9. Target Architecture Definition

GDB's target architecture defines what sort of machine-language programs GDB can work with, and how it works with them.

The target architecture object is implemented as the C structure struct gdbarch *. The structure, and its methods, are generated using the Bourne shell script `gdbarch.sh'.

9.1 Operating System ABI Variant Handling

GDB provides a mechanism for handling variations in OS ABIs. An OS ABI variant may have influence over any number of variables in the target architecture definition. There are two major components in the OS ABI mechanism: sniffers and handlers.

A sniffer examines a file matching a BFD architecture/flavour pair (the architecture may be wildcarded) in an attempt to determine the OS ABI of that file. Sniffers with a wildcarded architecture are considered to be generic, while sniffers for a specific architecture are considered to be specific. A match from a specific sniffer overrides a match from a generic sniffer. Multiple sniffers for an architecture/flavour may exist, in order to differentiate between two different operating systems which use the same basic file format. The OS ABI framework provides a generic sniffer for ELF-format files which examines the EI_OSABI field of the ELF header, as well as note sections known to be used by several operating systems.

A handler is used to fine-tune the gdbarch structure for the selected OS ABI. There may be only one handler for a given OS ABI for each BFD architecture.

The following OS ABI variants are defined in `osabi.h':

GDB_OSABI_UNKNOWN

The ABI of the inferior is unknown. The default gdbarch settings for the architecture will be used.

GDB_OSABI_SVR4

UNIX System V Release 4

GDB_OSABI_HURD

GNU using the Hurd kernel

GDB_OSABI_SOLARIS

Sun Solaris

GDB_OSABI_OSF1

OSF/1, including Digital UNIX and Compaq Tru64 UNIX

GDB_OSABI_LINUX

GNU using the Linux kernel

GDB_OSABI_FREEBSD_AOUT

FreeBSD using the a.out executable format

GDB_OSABI_FREEBSD_ELF

FreeBSD using the ELF executable format

GDB_OSABI_NETBSD_AOUT

NetBSD using the a.out executable format

GDB_OSABI_NETBSD_ELF

NetBSD using the ELF executable format

GDB_OSABI_WINCE

Windows CE

GDB_OSABI_GO32

DJGPP

GDB_OSABI_NETWARE

Novell NetWare

GDB_OSABI_ARM_EABI_V1

ARM Embedded ABI version 1

GDB_OSABI_ARM_EABI_V2

ARM Embedded ABI version 2

GDB_OSABI_ARM_APCS

Generic ARM Procedure Call Standard

Here are the functions that make up the OS ABI framework:

Function: const char *gdbarch_osabi_name (enum gdb_osabi osabi)

Return the name of the OS ABI corresponding to osabi.

Function: void gdbarch_register_osabi (enum bfd_architecture arch, enum gdb_osabi osabi, void (*init_osabi)(struct gdbarch_info info, struct gdbarch *gdbarch))

Register the OS ABI handler specified by init_osabi for the architecture/OS ABI pair specified by arch and osabi.

Function: void gdbarch_register_osabi_sniffer (enum bfd_architecture arch, enum bfd_flavour flavour, enum gdb_osabi (*sniffer)(bfd *abfd))

Register the OS ABI file sniffer specified by sniffer for the BFD architecture/flavour pair specified by arch and flavour. If arch is bfd_arch_unknown, the sniffer is considered to be generic, and is allowed to examine flavour-flavoured files for any architecture.

Function: enum gdb_osabi gdbarch_lookup_osabi (bfd *abfd)

Examine the file described by abfd to determine its OS ABI. The value GDB_OSABI_UNKNOWN is returned if the OS ABI cannot be determined.

Function: void gdbarch_init_osabi (struct gdbarch info info, struct gdbarch *gdbarch, enum gdb_osabi osabi)

Invoke the OS ABI handler corresponding to osabi to fine-tune the gdbarch structure specified by gdbarch. If a handler corresponding to osabi has not been registered for gdbarch's architecture, a warning will be issued and the debugging session will continue with the defaults already established for gdbarch.

9.2 Registers and Memory

GDB's model of the target machine is rather simple. GDB assumes the machine includes a bank of registers and a block of memory. Each register may have a different size.

GDB does not have a magical way to match up with the compiler's idea of which registers are which; however, it is critical that they do match up accurately. The only way to make this work is to get accurate information about the order that the compiler uses, and to reflect that in the REGISTER_NAME and related macros.

GDB can handle big-endian, little-endian, and bi-endian architectures.

9.3 Pointers Are Not Always Addresses

On almost all 32-bit architectures, the representation of a pointer is indistinguishable from the representation of some fixed-length number whose value is the byte address of the object pointed to. On such machines, the words "pointer" and "address" can be used interchangeably. However, architectures with smaller word sizes are often cramped for address space, so they may choose a pointer representation that breaks this identity, and allows a larger code address space.

For example, the Mitsubishi D10V is a 16-bit VLIW processor whose instructions are 32 bits long(3). If the D10V used ordinary byte addresses to refer to code locations, then the processor would only be able to address 64kb of instructions. However, since instructions must be aligned on four-byte boundaries, the low two bits of any valid instruction's byte address are always zero--byte addresses waste two bits. So instead of byte addresses, the D10V uses word addresses--byte addresses shifted right two bits--to refer to code. Thus, the D10V can use 16-bit words to address 256kb of code space.

However, this means that code pointers and data pointers have different forms on the D10V. The 16-bit word 0xC020 refers to byte address 0xC020 when used as a data address, but refers to byte address 0x30080 when used as a code address.

(The D10V also uses separate code and data address spaces, which also affects the correspondence between pointers and addresses, but we're going to ignore that here; this example is already too long.)

To cope with architectures like this--the D10V is not the only one!---GDB tries to distinguish between addresses, which are byte numbers, and pointers, which are the target's representation of an address of a particular type of data. In the example above, 0xC020 is the pointer, which refers to one of the addresses 0xC020 or 0x30080, depending on the type imposed upon it. GDB provides functions for turning a pointer into an address and vice versa, in the appropriate way for the current architecture.

Unfortunately, since addresses and pointers are identical on almost all processors, this distinction tends to bit-rot pretty quickly. Thus, each time you port GDB to an architecture which does distinguish between pointers and addresses, you'll probably need to clean up some architecture-independent code.

Here are functions which convert between pointers and addresses:

Function: CORE_ADDR extract_typed_address (void *buf, struct type *type)

Treat the bytes at buf as a pointer or reference of type type, and return the address it represents, in a manner appropriate for the current architecture. This yields an address GDB can use to read target memory, disassemble, etc. Note that buf refers to a buffer in GDB's memory, not the inferior's.

For example, if the current architecture is the Intel x86, this function extracts a little-endian integer of the appropriate length from buf and returns it. However, if the current architecture is the D10V, this function will return a 16-bit integer extracted from buf, multiplied by four if type is a pointer to a function.

If type is not a pointer or reference type, then this function will signal an internal error.

Function: CORE_ADDR store_typed_address (void *buf, struct type *type, CORE_ADDR addr)

Store the address addr in buf, in the proper format for a pointer of type type in the current architecture. Note that buf refers to a buffer in GDB's memory, not the inferior's.

For example, if the current architecture is the Intel x86, this function stores addr unmodified as a little-endian integer of the appropriate length in buf. However, if the current architecture is the D10V, this function divides addr by four if type is a pointer to a function, and then stores it in buf.

If type is not a pointer or reference type, then this function will signal an internal error.

Function: CORE_ADDR value_as_address (struct value *val)

Assuming that val is a pointer, return the address it represents, as appropriate for the current architecture.

This function actually works on integral values, as well as pointers. For pointers, it performs architecture-specific conversions as described above for extract_typed_address.

Function: CORE_ADDR value_from_pointer (struct type *type, CORE_ADDR addr)

Create and return a value representing a pointer of type type to the address addr, as appropriate for the current architecture. This function performs architecture-specific conversions as described above for store_typed_address.

GDB also provides functions that do the same tasks, but assume that pointers are simply byte addresses; they aren't sensitive to the current architecture, beyond knowing the appropriate endianness.

Function: CORE_ADDR extract_address (void *addr, int len)

Extract a len-byte number from addr in the appropriate endianness for the current architecture, and return it. Note that addr refers to GDB's memory, not the inferior's.

This function should only be used in architecture-specific code; it doesn't have enough information to turn bits into a true address in the appropriate way for the current architecture. If you can, use extract_typed_address instead.

Function: void store_address (void *addr, int len, LONGEST val)

Store val at addr as a len-byte integer, in the appropriate endianness for the current architecture. Note that addr refers to a buffer in GDB's memory, not the inferior's.

This function should only be used in architecture-specific code; it doesn't have enough information to turn a true address into bits in the appropriate way for the current architecture. If you can, use store_typed_address instead.

Here are some macros which architectures can define to indicate the relationship between pointers and addresses. These have default definitions, appropriate for architectures on which all pointers are simple unsigned byte addresses.

Target Macro: CORE_ADDR POINTER_TO_ADDRESS (struct type *type, char *buf)

Assume that buf holds a pointer of type type, in the appropriate format for the current architecture. Return the byte address the pointer refers to.

This function may safely assume that type is either a pointer or a C++ reference type.

Target Macro: void ADDRESS_TO_POINTER (struct type *type, char *buf, CORE_ADDR addr)

Store in buf a pointer of type type representing the address addr, in the appropriate format for the current architecture.

This function may safely assume that type is either a pointer or a C++ reference type.

9.4 Address Classes

Sometimes information about different kinds of addresses is available via the debug information. For example, some programming environments define addresses of several different sizes. If the debug information distinguishes these kinds of address classes through either the size info (e.g, DW_AT_byte_size in DWARF 2) or through an explicit address class attribute (e.g, DW_AT_address_class in DWARF 2), the following macros should be defined in order to disambiguate these types within GDB as well as provide the added information to a GDB user when printing type expressions.

Target Macro: int ADDRESS_CLASS_TYPE_FLAGS (int byte_size, int dwarf2_addr_class)

Returns the type flags needed to construct a pointer type whose size is byte_size and whose address class is dwarf2_addr_class. This function is normally called from within a symbol reader. See `dwarf2read.c'.

Target Macro: char *ADDRESS_CLASS_TYPE_FLAGS_TO_NAME (int type_flags)

Given the type flags representing an address class qualifier, return its name.

Target Macro: int ADDRESS_CLASS_NAME_to_TYPE_FLAGS (int name, int *var{type_flags_ptr})

Given an address qualifier name, set the int refererenced by type_flags_ptr to the type flags for that address class qualifier.

Since the need for address classes is rather rare, none of the address class macros defined by default. Predicate macros are provided to detect when they are defined.

Consider a hypothetical architecture in which addresses are normally 32-bits wide, but 16-bit addresses are also supported. Furthermore, suppose that the DWARF 2 information for this architecture simply uses a DW_AT_byte_size value of 2 to indicate the use of one of these "short" pointers. The following functions could be defined to implement the address class macros:

somearch_address_class_type_flags (int byte_size, int dwarf2_addr_class) { if (byte_size == 2) return TYPE_FLAG_ADDRESS_CLASS_1; else return 0; } static char * somearch_address_class_type_flags_to_name (int type_flags) { if (type_flags & TYPE_FLAG_ADDRESS_CLASS_1) return "short"; else return NULL; } int somearch_address_class_name_to_type_flags (char *name, int *type_flags_ptr) { if (strcmp (name, "short") == 0) { *type_flags_ptr = TYPE_FLAG_ADDRESS_CLASS_1; return 1; } else return 0; }

The qualifier @short is used in GDB's type expressions to indicate the presence of one of these "short" pointers. E.g, if the debug information indicates that short_ptr_var is one of these short pointers, GDB might show the following behavior:

(gdb) ptype short_ptr_var type = int * @short

9.5 Raw and Virtual Register Representations

Maintainer note: This section is pretty much obsolete. The functionality described here has largely been replaced by pseudo-registers and the mechanisms described in Using Different Register and Memory Data Representations. See also Bug Tracking Database and ARI Index for more up-to-date information.

Some architectures use one representation for a value when it lives in a register, but use a different representation when it lives in memory. In GDB's terminology, the raw representation is the one used in the target registers, and the virtual representation is the one used in memory, and within GDB struct value objects.

Maintainer note: Notice that the same mechanism is being used to both convert a register to a struct value and alternative register forms.

For almost all data types on almost all architectures, the virtual and raw representations are identical, and no special handling is needed. However, they do occasionally differ. For example:

• The x86 architecture supports an 80-bit long double type. However, when we store those values in memory, they occupy twelve bytes: the floating-point number occupies the first ten, and the final two bytes are unused. This keeps the values aligned on four-byte boundaries, allowing more efficient access. Thus, the x86 80-bit floating-point type is the raw representation, and the twelve-byte loosely-packed arrangement is the virtual representation.

• Some 64-bit MIPS targets present 32-bit registers to GDB as 64-bit registers, with garbage in their upper bits. GDB ignores the top 32 bits. Thus, the 64-bit form, with garbage in the upper 32 bits, is the raw representation, and the trimmed 32-bit representation is the virtual representation.

In general, the raw representation is determined by the architecture, or GDB's interface to the architecture, while the virtual representation can be chosen for GDB's convenience. GDB's register file, registers, holds the register contents in raw format, and the GDB remote protocol transmits register values in raw format.

Your architecture may define the following macros to request conversions between the raw and virtual format:

Target Macro: int REGISTER_CONVERTIBLE (int reg)

Return non-zero if register number reg's value needs different raw and virtual formats.

You should not use REGISTER_CONVERT_TO_VIRTUAL for a register unless this macro returns a non-zero value for that register.

Target Macro: int REGISTER_RAW_SIZE (int reg)

The size of register number reg's raw value. This is the number of bytes the register will occupy in registers, or in a GDB remote protocol packet.

Target Macro: int REGISTER_VIRTUAL_SIZE (int reg)

The size of register number reg's value, in its virtual format. This is the size a struct value's buffer will have, holding that register's value.

Target Macro: struct type *REGISTER_VIRTUAL_TYPE (int reg)

This is the type of the virtual representation of register number reg. Note that there is no need for a macro giving a type for the register's raw form; once the register's value has been obtained, GDB always uses the virtual form.

Target Macro: void REGISTER_CONVERT_TO_VIRTUAL (int reg, struct type *type, char *from, char *to)

Convert the value of register number reg to type, which should always be REGISTER_VIRTUAL_TYPE (reg). The buffer at from holds the register's value in raw format; the macro should convert the value to virtual format, and place it at to.

Note that REGISTER_CONVERT_TO_VIRTUAL and REGISTER_CONVERT_TO_RAW take their reg and type arguments in different orders.

You should only use REGISTER_CONVERT_TO_VIRTUAL with registers for which the REGISTER_CONVERTIBLE macro returns a non-zero value.

Target Macro: void REGISTER_CONVERT_TO_RAW (struct type *type, int reg, char *from, char *to)

Convert the value of register number reg to type, which should always be REGISTER_VIRTUAL_TYPE (reg). The buffer at from holds the register's value in raw format; the macro should convert the value to virtual format, and place it at to.

Note that REGISTER_CONVERT_TO_VIRTUAL and REGISTER_CONVERT_TO_RAW take their reg and type arguments in different orders.

9.6 Using Different Register and Memory Data Representations

Maintainer's note: The way GDB manipulates registers is undergoing significant change. Many of the macros and functions refered to in this section are likely to be subject to further revision. See A.R. Index and Bug Tracking Database for further information. cagney/2002-05-06.

Some architectures can represent a data object in a register using a form that is different to the objects more normal memory representation. For example:

• The Alpha architecture can represent 32 bit integer values in floating-point registers.

• The x86 architecture supports 80-bit floating-point registers. The long double data type occupies 96 bits in memory but only 80 bits when stored in a register.

In general, the register representation of a data type is determined by the architecture, or GDB's interface to the architecture, while the memory representation is determined by the Application Binary Interface.

For almost all data types on almost all architectures, the two representations are identical, and no special handling is needed. However, they do occasionally differ. Your architecture may define the following macros to request conversions between the register and memory representations of a data type:

Target Macro: int CONVERT_REGISTER_P (int reg)

Return non-zero if the representation of a data value stored in this register may be different to the representation of that same data value when stored in memory.

When non-zero, the macros REGISTER_TO_VALUE and VALUE_TO_REGISTER are used to perform any necessary conversion.

Target Macro: void REGISTER_TO_VALUE (int reg, struct type *type, char *from, char *to)

Convert the value of register number reg to a data object of type type. The buffer at from holds the register's value in raw format; the converted value should be placed in the buffer at to.

Note that REGISTER_TO_VALUE and VALUE_TO_REGISTER take their reg and type arguments in different orders.

You should only use REGISTER_TO_VALUE with registers for which the CONVERT_REGISTER_P macro returns a non-zero value.

Target Macro: void VALUE_TO_REGISTER (struct type *type, int reg, char *from, char *to)

Convert a data value of type type to register number reg' raw format.

Note that REGISTER_TO_VALUE and VALUE_TO_REGISTER take their reg and type arguments in different orders.

You should only use VALUE_TO_REGISTER with registers for which the CONVERT_REGISTER_P macro returns a non-zero value.

Target Macro: void REGISTER_CONVERT_TO_TYPE (int regnum, struct type *type, char *buf)

See `mips-tdep.c'. It does not do what you want.

9.7 Frame Interpretation

9.8 Inferior Call Setup

9.9 Compiler Characteristics

9.10 Target Conditionals

This section describes the macros that you can use to define the target machine.

ADDITIONAL_OPTIONS

ADDITIONAL_OPTION_CASES

ADDITIONAL_OPTION_HANDLER

ADDITIONAL_OPTION_HELP

These are a set of macros that allow the addition of additional command line options to GDB. They are currently used only for the unsupported i960 Nindy target, and should not be used in any other configuration.

ADDR_BITS_REMOVE (addr)

If a raw machine instruction address includes any bits that are not really part of the address, then define this macro to expand into an expression that zeroes those bits in addr. This is only used for addresses of instructions, and even then not in all contexts.

For example, the two low-order bits of the PC on the Hewlett-Packard PA 2.0 architecture contain the privilege level of the corresponding instruction. Since instructions must always be aligned on four-byte boundaries, the processor masks out these bits to generate the actual address of the instruction. ADDR_BITS_REMOVE should filter out these bits with an expression such as ((addr) & ~3).

ADDRESS_CLASS_NAME_TO_TYPE_FLAGS (name, type_flags_ptr)

If name is a valid address class qualifier name, set the int referenced by type_flags_ptr to the mask representing the qualifier and return 1. If name is not a valid address class qualifier name, return 0.

The value for type_flags_ptr should be one of TYPE_FLAG_ADDRESS_CLASS_1, TYPE_FLAG_ADDRESS_CLASS_2, or possibly some combination of these values or'd together. See section Address Classes.

ADDRESS_CLASS_NAME_TO_TYPE_FLAGS_P ()

Predicate which indicates whether ADDRESS_CLASS_NAME_TO_TYPE_FLAGS has been defined.

ADDRESS_CLASS_TYPE_FLAGS (byte_size, dwarf2_addr_class)

Given a pointers byte size (as described by the debug information) and the possible DW_AT_address_class value, return the type flags used by GDB to represent this address class. The value returned should be one of TYPE_FLAG_ADDRESS_CLASS_1, TYPE_FLAG_ADDRESS_CLASS_2, or possibly some combination of these values or'd together. See section Address Classes.

ADDRESS_CLASS_TYPE_FLAGS_P ()

Predicate which indicates whether ADDRESS_CLASS_TYPE_FLAGS has been defined.

ADDRESS_CLASS_TYPE_FLAGS_TO_NAME (type_flags)

Return the name of the address class qualifier associated with the type flags given by type_flags.

ADDRESS_CLASS_TYPE_FLAGS_TO_NAME_P ()

Predicate which indicates whether ADDRESS_CLASS_TYPE_FLAGS_TO_NAME has been defined. See section Address Classes.

ADDRESS_TO_POINTER (type, buf, addr)

Store in buf a pointer of type type representing the address addr, in the appropriate format for the current architecture. This macro may safely assume that type is either a pointer or a C++ reference type. See section Pointers Are Not Always Addresses.

BEFORE_MAIN_LOOP_HOOK

Define this to expand into any code that you want to execute before the main loop starts. Although this is not, strictly speaking, a target conditional, that is how it is currently being used. Note that if a configuration were to define it one way for a host and a different way for the target, GDB will probably not compile, let alone run correctly. This macro is currently used only for the unsupported i960 Nindy target, and should not be used in any other configuration.

BELIEVE_PCC_PROMOTION

Define if the compiler promotes a short or char parameter to an int, but still reports the parameter as its original type, rather than the promoted type.

BELIEVE_PCC_PROMOTION_TYPE

Define this if GDB should believe the type of a short argument when compiled by pcc, but look within a full int space to get its value. Only defined for Sun-3 at present.

BITS_BIG_ENDIAN

Define this if the numbering of bits in the targets does not match the endianness of the target byte order. A value of 1 means that the bits are numbered in a big-endian bit order, 0 means little-endian.

BREAKPOINT

This is the character array initializer for the bit pattern to put into memory where a breakpoint is set. Although it's common to use a trap instruction for a breakpoint, it's not required; for instance, the bit pattern could be an invalid instruction. The breakpoint must be no longer than the shortest instruction of the architecture.

BREAKPOINT has been deprecated in favor of BREAKPOINT_FROM_PC.

BIG_BREAKPOINT

LITTLE_BREAKPOINT

Similar to BREAKPOINT, but used for bi-endian targets.

BIG_BREAKPOINT and LITTLE_BREAKPOINT have been deprecated in favor of BREAKPOINT_FROM_PC.

REMOTE_BREAKPOINT

LITTLE_REMOTE_BREAKPOINT

BIG_REMOTE_BREAKPOINT

Similar to BREAKPOINT, but used for remote targets.

BIG_REMOTE_BREAKPOINT and LITTLE_REMOTE_BREAKPOINT have been deprecated in favor of BREAKPOINT_FROM_PC.

BREAKPOINT_FROM_PC (pcptr, lenptr)

Use the program counter to determine the contents and size of a breakpoint instruction. It returns a pointer to a string of bytes that encode a breakpoint instruction, stores the length of the string to *lenptr, and adjusts pc (if necessary) to point to the actual memory location where the breakpoint should be inserted.

Although it is common to use a trap instruction for a breakpoint, it's not required; for instance, the bit pattern could be an invalid instruction. The breakpoint must be no longer than the shortest instruction of the architecture.

Replaces all the other BREAKPOINT macros.

MEMORY_INSERT_BREAKPOINT (addr, contents_cache)

MEMORY_REMOVE_BREAKPOINT (addr, contents_cache)

Insert or remove memory based breakpoints. Reasonable defaults (default_memory_insert_breakpoint and default_memory_remove_breakpoint respectively) have been provided so that it is not necessary to define these for most architectures. Architectures which may want to define MEMORY_INSERT_BREAKPOINT and MEMORY_REMOVE_BREAKPOINT will likely have instructions that are oddly sized or are not stored in a conventional manner.

It may also be desirable (from an efficiency standpoint) to define custom breakpoint insertion and removal routines if BREAKPOINT_FROM_PC needs to read the target's memory for some reason.

CALL_DUMMY_P

A C expression that is non-zero when the target supports inferior function calls.

CALL_DUMMY_WORDS

Pointer to an array of LONGEST words of data containing host-byte-ordered REGISTER_BYTES sized values that partially specify the sequence of instructions needed for an inferior function call.

Should be deprecated in favor of a macro that uses target-byte-ordered data.

SIZEOF_CALL_DUMMY_WORDS

The size of CALL_DUMMY_WORDS. When CALL_DUMMY_P this must return a positive value. See also CALL_DUMMY_LENGTH.

CALL_DUMMY

A static initializer for CALL_DUMMY_WORDS. Deprecated.

CALL_DUMMY_LOCATION

See the file `inferior.h'.

CALL_DUMMY_STACK_ADJUST

Stack adjustment needed when performing an inferior function call.

Should be deprecated in favor of something like STACK_ALIGN.

CALL_DUMMY_STACK_ADJUST_P

Predicate for use of CALL_DUMMY_STACK_ADJUST.

Should be deprecated in favor of something like STACK_ALIGN.

CANNOT_FETCH_REGISTER (regno)

A C expression that should be nonzero if regno cannot be fetched from an inferior process. This is only relevant if FETCH_INFERIOR_REGISTERS is not defined.

CANNOT_STORE_REGISTER (regno)

A C expression that should be nonzero if regno should not be written to the target. This is often the case for program counters, status words, and other special registers. If this is not defined, GDB will assume that all registers may be written.

DO_DEFERRED_STORES

CLEAR_DEFERRED_STORES

Define this to execute any deferred stores of registers into the inferior, and to cancel any deferred stores.

Currently only implemented correctly for native Sparc configurations?

COERCE_FLOAT_TO_DOUBLE (formal, actual)

Return non-zero if GDB should promote float values to double when calling a non-prototyped function. The argument actual is the type of the value we want to pass to the function. The argument formal is the type of this argument, as it appears in the function's definition. Note that formal may be zero if we have no debugging information for the function, or if we're passing more arguments than are officially declared (for example, varargs). This macro is never invoked if the function definitely has a prototype.

How you should pass arguments to a function depends on whether it was defined in K&R style or prototype style. If you define a function using the K&R syntax that takes a float argument, then callers must pass that argument as a double. If you define the function using the prototype syntax, then you must pass the argument as a float, with no promotion.

Unfortunately, on certain older platforms, the debug info doesn't indicate reliably how each function was defined. A function type's TYPE_FLAG_PROTOTYPED flag may be unset, even if the function was defined in prototype style. When calling a function whose TYPE_FLAG_PROTOTYPED flag is unset, GDB consults the COERCE_FLOAT_TO_DOUBLE macro to decide what to do.

For modern targets, it is proper to assume that, if the prototype flag is unset, that can be trusted: float arguments should be promoted to double. You should use the function standard_coerce_float_to_double to get this behavior.

For some older targets, if the prototype flag is unset, that doesn't tell us anything. So we guess that, if we don't have a type for the formal parameter (i.e., the first argument to COERCE_FLOAT_TO_DOUBLE is null), then we should promote it; otherwise, we should leave it alone. The function default_coerce_float_to_double provides this behavior; it is the default value, for compatibility with older configurations.

int CONVERT_REGISTER_P(regnum)

Return non-zero if register regnum can represent data values in a non-standard form. See section Using Different Register and Memory Data Representations.

DBX_PARM_SYMBOL_CLASS

Hook for the SYMBOL_CLASS of a parameter when decoding DBX symbol information. In the i960, parameters can be stored as locals or as args, depending on the type of the debug record.

DECR_PC_AFTER_BREAK

Define this to be the amount by which to decrement the PC after the program encounters a breakpoint. This is often the number of bytes in BREAKPOINT, though not always. For most targets this value will be 0.

DECR_PC_AFTER_HW_BREAK

Similarly, for hardware breakpoints.

DISABLE_UNSETTABLE_BREAK (addr)

If defined, this should evaluate to 1 if addr is in a shared library in which breakpoints cannot be set and so should be disabled.

PRINT_FLOAT_INFO()

If defined, then the `info float' command will print information about the processor's floating point unit.

print_registers_info (gdbarch, frame, regnum, all)

If defined, pretty print the value of the register regnum for the specified frame. If the value of regnum is -1, pretty print either all registers (all is non zero) or a select subset of registers (all is zero).

The default method prints one register per line, and if all is zero omits floating-point registers.

PRINT_VECTOR_INFO()

If defined, then the `info vector' command will call this function to print information about the processor's vector unit.

By default, the `info vector' command will print all vector registers (the register's type having the vector attribute).

DWARF_REG_TO_REGNUM

Convert DWARF register number into GDB regnum. If not defined, no conversion will be performed.

DWARF2_REG_TO_REGNUM

Convert DWARF2 register number into GDB regnum. If not defined, no conversion will be performed.

ECOFF_REG_TO_REGNUM

Convert ECOFF register number into GDB regnum. If not defined, no conversion will be performed.

END_OF_TEXT_DEFAULT

This is an expression that should designate the end of the text section.

EXTRACT_RETURN_VALUE(type, regbuf, valbuf)

Define this to extract a function's return value of type type from the raw register state regbuf and copy that, in virtual format, into valbuf.

EXTRACT_STRUCT_VALUE_ADDRESS(regbuf)

When defined, extract from the array regbuf (containing the raw register state) the CORE_ADDR at which a function should return its structure value.

If not defined, EXTRACT_RETURN_VALUE is used.

EXTRACT_STRUCT_VALUE_ADDRESS_P()

Predicate for EXTRACT_STRUCT_VALUE_ADDRESS.

FLOAT_INFO

Deprecated in favor of PRINT_FLOAT_INFO.

FP_REGNUM

If the virtual frame pointer is kept in a register, then define this macro to be the number (greater than or equal to zero) of that register.

This should only need to be defined if TARGET_READ_FP is not defined.

FRAMELESS_FUNCTION_INVOCATION(fi)

Define this to an expression that returns 1 if the function invocation represented by fi does not have a stack frame associated with it. Otherwise return 0.

frame_align (address)

Define this to adjust address so that it meets the alignment requirements for the start of a new stack frame. A stack frame's alignment requirements are typically stronger than a target processors stack alignment requirements (see STACK_ALIGN).

This function is used to ensure that, when creating a dummy frame, both the initial stack pointer and (if needed) the address of the return value are correctly aligned.

Unlike STACK_ALIGN, this function always adjusts the address in the direction of stack growth.

By default, no frame based stack alignment is performed.

FRAME_ARGS_ADDRESS_CORRECT

See `stack.c'.

FRAME_CHAIN(frame)

Given frame, return a pointer to the calling frame.

FRAME_CHAIN_VALID(chain, thisframe)

Define this to be an expression that returns zero if the given frame is an outermost frame, with no caller, and nonzero otherwise. Several common definitions are available:

• file_frame_chain_valid is nonzero if the chain pointer is nonzero and given frame's PC is not inside the startup file (such as `crt0.o').

• func_frame_chain_valid is nonzero if the chain pointer is nonzero and the given frame's PC is not in main or a known entry point function (such as _start).

• generic_file_frame_chain_valid and generic_func_frame_chain_valid are equivalent implementations for targets using generic dummy frames.

FRAME_INIT_SAVED_REGS(frame)

See `frame.h'. Determines the address of all registers in the current stack frame storing each in frame->saved_regs. Space for frame->saved_regs shall be allocated by FRAME_INIT_SAVED_REGS using either frame_saved_regs_zalloc or frame_obstack_alloc.

FRAME_FIND_SAVED_REGS and EXTRA_FRAME_INFO are deprecated.

FRAME_NUM_ARGS (fi)

For the frame described by fi return the number of arguments that are being passed. If the number of arguments is not known, return -1.

FRAME_SAVED_PC(frame)

Given frame, return the pc saved there. This is the return address.

FUNCTION_EPILOGUE_SIZE

For some COFF targets, the x_sym.x_misc.x_fsize field of the function end symbol is 0. For such targets, you must define FUNCTION_EPILOGUE_SIZE to expand into the standard size of a function's epilogue.

FUNCTION_START_OFFSET

An integer, giving the offset in bytes from a function's address (as used in the values of symbols, function pointers, etc.), and the function's first genuine instruction.

This is zero on almost all machines: the function's address is usually the address of its first instruction. However, on the VAX, for example, each function starts with two bytes containing a bitmask indicating which registers to save upon entry to the function. The VAX call instructions check this value, and save the appropriate registers automatically. Thus, since the offset from the function's address to its first instruction is two bytes, FUNCTION_START_OFFSET would be 2 on the VAX.

GCC_COMPILED_FLAG_SYMBOL

GCC2_COMPILED_FLAG_SYMBOL

If defined, these are the names of the symbols that GDB will look for to detect that GCC compiled the file. The default symbols are gcc_compiled. and gcc2_compiled., respectively. (Currently only defined for the Delta 68.)

GDB_MULTI_ARCH

If defined and non-zero, enables support for multiple architectures within GDB.

This support can be enabled at two levels. At level one, only definitions for previously undefined macros are provided; at level two, a multi-arch definition of all architecture dependent macros will be defined.

GDB_TARGET_IS_HPPA

This determines whether horrible kludge code in `dbxread.c' and `partial-stab.h' is used to mangle multiple-symbol-table files from HPPA's. This should all be ripped out, and a scheme like `elfread.c' used instead.

GET_LONGJMP_TARGET

For most machines, this is a target-dependent parameter. On the DECstation and the Iris, this is a native-dependent parameter, since the header file `setjmp.h' is needed to define it.

This macro determines the target PC address that longjmp will jump to, assuming that we have just stopped at a longjmp breakpoint. It takes a CORE_ADDR * as argument, and stores the target PC value through this pointer. It examines the current state of the machine as needed.

GET_SAVED_REGISTER

Define this if you need to supply your own definition for the function get_saved_register.

IBM6000_TARGET

Shows that we are configured for an IBM RS/6000 target. This conditional should be eliminated (FIXME) and replaced by feature-specific macros. It was introduced in a haste and we are repenting at leisure.

I386_USE_GENERIC_WATCHPOINTS

An x86-based target can define this to use the generic x86 watchpoint support; see I386_USE_GENERIC_WATCHPOINTS.

SYMBOLS_CAN_START_WITH_DOLLAR

Some systems have routines whose names start with `$'. Giving this macro a non-zero value tells GDB's expression parser to check for such routines when parsing tokens that begin with `$'.

On HP-UX, certain system routines (millicode) have names beginning with `$' or `$$'. For example, $$dyncall is a millicode routine that handles inter-space procedure calls on PA-RISC.

INIT_EXTRA_FRAME_INFO (fromleaf, frame)

If additional information about the frame is required this should be stored in frame->extra_info. Space for frame->extra_info is allocated using frame_obstack_alloc.

INIT_FRAME_PC (fromleaf, prev)

This is a C statement that sets the pc of the frame pointed to by prev. [By default...]

INNER_THAN (lhs, rhs)

Returns non-zero if stack address lhs is inner than (nearer to the stack top) stack address rhs. Define this as lhs < rhs if the target's stack grows downward in memory, or lhs > rsh if the stack grows upward.

gdbarch_in_function_epilogue_p (gdbarch, pc)

Returns non-zero if the given pc is in the epilogue of a function. The epilogue of a function is defined as the part of a function where the stack frame of the function already has been destroyed up to the final `return from function call' instruction.

SIGTRAMP_START (pc)

SIGTRAMP_END (pc)

Define these to be the start and end address of the sigtramp for the given pc. On machines where the address is just a compile time constant, the macro expansion will typically just ignore the supplied pc.

IN_SOLIB_CALL_TRAMPOLINE (pc, name)

Define this to evaluate to nonzero if the program is stopped in the trampoline that connects to a shared library.

IN_SOLIB_RETURN_TRAMPOLINE (pc, name)

Define this to evaluate to nonzero if the program is stopped in the trampoline that returns from a shared library.

IN_SOLIB_DYNSYM_RESOLVE_CODE (pc)

Define this to evaluate to nonzero if the program is stopped in the dynamic linker.

SKIP_SOLIB_RESOLVER (pc)

Define this to evaluate to the (nonzero) address at which execution should continue to get past the dynamic linker's symbol resolution function. A zero value indicates that it is not important or necessary to set a breakpoint to get through the dynamic linker and that single stepping will suffice.

INTEGER_TO_ADDRESS (type, buf)

Define this when the architecture needs to handle non-pointer to address conversions specially. Converts that value to an address according to the current architectures conventions.

Pragmatics: When the user copies a well defined expression from their source code and passes it, as a parameter, to GDB's print command, they should get the same value as would have been computed by the target program. Any deviation from this rule can cause major confusion and annoyance, and needs to be justified carefully. In other words, GDB doesn't really have the freedom to do these conversions in clever and useful ways. It has, however, been pointed out that users aren't complaining about how GDB casts integers to pointers; they are complaining that they can't take an address from a disassembly listing and give it to x/i. Adding an architecture method like INTEGER_TO_ADDRESS certainly makes it possible for GDB to "get it right" in all circumstances.

See section Pointers Are Not Always Addresses.

IS_TRAPPED_INTERNALVAR (name)

This is an ugly hook to allow the specification of special actions that should occur as a side-effect of setting the value of a variable internal to GDB. Currently only used by the h8500. Note that this could be either a host or target conditional.

NEED_TEXT_START_END

Define this if GDB should determine the start and end addresses of the text section. (Seems dubious.)

NO_HIF_SUPPORT

(Specific to the a29k.)

POINTER_TO_ADDRESS (type, buf)

Assume that buf holds a pointer of type type, in the appropriate format for the current architecture. Return the byte address the pointer refers to. See section Pointers Are Not Always Addresses.

REGISTER_CONVERTIBLE (reg)

Return non-zero if reg uses different raw and virtual formats. See section Raw and Virtual Register Representations.

REGISTER_TO_VALUE(regnum, type, from, to)

Convert the raw contents of register regnum into a value of type type. See section Using Different Register and Memory Data Representations.

REGISTER_RAW_SIZE (reg)

Return the raw size of reg; defaults to the size of the register's virtual type. See section Raw and Virtual Register Representations.

REGISTER_VIRTUAL_SIZE (reg)

Return the virtual size of reg; defaults to the size of the register's virtual type. Return the virtual size of reg. See section Raw and Virtual Register Representations.

REGISTER_VIRTUAL_TYPE (reg)

Return the virtual type of reg. See section Raw and Virtual Register Representations.

REGISTER_CONVERT_TO_VIRTUAL(reg, type, from, to)

Convert the value of register reg from its raw form to its virtual form. See section Raw and Virtual Register Representations.

REGISTER_CONVERT_TO_RAW(type, reg, from, to)

Convert the value of register reg from its virtual form to its raw form. See section Raw and Virtual Register Representations.

RETURN_VALUE_ON_STACK(type)

Return non-zero if values of type TYPE are returned on the stack, using the "struct convention" (i.e., the caller provides a pointer to a buffer in which the callee should store the return value). This controls how the `finish' command finds a function's return value, and whether an inferior function call reserves space on the stack for the return value.

The full logic GDB uses here is kind of odd.

• If the type being returned by value is not a structure, union, or array, and RETURN_VALUE_ON_STACK returns zero, then GDB concludes the value is not returned using the struct convention.

• Otherwise, GDB calls USE_STRUCT_CONVENTION (see below). If that returns non-zero, GDB assumes the struct convention is in use.

In other words, to indicate that a given type is returned by value using the struct convention, that type must be either a struct, union, array, or something RETURN_VALUE_ON_STACK likes, and something that USE_STRUCT_CONVENTION likes.

Note that, in C and C++, arrays are never returned by value. In those languages, these predicates will always see a pointer type, never an array type. All the references above to arrays being returned by value apply only to other languages.

SOFTWARE_SINGLE_STEP_P()

Define this as 1 if the target does not have a hardware single-step mechanism. The macro SOFTWARE_SINGLE_STEP must also be defined.

SOFTWARE_SINGLE_STEP(signal, insert_breapoints_p)

A function that inserts or removes (depending on insert_breapoints_p) breakpoints at each possible destinations of the next instruction. See `sparc-tdep.c' and `rs6000-tdep.c' for examples.

SOFUN_ADDRESS_MAYBE_MISSING

Somebody clever observed that, the more actual addresses you have in the debug information, the more time the linker has to spend relocating them. So whenever there's some other way the debugger could find the address it needs, you should omit it from the debug info, to make linking faster.

SOFUN_ADDRESS_MAYBE_MISSING indicates that a particular set of hacks of this sort are in use, affecting N_SO and N_FUN entries in stabs-format debugging information. N_SO stabs mark the beginning and ending addresses of compilation units in the text segment. N_FUN stabs mark the starts and ends of functions.

SOFUN_ADDRESS_MAYBE_MISSING means two things:

• N_FUN stabs have an address of zero. Instead, you should find the addresses where the function starts by taking the function name from the stab, and then looking that up in the minsyms (the linker/assembler symbol table). In other words, the stab has the name, and the linker/assembler symbol table is the only place that carries the address.

• N_SO stabs have an address of zero, too. You just look at the N_FUN stabs that appear before and after the N_SO stab, and guess the starting and ending addresses of the compilation unit from them.

PCC_SOL_BROKEN

(Used only in the Convex target.)

PC_IN_CALL_DUMMY

See `inferior.h'.

PC_IN_SIGTRAMP (pc, name)

The sigtramp is a routine that the kernel calls (which then calls the signal handler). On most machines it is a library routine that is linked into the executable.

This function, given a program counter value in pc and the (possibly NULL) name of the function in which that pc resides, returns nonzero if the pc and/or name show that we are in sigtramp.

PC_LOAD_SEGMENT

If defined, print information about the load segment for the program counter. (Defined only for the RS/6000.)

PC_REGNUM

If the program counter is kept in a register, then define this macro to be the number (greater than or equal to zero) of that register.

This should only need to be defined if TARGET_READ_PC and TARGET_WRITE_PC are not defined.

NPC_REGNUM

The number of the "next program counter" register, if defined.

PARM_BOUNDARY

If non-zero, round arguments to a boundary of this many bits before pushing them on the stack.

PRINT_REGISTER_HOOK (regno)

If defined, this must be a function that prints the contents of the given register to standard output.

PRINT_TYPELESS_INTEGER

This is an obscure substitute for print_longest that seems to have been defined for the Convex target.

PROCESS_LINENUMBER_HOOK

A hook defined for XCOFF reading.

PROLOGUE_FIRSTLINE_OVERLAP

(Only used in unsupported Convex configuration.)

PS_REGNUM

If defined, this is the number of the processor status register. (This definition is only used in generic code when parsing "$ps".)

POP_FRAME

Used in `call_function_by_hand' to remove an artificial stack frame and in `return_command' to remove a real stack frame.

PUSH_ARGUMENTS (nargs, args, sp, struct_return, struct_addr)

Define this to push arguments onto the stack for inferior function call. Returns the updated stack pointer value.

PUSH_DUMMY_FRAME

Used in `call_function_by_hand' to create an artificial stack frame.

REGISTER_BYTES

The total amount of space needed to store GDB's copy of the machine's register state.

REGISTER_NAME(i)

Return the name of register i as a string. May return NULL or NUL to indicate that register i is not valid.

REGISTER_NAMES

Deprecated in favor of REGISTER_NAME.

REG_STRUCT_HAS_ADDR (gcc_p, type)

Define this to return 1 if the given type will be passed by pointer rather than directly.

SAVE_DUMMY_FRAME_TOS (sp)

Used in `call_function_by_hand' to notify the target dependent code of the top-of-stack value that will be passed to the the inferior code. This is the value of the SP after both the dummy frame and space for parameters/results have been allocated on the stack.

SDB_REG_TO_REGNUM

Define this to convert sdb register numbers into GDB regnums. If not defined, no conversion will be done.

SKIP_PERMANENT_BREAKPOINT

Advance the inferior's PC past a permanent breakpoint. GDB normally steps over a breakpoint by removing it, stepping one instruction, and re-inserting the breakpoint. However, permanent breakpoints are hardwired into the inferior, and can't be removed, so this strategy doesn't work. Calling SKIP_PERMANENT_BREAKPOINT adjusts the processor's state so that execution will resume just after the breakpoint. This macro does the right thing even when the breakpoint is in the delay slot of a branch or jump.

SKIP_PROLOGUE (pc)

A C expression that returns the address of the "real" code beyond the function entry prologue found at pc.

SKIP_TRAMPOLINE_CODE (pc)

If the target machine has trampoline code that sits between callers and the functions being called, then define this macro to return a new PC that is at the start of the real function.

SP_REGNUM

If the stack-pointer is kept in a register, then define this macro to be the number (greater than or equal to zero) of that register.

This should only need to be defined if TARGET_WRITE_SP and TARGET_WRITE_SP are not defined.

STAB_REG_TO_REGNUM

Define this to convert stab register numbers (as gotten from `r' declarations) into GDB regnums. If not defined, no conversion will be done.

STACK_ALIGN (addr)

Define this to increase addr so that it meets the alignment requirements for the processor's stack.

Unlike frame_align, this function always adjusts addr upwards.

By default, no stack alignment is performed.

STEP_SKIPS_DELAY (addr)

Define this to return true if the address is of an instruction with a delay slot. If a breakpoint has been placed in the instruction's delay slot, GDB will single-step over that instruction before resuming normally. Currently only defined for the Mips.

STORE_RETURN_VALUE (type, regcache, valbuf)

A C expression that writes the function return value, found in valbuf, into the regcache. type is the type of the value that is to be returned.

SUN_FIXED_LBRAC_BUG

(Used only for Sun-3 and Sun-4 targets.)

SYMBOL_RELOADING_DEFAULT

The default value of the "symbol-reloading" variable. (Never defined in current sources.)

TARGET_CHAR_BIT

Number of bits in a char; defaults to 8.

TARGET_CHAR_SIGNED

Non-zero if char is normally signed on this architecture; zero if it should be unsigned.

The ISO C standard requires the compiler to treat char as equivalent to either signed char or unsigned char; any character in the standard execution set is supposed to be positive. Most compilers treat char as signed, but char is unsigned on the IBM S/390, RS6000, and PowerPC targets.

TARGET_COMPLEX_BIT

Number of bits in a complex number; defaults to 2 * TARGET_FLOAT_BIT.

At present this macro is not used.

TARGET_DOUBLE_BIT

Number of bits in a double float; defaults to 8 * TARGET_CHAR_BIT.

TARGET_DOUBLE_COMPLEX_BIT

Number of bits in a double complex; defaults to 2 * TARGET_DOUBLE_BIT.

At present this macro is not used.

TARGET_FLOAT_BIT

Number of bits in a float; defaults to 4 * TARGET_CHAR_BIT.

TARGET_INT_BIT

Number of bits in an integer; defaults to 4 * TARGET_CHAR_BIT.

TARGET_LONG_BIT

Number of bits in a long integer; defaults to 4 * TARGET_CHAR_BIT.

TARGET_LONG_DOUBLE_BIT

Number of bits in a long double float; defaults to 2 * TARGET_DOUBLE_BIT.

TARGET_LONG_LONG_BIT

Number of bits in a long long integer; defaults to 2 * TARGET_LONG_BIT.

TARGET_PTR_BIT

Number of bits in a pointer; defaults to TARGET_INT_BIT.

TARGET_SHORT_BIT

Number of bits in a short integer; defaults to 2 * TARGET_CHAR_BIT.

TARGET_READ_PC

TARGET_WRITE_PC (val, pid)

TARGET_READ_SP

TARGET_WRITE_SP

TARGET_READ_FP

These change the behavior of read_pc, write_pc, read_sp, write_sp and read_fp. For most targets, these may be left undefined. GDB will call the read and write register functions with the relevant _REGNUM argument.

These macros are useful when a target keeps one of these registers in a hard to get at place; for example, part in a segment register and part in an ordinary register.

TARGET_VIRTUAL_FRAME_POINTER(pc, regp, offsetp)

Returns a (register, offset) pair representing the virtual frame pointer in use at the code address pc. If virtual frame pointers are not used, a default definition simply returns FP_REGNUM, with an offset of zero.

TARGET_HAS_HARDWARE_WATCHPOINTS

If non-zero, the target has support for hardware-assisted watchpoints. See section watchpoints, for more details and other related macros.

TARGET_PRINT_INSN (addr, info)

This is the function used by GDB to print an assembly instruction. It prints the instruction at address addr in debugged memory and returns the length of the instruction, in bytes. If a target doesn't define its own printing routine, it defaults to an accessor function for the global pointer tm_print_insn. This usually points to a function in the opcodes library (see section Opcodes). info is a structure (of type disassemble_info) defined in `include/dis-asm.h' used to pass information to the instruction decoding routine.

USE_STRUCT_CONVENTION (gcc_p, type)

If defined, this must be an expression that is nonzero if a value of the given type being returned from a function must have space allocated for it on the stack. gcc_p is true if the function being considered is known to have been compiled by GCC; this is helpful for systems where GCC is known to use different calling convention than other compilers.

VALUE_TO_REGISTER(type, regnum, from, to)

Convert a value of type type into the raw contents of register regnum's. See section Using Different Register and Memory Data Representations.

VARIABLES_INSIDE_BLOCK (desc, gcc_p)

For dbx-style debugging information, if the compiler puts variable declarations inside LBRAC/RBRAC blocks, this should be defined to be nonzero. desc is the value of n_desc from the N_RBRAC symbol, and gcc_p is true if GDB has noticed the presence of either the GCC_COMPILED_SYMBOL or the GCC2_COMPILED_SYMBOL. By default, this is 0.

OS9K_VARIABLES_INSIDE_BLOCK (desc, gcc_p)

Similarly, for OS/9000. Defaults to 1.

Motorola M68K target conditionals.

BPT_VECTOR

Define this to be the 4-bit location of the breakpoint trap vector. If not defined, it will default to 0xf.

REMOTE_BPT_VECTOR

Defaults to 1.

NAME_OF_MALLOC

A string containing the name of the function to call in order to allocate some memory in the inferior. The default value is "malloc".

9.11 Adding a New Target

The following files add a target to GDB:

`gdb/config/arch/ttt.mt'

Contains a Makefile fragment specific to this target. Specifies what object files are needed for target ttt, by defining `TDEPFILES=...' and `TDEPLIBS=...'. Also specifies the header file which describes ttt, by defining `TM_FILE= tm-ttt.h'.

You can also define `TM_CFLAGS', `TM_CLIBS', `TM_CDEPS', but these are now deprecated, replaced by autoconf, and may go away in future versions of GDB.

`gdb/ttt-tdep.c'

Contains any miscellaneous code required for this target machine. On some machines it doesn't exist at all. Sometimes the macros in `tm-ttt.h' become very complicated, so they are implemented as functions here instead, and the macro is simply defined to call the function. This is vastly preferable, since it is easier to understand and debug.

`gdb/arch-tdep.c'

`gdb/arch-tdep.h'

This often exists to describe the basic layout of the target machine's processor chip (registers, stack, etc.). If used, it is included by `ttt-tdep.h'. It can be shared among many targets that use the same processor.

`gdb/config/arch/tm-ttt.h'

(`tm.h' is a link to this file, created by configure). Contains macro definitions about the target machine's registers, stack frame format and instructions.

New targets do not need this file and should not create it.

`gdb/config/arch/tm-arch.h'

This often exists to describe the basic layout of the target machine's processor chip (registers, stack, etc.). If used, it is included by `tm-ttt.h'. It can be shared among many targets that use the same processor.

New targets do not need this file and should not create it.

If you are adding a new operating system for an existing CPU chip, add a `config/tm-os.h' file that describes the operating system facilities that are unusual (extra symbol table info; the breakpoint instruction needed; etc.). Then write a `arch/tm-os.h' that just #includes `tm-arch.h' and `config/tm-os.h'.

9.12 Converting an existing Target Architecture to Multi-arch

This section describes the current accepted best practice for converting an existing target architecture to the multi-arch framework.

The process consists of generating, testing, posting and committing a sequence of patches. Each patch must contain a single change, for instance:

• Directly convert a group of functions into macros (the conversion does not change the behavior of any of the functions).

• Replace a non-multi-arch with a multi-arch mechanism (e.g., FRAME_INFO).

• Enable multi-arch level one.

• Delete one or more files.

There isn't a size limit on a patch, however, a developer is strongly encouraged to keep the patch size down.

Since each patch is well defined, and since each change has been tested and shows no regressions, the patches are considered fairly obvious. Such patches, when submitted by developers listed in the `MAINTAINERS' file, do not need approval. Occasional steps in the process may be more complicated and less clear. The developer is expected to use their judgment and is encouraged to seek advice as needed.

9.12.1 Preparation

The first step is to establish control. Build (with `-Werror' enabled) and test the target so that there is a baseline against which the debugger can be compared.

At no stage can the test results regress or GDB stop compiling with `-Werror'.

9.12.2 Add the multi-arch initialization code

The objective of this step is to establish the basic multi-arch framework. It involves

• The addition of a arch_gdbarch_init function(4) that creates the architecture:

  static struct gdbarch * d10v_gdbarch_init (info, arches) struct gdbarch_info info; struct gdbarch_list *arches; { struct gdbarch *gdbarch; /* there is only one d10v architecture */ if (arches != NULL) return arches->gdbarch; gdbarch = gdbarch_alloc (&info, NULL); return gdbarch; }

• A per-architecture dump function to print any architecture specific information:

  static void mips_dump_tdep (struct gdbarch *current_gdbarch, struct ui_file *file) { ... code to print architecture specific info ... }

• A change to _initialize_arch_tdep to register this new architecture:

  void _initialize_mips_tdep (void) { gdbarch_register (bfd_arch_mips, mips_gdbarch_init, mips_dump_tdep);

• Add the macro GDB_MULTI_ARCH, defined as 0 (zero), to the file
  `config/arch/tm-arch.h'.

9.12.3 Update multi-arch incompatible mechanisms

Some mechanisms do not work with multi-arch. They include:

EXTRA_FRAME_INFO

Delete.

FRAME_FIND_SAVED_REGS

Replaced with FRAME_INIT_SAVED_REGS

At this stage you could also consider converting the macros into functions.

9.12.4 Prepare for multi-arch level to one

Temporally set GDB_MULTI_ARCH to GDB_MULTI_ARCH_PARTIAL and then build and start GDB (the change should not be committed). GDB may not build, and once built, it may die with an internal error listing the architecture methods that must be provided.

Fix any build problems (patch(es)).

Convert all the architecture methods listed, which are only macros, into functions (patch(es)).

Update arch_gdbarch_init to set all the missing architecture methods and wrap the corresponding macros in #if !GDB_MULTI_ARCH (patch(es)).

9.12.5 Set multi-arch level one

Change the value of GDB_MULTI_ARCH to GDB_MULTI_ARCH_PARTIAL (a single patch).

Any problems with throwing "the switch" should have been fixed already.

9.12.6 Convert remaining macros

Suggest converting macros into functions (and setting the corresponding architecture method) in small batches.

9.12.7 Set multi-arch level to two

This should go smoothly.

9.12.8 Delete the TM file

The `tm-arch.h' can be deleted. `arch.mt' and `configure.in' updated.