and zero otherwise.  There is no need to print a message about an option
828not being recognized.  This will be handled by the generic code.
829
830GAS will call @code{md_show_usage} when a usage message is printed; it should
831print a description of the machine specific options. @code{md_after_pase_args},
832if defined, is called after all options are processed, to let the backend
833override settings done by the generic option parsing.
834
835@item md_begin
836@cindex md_begin
837GAS will call this function at the start of the assembly, after the command
838line arguments have been parsed and all the machine independent initializations
839have been completed.
840
841@item md_cleanup
842@cindex md_cleanup
843If you define this macro, GAS will call it at the end of each input file.
844
845@item md_assemble
846@cindex md_assemble
847GAS will call this function for each input line which does not contain a
848pseudo-op.  The argument is a null terminated string.  The function should
849assemble the string as an instruction with operands.  Normally
850@code{md_assemble} will do this by calling @code{frag_more} and writing out
851some bytes (@pxref{Frags}).  @code{md_assemble} will call @code{fix_new} to
852create fixups as needed (@pxref{Fixups}).  Targets which need to do special
853purpose relaxation will call @code{frag_var}.
854
855@item md_pseudo_table
856@cindex md_pseudo_table
857This is a const array of type @code{pseudo_typeS}.  It is a mapping from
858pseudo-op names to functions.  You should use this table to implement
859pseudo-ops which are specific to the CPU.
860
861@item tc_conditional_pseudoop
862@cindex tc_conditional_pseudoop
863If this macro is defined, GAS will call it with a @code{pseudo_typeS} argument.
864It should return non-zero if the pseudo-op is a conditional which controls
865whether code is assembled, such as @samp{.if}.  GAS knows about the normal
866conditional pseudo-ops, and you should normally not have to define this macro.
867
868@item comment_chars
869@cindex comment_chars
870This is a null terminated @code{const char} array of characters which start a
871comment.
872
873@item tc_comment_chars
874@cindex tc_comment_chars
875If this macro is defined, GAS will use it instead of @code{comment_chars}.
876This has the advantage that this macro does not have to refer to a constant
877array.
878
879@item tc_symbol_chars
880@cindex tc_symbol_chars
881If this macro is defined, it is a pointer to a null terminated list of
882characters which may appear in an operand.  GAS already assumes that all
883alphanumeric characters, and @samp{$}, @samp{.}, and @samp{_} may appear in an
884operand (see @samp{symbol_chars} in @file{app.c}).  This macro may be defined
885to treat additional characters as appearing in an operand.  This affects the
886way in which GAS removes whitespace before passing the string to
887@samp{md_assemble}.
888
889@item line_comment_chars
890@cindex line_comment_chars
891This is a null terminated @code{const char} array of characters which start a
892comment when they appear at the start of a line.
893
894@item line_separator_chars
895@cindex line_separator_chars
896This is a null terminated @code{const char} array of characters which separate
897lines (null and newline are such characters by default, and need not be
898listed in this array).  Note that line_separator_chars do not separate lines
899if found in a comment, such as after a character in line_comment_chars or
900comment_chars.
901
902@item tc_line_separator_chars
903@cindex tc_line_separator_chars
904If this macro is defined, GAS will use it instead of
905@code{line_separator_chars}.  This has the advantage that this macro does not
906have to refer to a constant array.
907
908
909@item EXP_CHARS
910@cindex EXP_CHARS
911This is a null terminated @code{const char} array of characters which may be
912used as the exponent character in a floating point number.  This is normally
913@code{"eE"}.
914
915@item FLT_CHARS
916@cindex FLT_CHARS
917This is a null terminated @code{const char} array of characters which may be
918used to indicate a floating point constant.  A zero followed by one of these
919characters is assumed to be followed by a floating point number; thus they
920operate the way that @code{0x} is used to indicate a hexadecimal constant.
921Usually this includes @samp{r} and @samp{f}.
922
923@item LEX_AT
924@cindex LEX_AT
925You may define this macro to the lexical type of the @kbd{@@} character.  The
926default is zero.
927
928Lexical types are a combination of @code{LEX_NAME} and @code{LEX_BEGIN_NAME},
929both defined in @file{read.h}.  @code{LEX_NAME} indicates that the character
930may appear in a name.  @code{LEX_BEGIN_NAME} indicates that the character may
931appear at the beginning of a name.
932
933@item LEX_BR
934@cindex LEX_BR
935You may define this macro to the lexical type of the brace characters @kbd{@{},
936@kbd{@}}, @kbd{[}, and @kbd{]}.  The default value is zero.
937
938@item LEX_PCT
939@cindex LEX_PCT
940You may define this macro to the lexical type of the @kbd{%} character.  The
941default value is zero.
942
943@item LEX_QM
944@cindex LEX_QM
945You may define this macro to the lexical type of the @kbd{?} character.  The
946default value it zero.
947
948@item LEX_DOLLAR
949@cindex LEX_DOLLAR
950You may define this macro to the lexical type of the @kbd{$} character.  The
951default value is @code{LEX_NAME | LEX_BEGIN_NAME}.
952
953@item NUMBERS_WITH_SUFFIX
954@cindex NUMBERS_WITH_SUFFIX
955When this macro is defined to be non-zero, the parser allows the radix of a
956constant to be indicated with a suffix.  Valid suffixes are binary (B),
957octal (Q), and hexadecimal (H).  Case is not significant.
958
959@item SINGLE_QUOTE_STRINGS
960@cindex SINGLE_QUOTE_STRINGS
961If you define this macro, GAS will treat single quotes as string delimiters.
962Normally only double quotes are accepted as string delimiters.
963
964@item NO_STRING_ESCAPES
965@cindex NO_STRING_ESCAPES
966If you define this macro, GAS will not permit escape sequences in a string.
967
968@item ONLY_STANDARD_ESCAPES
969@cindex ONLY_STANDARD_ESCAPES
970If you define this macro, GAS will warn about the use of nonstandard escape
971sequences in a string.
972
973@item md_start_line_hook
974@cindex md_start_line_hook
975If you define this macro, GAS will call it at the start of each line.
976
977@item LABELS_WITHOUT_COLONS
978@cindex LABELS_WITHOUT_COLONS
979If you define this macro, GAS will assume that any text at the start of a line
980is a label, even if it does not have a colon.
981
982@item TC_START_LABEL
983@itemx TC_START_LABEL_WITHOUT_COLON
984@cindex TC_START_LABEL
985You may define this macro to control what GAS considers to be a label.  The
986default definition is to accept any name followed by a colon character.
987
988@item TC_START_LABEL_WITHOUT_COLON
989@cindex TC_START_LABEL_WITHOUT_COLON
990Same as TC_START_LABEL, but should be used instead of TC_START_LABEL when
991LABELS_WITHOUT_COLONS is defined.
992
993@item TC_FAKE_LABEL
994@cindex TC_FAKE_LABEL
995You may define this macro to control what GAS considers to be a fake
996label.  The default fake label is FAKE_LABEL_NAME.
997
998@item NO_PSEUDO_DOT
999@cindex NO_PSEUDO_DOT
1000If you define this macro, GAS will not require pseudo-ops to start with a
1001@kbd{.} character.
1002
1003@item TC_EQUAL_IN_INSN
1004@cindex TC_EQUAL_IN_INSN
1005If you define this macro, it should return nonzero if the instruction is
1006permitted to contain an @kbd{=} character.  GAS will call it with two
1007arguments, the character before the @kbd{=} character, and the value of
1008the string preceding the equal sign. GAS uses this macro to decide if a
1009@kbd{=} is an assignment or an instruction.
1010
1011@item TC_EOL_IN_INSN
1012@cindex TC_EOL_IN_INSN
1013If you define this macro, it should return nonzero if the current input line
1014pointer should be treated as the end of a line.
1015
1016@item TC_CASE_SENSITIVE
1017@cindex TC_CASE_SENSITIVE
1018Define this macro if instruction mnemonics and pseudos are case sensitive.
1019The default is to have it undefined giving case insensitive names.
1020
1021@item md_parse_name
1022@cindex md_parse_name
1023If this macro is defined, GAS will call it for any symbol found in an
1024expression.  You can define this to handle special symbols in a special way.
1025If a symbol always has a certain value, you should normally enter it in the
1026symbol table, perhaps using @code{reg_section}.
1027
1028@item md_undefined_symbol
1029@cindex md_undefined_symbol
1030GAS will call this function when a symbol table lookup fails, before it
1031creates a new symbol.  Typically this would be used to supply symbols whose
1032name or value changes dynamically, possibly in a context sensitive way.
1033Predefined symbols with fixed values, such as register names or condition
1034codes, are typically entered directly into the symbol table when @code{md_begin}
1035is called.  One argument is passed, a @code{char *} for the symbol.
1036
1037@item md_resolve_symbol
1038@cindex md_resolve_symbol
1039If this macro is defined, GAS will call it upon resolving machine-dependent
1040symbols (that is, for any symbol with operation O_md1..O_md32 inclusively).
1041If this functions returns zero, then the symbol could not be resolved.
1042
1043@item md_operand
1044@cindex md_operand
1045GAS will call this function with one argument, an @code{expressionS}
1046pointer, for any expression that can not be recognized.  When the function
1047is called, @code{input_line_pointer} will point to the start of the
1048expression.
1049
1050@item md_optimize_expr
1051@cindex md_optimize_expr
1052GAS will call this function before trying to carry out certain operations,
1053like the adding of two constants.  The function is passed the left-hand
1054operand, an @code{expressionS} pointer, the operator, an @code{operatorT}
1055value, and the right-hand operand, again an @code{expressionS} pointer.  For
1056unary expressions NULL is passed as first argument.
1057
1058@item md_register_arithmetic
1059@cindex md_register_arithmetic
1060If this macro is defined and evaluates to zero then GAS will not fold
1061expressions that add or subtract a constant to/from a register to give
1062another register.  For example GAS's default behaviour is to fold the
1063expression "r8 + 1" into "r9", which is probably not the result
1064intended by the programmer.  The default is to allow such folding,
1065since this maintains backwards compatibility with earlier releases of
1066GAS.
1067
1068@item tc_unrecognized_line
1069@cindex tc_unrecognized_line
1070If you define this macro, GAS will call it when it finds a line that it can not
1071parse.
1072
1073@item md_do_align
1074@cindex md_do_align
1075You may define this macro to handle an alignment directive.  GAS will call it
1076when the directive is seen in the input file.  For example, the i386 backend
1077uses this to generate efficient nop instructions of varying lengths, depending
1078upon the number of bytes that the alignment will skip.
1079
1080@item HANDLE_ALIGN
1081@cindex HANDLE_ALIGN
1082You may define this macro to do special handling for an alignment directive.
1083GAS will call it at the end of the assembly.
1084
1085@item TC_IMPLICIT_LCOMM_ALIGNMENT (@var{size}, @var{p2var})
1086@cindex TC_IMPLICIT_LCOMM_ALIGNMENT
1087An @code{.lcomm} directive with no explicit alignment parameter will use this
1088macro to set @var{p2var} to the alignment that a request for @var{size} bytes
1089will have.  The alignment is expressed as a power of two.  If no alignment
1090should take place, the macro definition should do nothing.  Some targets define
1091a @code{.bss} directive that is also affected by this macro.  The default
1092definition will set @var{p2var} to the truncated power of two of sizes up to
1093eight bytes.
1094
1095@item md_flush_pending_output
1096@cindex md_flush_pending_output
1097If you define this macro, GAS will call it each time it skips any space because of a
1098space filling or alignment or data allocation pseudo-op.
1099
1100@item TC_PARSE_CONS_EXPRESSION
1101@cindex TC_PARSE_CONS_EXPRESSION
1102You may define this macro to parse an expression used in a data allocation
1103pseudo-op such as @code{.word}.  You can use this to recognize relocation
1104directives that may appear in such directives.
1105
1106@item REPEAT_CONS_EXPRESSION
1107@cindex REPEAT_CONS_EXPRESSION
1108If you define this macro, GAS will recognize repeat counts in data allocation
1109pseudo-ops, as used on the MIPS.
1110
1111@item md_cons_align
1112@cindex md_cons_align
1113You may define this macro to do any special alignment before a data allocation
1114pseudo-op.
1115
1116@item TC_CONS_FIX_NEW
1117@cindex TC_CONS_FIX_NEW
1118You may define this macro to generate a fixup for a data allocation pseudo-op.
1119
1120@item TC_ADDRESS_BYTES
1121@cindex TC_ADDRESS_BYTES
1122Define this macro to specify the number of bytes used to store an address.
1123Used to implement @code{dc.a}.  If not defined by the target, a default will
1124be supplied.  Targets are assumed to have a reloc for this size.
1125
1126@item TC_INIT_FIX_DATA (@var{fixp})
1127@cindex TC_INIT_FIX_DATA
1128A C statement to initialize the target specific fields of fixup @var{fixp}.
1129These fields are defined with the @code{TC_FIX_TYPE} macro.
1130
1131@item TC_FIX_DATA_PRINT (@var{stream}, @var{fixp})
1132@cindex TC_FIX_DATA_PRINT
1133A C statement to output target specific debugging information for
1134fixup @var{fixp} to @var{stream}.  This macro is called by @code{print_fixup}.
1135
1136@item TC_FRAG_INIT (@var{fragp}, @var{max_bytes})
1137@cindex TC_FRAG_INIT
1138A C statement to initialize the target specific fields of frag @var{fragp}
1139with maximum number of bytes @var{max_bytes}.  These fields are defined
1140with the @code{TC_FRAG_TYPE} macro.
1141
1142@item md_number_to_chars
1143@cindex md_number_to_chars
1144This should just call either @code{number_to_chars_bigendian} or
1145@code{number_to_chars_littleendian}, whichever is appropriate.  On targets like
1146the MIPS which support options to change the endianness, which function to call
1147is a runtime decision.  On other targets, @code{md_number_to_chars} can be a
1148simple macro.
1149
1150@item md_atof (@var{type},@var{litP},@var{sizeP})
1151@cindex md_atof
1152This function is called to convert an ASCII string into a floating point value
1153in format used by the CPU.  It takes three arguments.  The first is @var{type}
1154which is a byte describing the type of floating point number to be created.  It
1155is one of the characters defined in the @code{FLT_CHARS} macro.  Possible
1156values are @var{'f'} or @var{'s'} for single precision, @var{'d'} or @var{'r'}
1157for double precision and @var{'x'} or @var{'p'} for extended precision.  Either
1158lower or upper case versions of these letters can be used.  Note: some targets
1159do not support all of these types, and some targets may also support other
1160types not mentioned here.
1161
1162The second parameter is @var{litP} which is a pointer to a byte array where the
1163converted value should be stored.  The value is converted into LITTLENUMs and
1164is stored in the target's endian-ness order.  (@var{LITTLENUM} is defined in
1165gas/bignum.h).  Single precision values occupy 2 littlenums.  Double precision
1166values occupy 4 littlenums and extended precision values occupy either 5 or 6
1167littlenums, depending upon the target.
1168
1169The third argument is @var{sizeP}, which is a pointer to a integer that should
1170be filled in with the number of chars emitted into the byte array.
1171
1172The function should return NULL upon success or an error string upon failure.
1173
1174@item TC_LARGEST_EXPONENT_IS_NORMAL
1175@cindex TC_LARGEST_EXPONENT_IS_NORMAL (@var{precision})
1176This macro is used only by @file{atof-ieee.c}.  It should evaluate to true
1177if floats of the given precision use the largest exponent for normal numbers
1178instead of NaNs and infinities.  @var{precision} is @samp{F_PRECISION} for
1179single precision, @samp{D_PRECISION} for double precision, or
1180@samp{X_PRECISION} for extended double precision.
1181
1182The macro has a default definition which returns 0 for all cases.
1183
1184@item WORKING_DOT_WORD
1185@itemx md_short_jump_size
1186@itemx md_long_jump_size
1187@itemx md_create_short_jump
1188@itemx md_create_long_jump
1189@itemx TC_CHECK_ADJUSTED_BROKEN_DOT_WORD
1190@cindex WORKING_DOT_WORD
1191@cindex md_short_jump_size
1192@cindex md_long_jump_size
1193@cindex md_create_short_jump
1194@cindex md_create_long_jump
1195@cindex TC_CHECK_ADJUSTED_BROKEN_DOT_WORD
1196If @code{WORKING_DOT_WORD} is defined, GAS will not do broken word processing
1197(@pxref{Broken words}).  Otherwise, you should set @code{md_short_jump_size} to
1198the size of a short jump (a jump that is just long enough to jump around a
1199number of long jumps) and @code{md_long_jump_size} to the size of a long jump
1200(a jump that can go anywhere in the function).  You should define
1201@code{md_create_short_jump} to create a short jump around a number of long
1202jumps, and define @code{md_create_long_jump} to create a long jump.
1203If defined, the macro TC_CHECK_ADJUSTED_BROKEN_DOT_WORD will be called for each
1204adjusted word just before the word is output.  The macro takes two arguments,
1205an @code{addressT} with the adjusted word and a pointer to the current
1206@code{struct broken_word}.
1207
1208@item md_estimate_size_before_relax
1209@cindex md_estimate_size_before_relax
1210This function returns an estimate of the size of a @code{rs_machine_dependent}
1211frag before any relaxing is done.  It may also create any necessary
1212relocations.
1213
1214@item md_relax_frag
1215@cindex md_relax_frag
1216This macro may be defined to relax a frag.  GAS will call this with the
1217segment, the frag, and the change in size of all previous frags;
1218@code{md_relax_frag} should return the change in size of the frag.
1219@xref{Relaxation}.
1220
1221@item TC_GENERIC_RELAX_TABLE
1222@cindex TC_GENERIC_RELAX_TABLE
1223If you do not define @code{md_relax_frag}, you may define
1224@code{TC_GENERIC_RELAX_TABLE} as a table of @code{relax_typeS} structures.  The
1225machine independent code knows how to use such a table to relax PC relative
1226references.  See @file{tc-m68k.c} for an example.  @xref{Relaxation}.
1227
1228@item md_generic_table_relax_frag
1229@cindex md_generic_table_relax_frag
1230If defined, it is a C statement that is invoked, instead of
1231the default implementation, to scan @code{TC_GENERIC_RELAX_TABLE}.
1232
1233@item md_prepare_relax_scan
1234@cindex md_prepare_relax_scan
1235If defined, it is a C statement that is invoked prior to scanning
1236the relax table.
1237
1238@item LINKER_RELAXING_SHRINKS_ONLY
1239@cindex LINKER_RELAXING_SHRINKS_ONLY
1240If you define this macro, and the global variable @samp{linkrelax} is set
1241(because of a command-line option, or unconditionally in @code{md_begin}), a
1242@samp{.align} directive will cause extra space to be allocated.  The linker can
1243then discard this space when relaxing the section.
1244
1245@item TC_LINKRELAX_FIXUP (@var{segT})
1246@cindex TC_LINKRELAX_FIXUP
1247If defined, this macro allows control over whether fixups for a
1248given section will be processed when the @var{linkrelax} variable is
1249set.  The macro is given the N_TYPE bits for the section in its
1250@var{segT} argument.  If the macro evaluates to a non-zero value
1251then the fixups will be converted into relocs, otherwise they will
1252be passed to @var{md_apply_fix} as normal.
1253
1254@item md_convert_frag
1255@cindex md_convert_frag
1256GAS will call this for each rs_machine_dependent fragment.
1257The instruction is completed using the data from the relaxation pass.
1258It may also create any necessary relocations.
1259@xref{Relaxation}.
1260
1261@item TC_FINALIZE_SYMS_BEFORE_SIZE_SEG
1262@cindex TC_FINALIZE_SYMS_BEFORE_SIZE_SEG
1263Specifies the value to be assigned to @code{finalize_syms} before the function
1264@code{size_segs} is called.  Since @code{size_segs} calls @code{cvt_frag_to_fill}
1265which can call @code{md_convert_frag}, this constant governs whether the symbols
1266accessed in @code{md_convert_frag} will be fully resolved.  In particular it
1267governs whether local symbols will have been resolved, and had their frag
1268information removed.  Depending upon the processing performed by
1269@code{md_convert_frag} the frag information may or may not be necessary, as may
1270the resolved values of the symbols.  The default value is 1.
1271
1272@item TC_VALIDATE_FIX (@var{fixP}, @var{seg}, @var{skip})
1273@cindex TC_VALIDATE_FIX
1274This macro is evaluated for each fixup (when @var{linkrelax} is not set).
1275It may be used to change the fixup in @code{struct fix *@var{fixP}} before
1276the generic code sees it, or to fully process the fixup.  In the latter case,
1277a @code{goto @var{skip}} will bypass the generic code.
1278
1279@item md_apply_fix (@var{fixP}, @var{valP}, @var{seg})
1280@cindex md_apply_fix
1281GAS will call this for each fixup that passes the @code{TC_VALIDATE_FIX} test
1282when @var{linkrelax} is not set.  It should store the correct value in the
1283object file.  @code{struct fix *@var{fixP}} is the fixup @code{md_apply_fix}
1284is operating on.  @code{valueT *@var{valP}} is the value to store into the
1285object files, or at least is the generic code's best guess.  Specifically,
1286*@var{valP} is the value of the fixup symbol, perhaps modified by
1287@code{MD_APPLY_SYM_VALUE}, plus @code{@var{fixP}->fx_offset} (symbol addend),
1288less @code{MD_PCREL_FROM_SECTION} for pc-relative fixups.
1289@code{segT @var{seg}} is the section the fix is in.
1290@code{fixup_segment} performs a generic overflow check on *@var{valP} after
1291@code{md_apply_fix} returns.  If the overflow check is relevant for the target
1292machine, then @code{md_apply_fix} should modify *@var{valP}, typically to the
1293value stored in the object file.
1294
1295@item TC_FORCE_RELOCATION (@var{fix})
1296@cindex TC_FORCE_RELOCATION
1297If this macro returns non-zero, it guarantees that a relocation will be emitted
1298even when the value can be resolved locally, as @code{fixup_segment} tries to
1299reduce the number of relocations emitted.  For example, a fixup expression
1300against an absolute symbol will normally not require a reloc.  If undefined,
1301a default of @w{@code{(S_FORCE_RELOC ((@var{fix})->fx_addsy))}} is used.
1302
1303@item TC_FORCE_RELOCATION_ABS (@var{fix})
1304@cindex TC_FORCE_RELOCATION_ABS
1305Like @code{TC_FORCE_RELOCATION}, but used only for fixup expressions against an
1306absolute symbol.  If undefined, @code{TC_FORCE_RELOCATION} will be used.
1307
1308@item TC_FORCE_RELOCATION_LOCAL (@var{fix})
1309@cindex TC_FORCE_RELOCATION_LOCAL
1310Like @code{TC_FORCE_RELOCATION}, but used only for fixup expressions against a
1311symbol in the current section.  If undefined, fixups that are not
1312@code{fx_pcrel} or for which @code{TC_FORCE_RELOCATION}
1313returns non-zero, will emit relocs.
1314
1315@item TC_FORCE_RELOCATION_SUB_SAME (@var{fix}, @var{seg})
1316@cindex TC_FORCE_RELOCATION_SUB_SAME
1317This macro controls resolution of fixup expressions involving the
1318difference of two symbols in the same section.  If this macro returns zero,
1319the subtrahend will be resolved and @code{fx_subsy} set to @code{NULL} for
1320@code{md_apply_fix}.  If undefined, the default of
1321@w{@code{! SEG_NORMAL (@var{seg})}} will be used.
1322
1323@item TC_FORCE_RELOCATION_SUB_ABS (@var{fix}, @var{seg})
1324@cindex TC_FORCE_RELOCATION_SUB_ABS
1325Like @code{TC_FORCE_RELOCATION_SUB_SAME}, but used when the subtrahend is an
1326absolute symbol.  If the macro is undefined a default of @code{0} is used.
1327
1328@item TC_FORCE_RELOCATION_SUB_LOCAL (@var{fix}, @var{seg})
1329@cindex TC_FORCE_RELOCATION_SUB_LOCAL
1330Like @code{TC_FORCE_RELOCATION_SUB_ABS}, but the subtrahend is a symbol in the
1331same section as the fixup.
1332
1333@item TC_VALIDATE_FIX_SUB (@var{fix}, @var{seg})
1334@cindex TC_VALIDATE_FIX_SUB
1335This macro is evaluated for any fixup with a @code{fx_subsy} that
1336@code{fixup_segment} cannot reduce to a number.  If the macro returns
1337@code{false} an error will be reported.
1338
1339@item TC_GLOBAL_REGISTER_SYMBOL_OK
1340@cindex TC_GLOBAL_REGISTER_SYMBOL_OK
1341Define this macro if global register symbols are supported. The default
1342is to disallow global register symbols.
1343
1344@item MD_APPLY_SYM_VALUE (@var{fix})
1345@cindex MD_APPLY_SYM_VALUE
1346This macro controls whether the symbol value becomes part of the value passed
1347to @code{md_apply_fix}.  If the macro is undefined, or returns non-zero, the
1348symbol value will be included.  For ELF, a suitable definition might simply be
1349@code{0}, because ELF relocations don't include the symbol value in the addend.
1350
1351@item S_FORCE_RELOC (@var{sym}, @var{strict})
1352@cindex S_FORCE_RELOC
1353This function returns true for symbols
1354that should not be reduced to section symbols or eliminated from expressions,
1355because they may be overridden by the linker.  ie. for symbols that are
1356undefined or common, and when @var{strict} is set, weak, or global (for ELF
1357assemblers that support ELF shared library linking semantics).
1358
1359@item EXTERN_FORCE_RELOC
1360@cindex EXTERN_FORCE_RELOC
1361This macro controls whether @code{S_FORCE_RELOC} returns true for global
1362symbols.  If undefined, the default is @code{true} for ELF assemblers, and
1363@code{false} for non-ELF.
1364
1365@item tc_gen_reloc
1366@cindex tc_gen_reloc
1367GAS will call this to generate a reloc.  GAS will pass
1368the resulting reloc to @code{bfd_install_relocation}.  This currently works
1369poorly, as @code{bfd_install_relocation} often does the wrong thing, and
1370instances of @code{tc_gen_reloc} have been written to work around the problems,
1371which in turns makes it difficult to fix @code{bfd_install_relocation}.
1372
1373@item RELOC_EXPANSION_POSSIBLE
1374@cindex RELOC_EXPANSION_POSSIBLE
1375If you define this macro, it means that @code{tc_gen_reloc} may return multiple
1376relocation entries for a single fixup.  In this case, the return value of
1377@code{tc_gen_reloc} is a pointer to a null terminated array.
1378
1379@item MAX_RELOC_EXPANSION
1380@cindex MAX_RELOC_EXPANSION
1381You must define this if @code{RELOC_EXPANSION_POSSIBLE} is defined; it
1382indicates the largest number of relocs which @code{tc_gen_reloc} may return for
1383a single fixup.
1384
1385@item tc_fix_adjustable
1386@cindex tc_fix_adjustable
1387You may define this macro to indicate whether a fixup against a locally defined
1388symbol should be adjusted to be against the section symbol.  It should return a
1389non-zero value if the adjustment is acceptable.
1390
1391@item MD_PCREL_FROM_SECTION (@var{fixp}, @var{section})
1392@cindex MD_PCREL_FROM_SECTION
1393If you define this macro, it should return the position from which the PC
1394relative adjustment for a PC relative fixup should be made.  On many
1395processors, the base of a PC relative instruction is the next instruction,
1396so this macro would return the length of an instruction, plus the address of
1397the PC relative fixup.  The latter can be calculated as
1398@var{fixp}->fx_where + @var{fixp}->fx_frag->fr_address .
1399
1400@item md_pcrel_from
1401@cindex md_pcrel_from
1402This is the default value of @code{MD_PCREL_FROM_SECTION}.  The difference is
1403that @code{md_pcrel_from} does not take a section argument.
1404
1405@item tc_frob_label
1406@cindex tc_frob_label
1407If you define this macro, GAS will call it each time a label is defined.
1408
1409@item tc_new_dot_label
1410@cindex tc_new_dot_label
1411If you define this macro, GAS will call it each time a fake label is created
1412off the special dot symbol.
1413
1414@item md_section_align
1415@cindex md_section_align
1416GAS will call this function for each section at the end of the assembly, to
1417permit the CPU backend to adjust the alignment of a section.  The function
1418must take two arguments, a @code{segT} for the section and a @code{valueT}
1419for the size of the section, and return a @code{valueT} for the rounded
1420size.
1421
1422@item md_macro_start
1423@cindex md_macro_start
1424If defined, GAS will call this macro when it starts to include a macro
1425expansion.  @code{macro_nest} indicates the current macro nesting level, which
1426includes the one being expanded.
1427
1428@item md_macro_info
1429@cindex md_macro_info
1430If defined, GAS will call this macro after the macro expansion has been
1431included in the input and after parsing the macro arguments.  The single
1432argument is a pointer to the macro processing's internal representation of the
1433macro (macro_entry *), which includes expansion of the formal arguments.
1434
1435@item md_macro_end
1436@cindex md_macro_end
1437Complement to md_macro_start.  If defined, it is called when finished
1438processing an inserted macro expansion, just before decrementing macro_nest.
1439
1440@item DOUBLEBAR_PARALLEL
1441@cindex DOUBLEBAR_PARALLEL
1442Affects the preprocessor so that lines containing '||' don't have their
1443whitespace stripped following the double bar.  This is useful for targets that
1444implement parallel instructions.
1445
1446@item KEEP_WHITE_AROUND_COLON
1447@cindex KEEP_WHITE_AROUND_COLON
1448Normally, whitespace is compressed and removed when, in the presence of the
1449colon, the adjoining tokens can be distinguished.  This option affects the
1450preprocessor so that whitespace around colons is preserved.  This is useful
1451when colons might be removed from the input after preprocessing but before
1452assembling, so that adjoining tokens can still be distinguished if there is
1453whitespace, or concatenated if there is not.
1454
1455@item tc_frob_section
1456@cindex tc_frob_section
1457If you define this macro, GAS will call it for each
1458section at the end of the assembly.
1459
1460@item tc_frob_file_before_adjust
1461@cindex tc_frob_file_before_adjust
1462If you define this macro, GAS will call it after the symbol values are
1463resolved, but before the fixups have been changed from local symbols to section
1464symbols.
1465
1466@item tc_frob_symbol
1467@cindex tc_frob_symbol
1468If you define this macro, GAS will call it for each symbol.  You can indicate
1469that the symbol should not be included in the object file by defining this
1470macro to set its second argument to a non-zero value.
1471
1472@item tc_frob_file
1473@cindex tc_frob_file
1474If you define this macro, GAS will call it after the symbol table has been
1475completed, but before the relocations have been generated.
1476
1477@item tc_frob_file_after_relocs
1478If you define this macro, GAS will call it after the relocs have been
1479generated.
1480
1481@item tc_cfi_reloc_for_encoding
1482@cindex tc_cfi_reloc_for_encoding
1483This macro is used to indicate whether a cfi encoding requires a relocation.
1484It should return the required relocation type.  Defining this macro implies
1485that Compact EH is supported.
1486
1487@item md_post_relax_hook
1488If you define this macro, GAS will call it after relaxing and sizing the
1489segments.
1490
1491@item LISTING_HEADER
1492A string to use on the header line of a listing.  The default value is simply
1493@code{"GAS LISTING"}.
1494
1495@item LISTING_WORD_SIZE
1496The number of bytes to put into a word in a listing.  This affects the way the
1497bytes are clumped together in the listing.  For example, a value of 2 might
1498print @samp{1234 5678} where a value of 1 would print @samp{12 34 56 78}.  The
1499default value is 4.
1500
1501@item LISTING_LHS_WIDTH
1502The number of words of data to print on the first line of a listing for a
1503particular source line, where each word is @code{LISTING_WORD_SIZE} bytes.  The
1504default value is 1.
1505
1506@item LISTING_LHS_WIDTH_SECOND
1507Like @code{LISTING_LHS_WIDTH}, but applying to the second and subsequent line
1508of the data printed for a particular source line.  The default value is 1.
1509
1510@item LISTING_LHS_CONT_LINES
1511The maximum number of continuation lines to print in a listing for a particular
1512source line.  The default value is 4.
1513
1514@item LISTING_RHS_WIDTH
1515The maximum number of characters to print from one line of the input file.  The
1516default value is 100.
1517
1518@item TC_COFF_SECTION_DEFAULT_ATTRIBUTES
1519@cindex TC_COFF_SECTION_DEFAULT_ATTRIBUTES
1520The COFF @code{.section} directive will use the value of this macro to set
1521a new section's attributes when a directive has no valid flags or when the
1522flag is @code{w}. The default value of the macro is @code{SEC_LOAD | SEC_DATA}.
1523
1524@item DWARF2_FORMAT (@var{sec})
1525@cindex DWARF2_FORMAT
1526If you define this, it should return one of @code{dwarf2_format_32bit},
1527@code{dwarf2_format_64bit}, or @code{dwarf2_format_64bit_irix} to indicate
1528the size of internal DWARF section offsets and the format of the DWARF initial
1529length fields.  When @code{dwarf2_format_32bit} is returned, the initial
1530length field will be 4 bytes long and section offsets are 32 bits in size.
1531For @code{dwarf2_format_64bit} and @code{dwarf2_format_64bit_irix}, section
1532offsets are 64 bits in size, but the initial length field differs.  An 8 byte
1533initial length is indicated by @code{dwarf2_format_64bit_irix} and
1534@code{dwarf2_format_64bit} indicates a 12 byte initial length field in
1535which the first four bytes are 0xffffffff and the next 8 bytes are
1536the section's length.
1537
1538If you don't define this, @code{dwarf2_format_32bit} will be used as
1539the default.
1540
1541This define only affects debug
1542sections generated by the assembler.  DWARF 2 sections generated by
1543other tools will be unaffected by this setting.
1544
1545@item DWARF2_ADDR_SIZE (@var{bfd})
1546@cindex DWARF2_ADDR_SIZE
1547It should return the size of an address, as it should be represented in
1548debugging info.  If you don't define this macro, the default definition uses
1549the number of bits per address, as defined in @var{bfd}, divided by 8.
1550
1551@item   MD_DEBUG_FORMAT_SELECTOR
1552@cindex MD_DEBUG_FORMAT_SELECTOR
1553If defined this macro is the name of a function to be called when the
1554@samp{--gen-debug} switch is detected on the assembler's command line.  The
1555prototype for the function looks like this:
1556
1557@smallexample
1558   enum debug_info_type MD_DEBUG_FORMAT_SELECTOR (int * use_gnu_extensions)
1559@end smallexample
1560
1561The function should return the debug format that is preferred by the CPU
1562backend.  This format will be used when generating assembler specific debug
1563information.
1564
1565@item md_emit_single_noop_insn
1566@itemx md_single_noop_insn
1567These macro facilitate the @var{.nop} directive.  If defined the
1568@var{md_emit_single_noop_insn} macro provides code to insert a single no-op
1569instruction into the output stream.  If this involves calling @var{md_assemble}
1570with a fixed string then the alternative macro @var{md_single_noop_insn} can be
1571defined, specifying the string to pass.  If neither of these macros are defined
1572then the @var{.nop} directive will call @var{md_assemble} with the string
1573@option{nop}.
1574
1575@item md_allow_local_subtract (@var{left}, @var{right}, @var{section})
1576If defined, GAS will call this macro when evaluating an expression which is the
1577difference of two symbols defined in the same section.  It takes three
1578arguments: @code{expressioS * @var{left}} which is the symbolic expression on
1579the left hand side of the subtraction operation, @code{expressionS *
1580@var{right}} which is the symbolic expression on the right hand side of the
1581subtraction, and @code{segT @var{section}} which is the section containing the two
1582symbols.  The macro should return a non-zero value if the expression should be
1583evaluated.  Targets which implement link time relaxation which may change the
1584position of the two symbols relative to each other should ensure that this
1585macro returns zero in situations where this can occur.
1586
1587@item md_allow_eh_opt
1588If defined, GAS will check this macro before performing any optimizations on
1589the DWARF call frame debug information that is emitted.  Targets which
1590implement link time relaxation may need to define this macro and set it to zero
1591if it is possible to change the size of a function's prologue.
1592
1593@item TARGET_MULTIPLE_EH_FRAME_SECTIONS
1594If defined, GAS will create multiple .eh_frame.* sections according to
1595the name of owner's function sections.
1596@end table
1597
1598@node Object format backend
1599@subsection Writing an object format backend
1600@cindex object format backend
1601@cindex @file{obj-@var{fmt}}
1602
1603As with the CPU backend, the object format backend must define a few things,
1604and may define some other things.  The interface to the object format backend
1605is generally simpler; most of the support for an object file format consists of
1606defining a number of pseudo-ops.
1607
1608The object format @file{.h} file must include @file{targ-cpu.h}.
1609
1610@table @code
1611@item OBJ_@var{format}
1612@cindex OBJ_@var{format}
1613By convention, you should define this macro in the @file{.h} file.  For
1614example, @file{obj-elf.h} defines @code{OBJ_ELF}.  You might have to use this
1615if it is necessary to add object file format specific code to the CPU file.
1616
1617@item obj_begin
1618If you define this macro, GAS will call it at the start of the assembly, after
1619the command-line arguments have been parsed and all the machine independent
1620initializations have been completed.
1621
1622@item obj_app_file
1623@cindex obj_app_file
1624If you define this macro, GAS will invoke it when it sees a @code{.file}
1625pseudo-op or a @samp{#} line as used by the C preprocessor.
1626
1627@item OBJ_COPY_SYMBOL_ATTRIBUTES
1628@cindex OBJ_COPY_SYMBOL_ATTRIBUTES
1629You should define this macro to copy object format specific information from
1630one symbol to another.  GAS will call it when one symbol is equated to
1631another.
1632
1633@item obj_sec_sym_ok_for_reloc
1634@cindex obj_sec_sym_ok_for_reloc
1635You may define this macro to indicate that it is OK to use a section symbol in
1636a relocation entry.  If it is not, GAS will define a new symbol at the start
1637of a section.
1638
1639@item EMIT_SECTION_SYMBOLS
1640@cindex EMIT_SECTION_SYMBOLS
1641You should define this macro with a zero value if you do not want to include
1642section symbols in the output symbol table.  The default value for this macro
1643is one.
1644
1645@item obj_adjust_symtab
1646@cindex obj_adjust_symtab
1647If you define this macro, GAS will invoke it just before setting the symbol
1648table of the output BFD.  For example, the COFF support uses this macro to
1649generate a @code{.file} symbol if none was generated previously.
1650
1651@item SEPARATE_STAB_SECTIONS
1652@cindex SEPARATE_STAB_SECTIONS
1653You may define this macro to a nonzero value to indicate that stabs should be
1654placed in separate sections, as in ELF.
1655
1656@item INIT_STAB_SECTION
1657@cindex INIT_STAB_SECTION
1658You may define this macro to initialize the stabs section in the output file.
1659
1660@item OBJ_PROCESS_STAB
1661@cindex OBJ_PROCESS_STAB
1662You may define this macro to do specific processing on a stabs entry.
1663
1664@item obj_frob_section
1665@cindex obj_frob_section
1666If you define this macro, GAS will call it for each section at the end of the
1667assembly.
1668
1669@item obj_frob_file_before_adjust
1670@cindex obj_frob_file_before_adjust
1671If you define this macro, GAS will call it after the symbol values are
1672resolved, but before the fixups have been changed from local symbols to section
1673symbols.
1674
1675@item obj_frob_symbol
1676@cindex obj_frob_symbol
1677If you define this macro, GAS will call it for each symbol.  You can indicate
1678that the symbol should not be included in the object file by defining this
1679macro to set its second argument to a non-zero value.
1680
1681@item obj_set_weak_hook
1682@cindex obj_set_weak_hook
1683If you define this macro, @code{S_SET_WEAK} will call it before modifying the
1684symbol's flags.
1685
1686@item obj_clear_weak_hook
1687@cindex obj_clear_weak_hook
1688If you define this macro, @code{S_CLEAR_WEAKREFD} will call it after cleaning
1689the @code{weakrefd} flag, but before modifying any other flags.
1690
1691@item obj_frob_file
1692@cindex obj_frob_file
1693If you define this macro, GAS will call it after the symbol table has been
1694completed, but before the relocations have been generated.
1695
1696@item obj_frob_file_after_relocs
1697If you define this macro, GAS will call it after the relocs have been
1698generated.
1699
1700@item SET_SECTION_RELOCS (@var{sec}, @var{relocs}, @var{n})
1701@cindex SET_SECTION_RELOCS
1702If you define this, it will be called after the relocations have been set for
1703the section @var{sec}.  The list of relocations is in @var{relocs}, and the
1704number of relocations is in @var{n}.
1705@end table
1706
1707@node Emulations
1708@subsection Writing emulation files
1709
1710Normally you do not have to write an emulation file.  You can just use
1711@file{te-generic.h}.
1712
1713If you do write your own emulation file, it must include @file{obj-format.h}.
1714
1715An emulation file will often define @code{TE_@var{EM}}; this may then be used
1716in other files to change the output.
1717
1718@node Relaxation
1719@section Relaxation
1720@cindex relaxation
1721
1722@dfn{Relaxation} is a generic term used when the size of some instruction or
1723data depends upon the value of some symbol or other data.
1724
1725GAS knows to relax a particular type of PC relative relocation using a table.
1726You can also define arbitrarily complex forms of relaxation yourself.
1727
1728@menu
1729* Relaxing with a table::       Relaxing with a table
1730* General relaxing::            General relaxing
1731@end menu
1732
1733@node Relaxing with a table
1734@subsection Relaxing with a table
1735
1736If you do not define @code{md_relax_frag}, and you do define
1737@code{TC_GENERIC_RELAX_TABLE}, GAS will relax @code{rs_machine_dependent} frags
1738based on the frag subtype and the displacement to some specified target
1739address.  The basic idea is that several machines have different addressing
1740modes for instructions that can specify different ranges of values, with
1741successive modes able to access wider ranges, including the entirety of the
1742previous range.  Smaller ranges are assumed to be more desirable (perhaps the
1743instruction requires one word instead of two or three); if this is not the
1744case, don't describe the smaller-range, inferior mode.
1745
1746The @code{fr_subtype} field of a frag is an index into a CPU-specific
1747relaxation table.  That table entry indicates the range of values that can be
1748stored, the number of bytes that will have to be added to the frag to
1749accommodate the addressing mode, and the index of the next entry to examine if
1750the value to be stored is outside the range accessible by the current
1751addressing mode.  The @code{fr_symbol} field of the frag indicates what symbol
1752is to be accessed; the @code{fr_offset} field is added in.
1753
1754If the @code{TC_PCREL_ADJUST} macro is defined, which currently should only happen
1755for the NS32k family, the @code{TC_PCREL_ADJUST} macro is called on the frag to
1756compute an adjustment to be made to the displacement.
1757
1758The value fitted by the relaxation code is always assumed to be a displacement
1759from the current frag.  (More specifically, from @code{fr_fix} bytes into the
1760frag.)
1761@ignore
1762This seems kinda silly.  What about fitting small absolute values?  I suppose
1763@code{md_assemble} is supposed to take care of that, but if the operand is a
1764difference between symbols, it might not be able to, if the difference was not
1765computable yet.
1766@end ignore
1767
1768The end of the relaxation sequence is indicated by a ``next'' value of 0.  This
1769means that the first entry in the table can't be used.
1770
1771For some configurations, the linker can do relaxing within a section of an
1772object file.  If call instructions of various sizes exist, the linker can
1773determine which should be used in each instance, when a symbol's value is
1774resolved.  In order for the linker to avoid wasting space and having to insert
1775no-op instructions, it must be able to expand or shrink the section contents
1776while still preserving intra-section references and meeting alignment
1777requirements.
1778
1779For the H8/300, I think the linker expands calls that can't reach, and doesn't
1780worry about alignment issues; the cpu probably never needs any significant
1781alignment beyond the instruction size.
1782
1783The relaxation table type contains these fields:
1784
1785@table @code
1786@item long rlx_forward
1787Forward reach, must be non-negative.
1788@item long rlx_backward
1789Backward reach, must be zero or negative.
1790@item rlx_length
1791Length in bytes of this addressing mode.
1792@item rlx_more
1793Index of the next-longer relax state, or zero if there is no next relax state.
1794@end table
1795
1796The relaxation is done in @code{relax_segment} in @file{write.c}.  The
1797difference in the length fields between the original mode and the one finally
1798chosen by the relaxing code is taken as the size by which the current frag will
1799be increased in size.  For example, if the initial relaxing mode has a length
1800of 2 bytes, and because of the size of the displacement, it gets upgraded to a
1801mode with a size of 6 bytes, it is assumed that the frag will grow by 4 bytes.
1802(The initial two bytes should have been part of the fixed portion of the frag,
1803since it is already known that they will be output.)  This growth must be
1804effected by @code{md_convert_frag}; it should increase the @code{fr_fix} field
1805by the appropriate size, and fill in the appropriate bytes of the frag.
1806(Enough space for the maximum growth should have been allocated in the call to
1807frag_var as the second argument.)
1808
1809If relocation records are needed, they should be emitted by
1810@code{md_estimate_size_before_relax}.  This function should examine the target
1811symbol of the supplied frag and correct the @code{fr_subtype} of the frag if
1812needed.  When this function is called, if the symbol has not yet been defined,
1813it will not become defined later; however, its value may still change if the
1814section it is in gets relaxed.
1815
1816Usually, if the symbol is in the same section as the frag (given by the
1817@var{sec} argument), the narrowest likely relaxation mode is stored in
1818@code{fr_subtype}, and that's that.
1819
1820If the symbol is undefined, or in a different section (and therefore movable
1821to an arbitrarily large distance), the largest available relaxation mode is
1822specified, @code{fix_new} is called to produce the relocation record,
1823@code{fr_fix} is increased to include the relocated field (remember, this
1824storage was allocated when @code{frag_var} was called), and @code{frag_wane} is
1825called to convert the frag to an @code{rs_fill} frag with no variant part.
1826Sometimes changing addressing modes may also require rewriting the instruction.
1827It can be accessed via @code{fr_opcode} or @code{fr_fix}.
1828
1829If you generate frags separately for the basic insn opcode and any relaxable
1830operands, do not call @code{fix_new} thinking you can emit fixups for the
1831opcode field from the relaxable frag.  It is not guaranteed to be the same frag.
1832If you need to emit fixups for the opcode field from inspection of the
1833relaxable frag, then you need to generate a common frag for both the basic
1834opcode and relaxable fields, or you need to provide the frag for the opcode to
1835pass to @code{fix_new}.  The latter can be done by passing a pointer to the
1836opcode in the call to @code{frag_var} or @code{frag_variant}, and accessing
1837it via @code{fr_opcode}.  See the @file{tc-vax.c}  and @file{tc-m68k.c} for
1838examples.  It is also possible for  @code{TC_FRAG_TYPE} to include a pointer to
1839the opcode and defining @code{TC_FRAG_INIT} to set the pointer.
1840
1841Sometimes @code{fr_var} is increased instead, and @code{frag_wane} is not
1842called.  I'm not sure, but I think this is to keep @code{fr_fix} referring to
1843an earlier byte, and @code{fr_subtype} set to @code{rs_machine_dependent} so
1844that @code{md_convert_frag} will get called.
1845
1846@node General relaxing
1847@subsection General relaxing
1848
1849If using a simple table is not suitable, you may implement arbitrarily complex
1850relaxation semantics yourself.  For example, the MIPS backend uses this to emit
1851different instruction sequences depending upon the size of the symbol being
1852accessed.
1853
1854When you assemble an instruction that may need relaxation, you should allocate
1855a frag using @code{frag_var} or @code{frag_variant} with a type of
1856@code{rs_machine_dependent}.  You should store some sort of information in the
1857@code{fr_subtype} field so that you can figure out what to do with the frag
1858later.
1859
1860When GAS reaches the end of the input file, it will look through the frags and
1861work out their final sizes.
1862
1863GAS will first call @code{md_estimate_size_before_relax} on each
1864@code{rs_machine_dependent} frag.  This function must return an estimated size
1865for the frag.
1866
1867GAS will then loop over the frags, calling @code{md_relax_frag} on each
1868@code{rs_machine_dependent} frag.  This function should return the change in
1869size of the frag.  GAS will keep looping over the frags until none of the frags
1870changes size.
1871
1872@node Broken words
1873@section Broken words
1874@cindex internals, broken words
1875@cindex broken words
1876
1877Some compilers, including GCC, will sometimes emit switch tables specifying
187816-bit @code{.word} displacements to branch targets, and branch instructions
1879that load entries from that table to compute the target address.  If this is
1880done on a 32-bit machine, there is a chance (at least with really large
1881functions) that the displacement will not fit in 16 bits.  The assembler
1882handles this using a concept called @dfn{broken words}.  This idea is well
1883named, since there is an implied promise that the 16-bit field will in fact
1884hold the specified displacement.
1885
1886If broken word processing is enabled, and a situation like this is encountered,
1887the assembler will insert a jump instruction into the instruction stream, close
1888enough to be reached with the 16-bit displacement.  This jump instruction will
1889transfer to the real desired target address.  Thus, as long as the @code{.word}
1890value really is used as a displacement to compute an address to jump to, the
1891net effect will be correct (minus a very small efficiency cost).  If
1892@code{.word} directives with label differences for values are used for other
1893purposes, however, things may not work properly.  For targets which use broken
1894words, the @samp{-K} option will warn when a broken word is discovered.
1895
1896The broken word code is turned off by the @code{WORKING_DOT_WORD} macro.  It
1897isn't needed if @code{.word} emits a value large enough to contain an address
1898(or, more correctly, any possible difference between two addresses).
1899
1900@node Internal functions
1901@section Internal functions
1902
1903This section describes basic internal functions used by GAS.
1904
1905@menu
1906* Warning and error messages::  Warning and error messages
1907@end menu
1908
1909@node Warning and error messages
1910@subsection Warning and error messages
1911
1912@deftypefun  @{@} int had_warnings (void)
1913@deftypefunx @{@} int had_errors (void)
1914Returns non-zero if any warnings or errors, respectively, have been printed
1915during this invocation.
1916@end deftypefun
1917
1918@deftypefun  @{@} void as_tsktsk (const char *@var{format}, ...)
1919@deftypefunx @{@} void as_warn (const char *@var{format}, ...)
1920@deftypefunx @{@} void as_bad (const char *@var{format}, ...)
1921@deftypefunx @{@} void as_fatal (const char *@var{format}, ...)
1922These functions display messages about something amiss with the input file, or
1923internal problems in the assembler itself.  The current file name and line
1924number are printed, followed by the supplied message, formatted using
1925@code{vfprintf}, and a final newline.
1926
1927An error indicated by @code{as_bad} will result in a non-zero exit status when
1928the assembler has finished.  Calling @code{as_fatal} will result in immediate
1929termination of the assembler process.
1930@end deftypefun
1931
1932@deftypefun @{@} void as_warn_where (char *@var{file}, unsigned int @var{line}, const char *@var{format}, ...)
1933@deftypefunx @{@} void as_bad_where (char *@var{file}, unsigned int @var{line}, const char *@var{format}, ...)
1934These variants permit specification of the file name and line number, and are
1935used when problems are detected when reprocessing information saved away when
1936processing some earlier part of the file.  For example, fixups are processed
1937after all input has been read, but messages about fixups should refer to the
1938original filename and line number that they are applicable to.
1939@end deftypefun
1940
1941@node Test suite
1942@section Test suite
1943@cindex test suite
1944
1945The test suite is kind of lame for most processors.  Often it only checks to
1946see if a couple of files can be assembled without the assembler reporting any
1947errors.  For more complete testing, write a test which either examines the
1948assembler listing, or runs @code{objdump} and examines its output.  For the
1949latter, the TCL procedure @code{run_dump_test} may come in handy.  It takes the
1950base name of a file, and looks for @file{@var{file}.d}.  This file should
1951contain as its initial lines a set of variable settings in @samp{#} comments,
1952in the form:
1953
1954@example
1955        #@var{varname}: @var{value}
1956@end example
1957
1958The @var{varname} may be @code{objdump}, @code{nm}, or @code{as}, in which case
1959it specifies the options to be passed to the specified programs.  Exactly one
1960of @code{objdump} or @code{nm} must be specified, as that also specifies which
1961program to run after the assembler has finished.  If @var{varname} is
1962@code{source}, it specifies the name of the source file; otherwise,
1963@file{@var{file}.s} is used.  If @var{varname} is @code{name}, it specifies the
1964name of the test to be used in the @code{pass} or @code{fail} messages.
1965
1966The non-commented parts of the file are interpreted as regular expressions, one
1967per line.  Blank lines in the @code{objdump} or @code{nm} output are skipped,
1968as are blank lines in the @code{.d} file; the other lines are tested to see if
1969the regular expression matches the program output.  If it does not, the test
1970fails.
1971
1972Note that this means the tests must be modified if the @code{objdump} output
1973style is changed.
1974
1975@bye
1976@c Local Variables:
1977@c fill-column: 79
1978@c End:
1979