changelog

gdb-6.0 linux 下的调试工具

	* configure.in (enable-sim-endain): Correct typo in usage (from
 	Erik Landry <[e-mail]>).
	* configure: Re-generate.

Fri Mar 14 18:23:02 1997  Andrew Cagney  <[e-mail]>

	* Makefile.in (targ-vals.def): Re-do rule so that it works with
 	FreeBSD's make.  Didn't like $<.

Thu Mar 13 12:55:48 1997  Doug Evans  <[e-mail]>

	* sim_calls.c (sim_open): New SIM_DESC result.  Argument is now
	in argv form.
	(other sim_*): New SIM_DESC argument.

Thu Feb 13 10:35:14 1997  Andrew Cagney  <[e-mail]>

	* ppc-opcode-simple-array, ppc-opcode-simple-goto,
 	ppc-opcode-simple-switch, ppc-opcode-complex-array,
 	ppc-opcode-complex-goto, ppc-opcode-complex-switch,
 	ppc-opcode-jump, ppc-opcode-goto, ppc-opcode-flat: Delete,
 	superseeded by --sim-decode-mechanism option.

	* ppc-opcode-simple, dc-simple: Rename to be 8.3
	* ppc-opcode-complex, dc-complex: Ditto.
	* ppc-opcode-stupid, dc-stupid: Ditto.
	* ppc-opcode-test-1, dc-test.01: Ditto.
	* ppc-opcode-test-2, dc-test.02: Ditto.

	* configure.in (--enable-sim-opcode): Change prefix to dc- instead
 	of ppc-opcode-.

Wed Feb 12 19:33:45 1997  Andrew Cagney  <[e-mail]>

	* Many of the ppc-opcode-* files are identical baring the type of
 	lookup table.  Instead of having multiple tables, igen can do this
	via an additional option.
	
	* ld-decode.h, ld-decode.c (force_decode_gen_type): New function,
 	allow the type of generated table specified in the decode file to
 	be overridden.

	* ld-decode.c (load_decode_table): Allow the table type to be
 	overridden.

	* igen.c (main): Add -T <mechanism> option so that an overriding
 	instruction decode mechanism can be specified.

	* configure.in: New option --sim-decode-mechanism to control
 	igen's new -T <mechanism> flag.
	* Makefile.in (IGEN_FLAGS): Add IGEN_IDECODE_MECHANISM set by the
 	configure script.
	* configure: Regenerate.
	
Tue Feb 11 13:49:10 1997  Michael Meissner  <[e-mail]>

	* events.c (event_queue_create): Don't use NULL to initialize an
	integer field.
	(even_queue_{init,schedule_after_signal,tick}): Conditionalize use
	of sigprocmask to appropriate autoconf test.

	* main.c ({cntrl_c,main}): Use RETSIGTYPE for signal return type,
	don't assume void.
	* sim_calls.c (sim_{ctrl_c,resume}): Ditto.

	* Makefile.in (callback.o): Define HAVE_CONFIG_H, so callback.c
	includes our config.h.

Tue Feb  4 13:42:59 1997  Doug Evans  <[e-mail]>

	* configure.in: Fix typo in test for callback.c.
	* configure: Regenerated.

Fri Feb  7 10:04:25 1997  Andrew Cagney  <[e-mail]>

	* emul_chirp.c (emul_chirp_create): Handle a virtbase of -1 being
 	found in the device tree.

Wed Feb  5 10:56:27 1997  Andrew Cagney  <[e-mail]>

	* Property create/initialization still wasn't correctly ordered.
  	Should be delaying everything related to ihandle creation until
 	after the rest of the tree has been established.
	
	* device.c (device_find_ihandle_runtime_property): Update.
	(device_add_ihandle_runtime_property): Update.

	* tree.c (parse_ihandle_property): Delay lookup of the device to
 	be opened until the ihandle initialization phase.
	* tree.c (print_properties): Update.

Wed Feb  5 10:56:27 1997  Andrew Cagney  <[e-mail]>
	
	* gen-icache.c (print_icache_extraction): Add a reason parameter.
  	Augment each extracted field with a comment citing the codes
 	origin.  Should simplify tracking down incorrect cache
 	extractions.

Tue Feb  4 17:44:51 1997  Andrew Cagney  <[e-mail]>

	* gen-icache.c: Generalize code handling XXX_is_NNN so that it
 	works for normal and boolean table entries.

	* psim.c (psim_write_memory): last_cpu == -1 or nr_cpus is now
 	valid.  Handle this just like *_{read,write}_register now handles
 	it.

Mon Feb  3 17:18:16 1997  Andrew Cagney  <[e-mail]>

	* events.c (insert_event_entry): Correct loop termination
 	assertions.

Fri Jan 31 16:20:26 1997  Andrew Cagney  <[e-mail]>

	* psim.c (psim_options): Add new option -c for max-iterations or
 	count.
	(psim_usage): Document.
	(psim_max_iterations_exceeded): New function, abort simulation if
 	max iterations exceeded.

	* gen-idecode.c: Re-work the table lookup code so that it assumes
 	that the entry is a leaf by default.  Simplify the boolean table
 	entry code so that it involves a mask + test instead of shift +
 	shift + mask + test.

	* gen-idecode.c: Correct generated igen body so that it no drops
 	or doubles clock interrupts.

	
Thu Jan 30 11:23:20 1997  Andrew Cagney  <[e-mail]>

	* Makefile.in (BUILT_SRC_WO_CONFIG): Change targ-vals.* to
 	@sim_targ_vals@

	* configure.in (sim_callback, sim_targ_vals): Set sim_targ_valls
 	if common callback is present.

Wed Jan 29 12:32:41 1997  Michael Meissner  <[e-mail]>

	* configure.in (sim_callback): If the gdb is post 4.16, configure
	callback support from the common directory.
	* configure: Regenerate.

	* Makefile.in (BUILT_SRC_WO_CONFIG): Add targ-vals.{h,def} and
	targ-map.c.
	(GDB_OBJ): Add callback support configured in.
	(gentmap,targ-vals.def): Build from common directory.
	(targ-vals.h,targ-map.c): Build by running gentmap.
	(callback.o): Build from source in common directory.
	(targ-map.o): Add dependency.
	(clean): Remove gentmap.

Wed Jan 29 12:14:19 1997  Andrew Cagney  <[e-mail]>

	* igen wasn't aborting if the opcode table contained no valid
 	fields.
	
	* misc.c (name2i): Possibly abort if an invalid name is
 	encountered.
	* ld-decode.c: Abort if the table type isn't found.

Wed Jan 29 12:14:19 1997  Andrew Cagney  <[e-mail]>

	* When performance monitoring is disabled, it is still possible to
 	determine the simulation speed by looking at the number of elapsed
 	ticks recorded by the event queue.
	
	* psim.c (psim_write_register, psim_read_register): Force the cpu
 	to zero when it is either of `-1' or `nr_cpus'.  In both cases the
 	next cpu would be zero any way.

	* mon.c (mon_print_info): If possible, print the system cycle
 	performance.  This is an indication of the number of instructions
 	per second.

Wed Jan 29 12:14:19 1997  Andrew Cagney  <[e-mail]>

	* The code to allow an event queue to be updated during a signal
 	was missing.  For main.c, a cntrl-c simulation termination wasn't
 	handled cleanly.
	
	* The simulation would not correctly restart if an event requested
 	that the simulation be halted.
	
	* psim.c (psim_options): Add hack to -i option to optionally
 	include a level vis -i2.
	(psim_usage): Document.

	* main.c (cntrl_c, cntrl_c_simulation): New functions.  When a
 	cntrl-c occures schedule an event to halt the simulation.
	(main): Catch CNTRL-C signals with the function cntrl_c.

	* events.c (event_queue_process): Mask interrupts while
 	manipulating the async event queue.
	(event_queue_init): Ditto.
	(event_queue_schedule_after_signal): Ditto.

	* events.c (event_queue_process): Mark the event queue as being in
 	the processing state when processing has started.  Adjust code
	so that it is tolerant of halts.
	(event_queue_init): Start the event queue out with processing
 	false.
	(event_queue_tick): Check that processing isn't still being
 	performed.
	
	* gen-idecode.c (print_run_until_stop_body): Call
 	event_queue_process_events to clear possibly pending events before
 	starting a simulation run.  Re-arange main loop so that simulator
 	is correctly restarted when an event halts the simulation.

	* psim.c (psim_halt): Handle an event halting the simulation.
	* psim.c (psim_init): Adjust initial cpu - == -1 - to match
 	reworked idecode.

Wed Jan 29 12:14:19 1997  Andrew Cagney  <[e-mail]>

	* ppc-opcode-complex: Correct typo - was expanding ORA instead of
 	RA.  Based on instruction frequency stats, expand additional
 	instructions.
	* ppc-instructions: Change all `RA == 0' to RA_is_0.
	
	* ppc-opcode-stupid: Move all but the basic table in -complex into
 	here.  Update to new format.
	
	* Makefile.in (tmp-defines): New target. Force defines.h to always
 	be built.  Hence get ppc-opcode-goto to build.

Tue Jan 28 13:00:19 1997  Andrew Cagney  <[e-mail]>

	* hw_com.c (hw_com_instance_read, hw_com_instance_write):
 	Implement.
	
Thu Jan 23 09:07:26 1997  Andrew Cagney  <[e-mail]>

	* hw_trace.c (hw_trace_init_data): Delete.  The trace options need
 	to be initialized independant of the rest of the simulation
 	initalization.  Otherwize a trace option explictly set from gdb
 	could be overridden by hw_trace.

	* psim.c (psim_options): Clarify reason why the trace ioctl occures.

	* FIXME: The trace code is too scattered - hw_trace.c, psim.c,
 	debug.c.  It could be much simpler.

Thu Jan 23 09:07:26 1997  Andrew Cagney  <[e-mail]>
	
	* Some devices support removable media.  Add hooks to the disk
 	device so that it supports this.
	
	* device.c (device_add_string_array_property,
 	device_find_string_array_property): New functions, manipulate
 	properties containing an array of strings.
	(device_find_string_property): Allow a string array.
	(device_init_static_properties): Update.
	(device_init_runtime_properties): Update.

	* hw_disk.c (hw_disk_ioctl): Add ioctl for changing the disk
 	media.  If no file image is specified, use the next one in the
 	image property list.
	(hw_disk_init_address): Change the file property so that it is a
 	string array - use the first entry for the initial file image.

	* tree.c (print_string_aray_property): New function - print a
 	string array.
	(print_properties): Adjust.
	(print_string): Write a string, handling double quotes.

	* device.h: Define an ioctl to `change-media' with an optional new
 	media image.
	
	* hw_disk.c: Allow floppy disk devices to be specified.

	* psim.c (psim_command): New function, parse more complex psim
 	commands such as "change-media" and "trace".
	* sim_calls.c (sim_do_command): Use.

Wed Jan 22 09:38:33 1997  Andrew Cagney  <[e-mail]>

	* For expressions like (RA == 0) that are entered in to the cache
 	as RA_is_0. If possible generate the result of the expression so
 	that the compiler gets a better chance of eliminating dead
 	branches.

	* gen-icache.c (print_icache_extraction): For a cache entry of
 	the form <name>_is_<const> where it is a boolean field, generate
 	the result of the expression instead of the expression its self.
	(print_icache_body): Remove code that was looking for *_is_0 and
 	then generating corresponding definitions.

	* gen-icache.c (print_icache_struct): If there is no cache, do not
 	output expressions in idecode.h file.
	
	* gen-icache.c (print_icache_body): Output them here.
	
	* ppc-opcode-complex: Clarify constant values for SPR==LR register
 	expansion.

	* ppc-cache-rules (RA_is_0, SPR_is_256): Two new cache entries.

Wed Jan 22 12:24:52 1997  Andrew Cagney  <[e-mail]>

	* The code that put values in and extracted values from the cache
 	was too compilicated.  The cache table did not allow values to be
 	computed from cache entries.  #defines could only be used when a
 	cache was present, remove the restriction.

	* ld-cache.h, ld-cache.c: Add a new cache entry type - SCRATCH.  A
 	scratch variable is defined when a cache entry is beinf
 	filled. Change the definition of a COMPUTE variable to be defined
 	when the cache entry is being used.
	* gen-icache.c: Update.
	
	* ld-cache.h, ld-cache.c: Change field names so that their meaning
 	is more obvious.  old_name->field_name, new_name->derived_name.
	* gen-icache.c: Update
	
	* gen-icache.h, gen-icache.c (print_icache_body): Make the three
 	different types of cache code - put into cache, extract from
 	cache, no cache - an explicit argument to print_icache_body.
	* gen-icache.c (print_icache_extraction): Ditto.
	
	* gen-semantics.c (print_c_semantic): Update use.
	* gen-idecode.c (print_jump_insn): Update use.
	* gen-icache.c (print_icache_function): Update use.

	* igen.c (main): Change 'R' option so that it does not force the
 	cache.

	* configure.in (enable-sim-icache): Clarify description.  Make
 	#define one of the defaults regardless of the cache.  Probably
 	should revamp and add a separate option.

Tue Jan 21 13:26:10 1997  Andrew Cagney  <[e-mail]>

	* pk_disklabel.c (block_is_fdisk): Tidy up traces - use dos
 	partition numbering.
	(pk_disklabel_create_instance): Partition 1..4 are valid - not
 	1..3.
	(is_iso9660): New function, verify a CD9660 File system.
	(pk_disklabel_create_instance): Start expanding so that active
 	partition selection is supported.

Mon Jan 20 11:20:15 1997  Andrew Cagney  <[e-mail]>

	* The cap object was retaining a reference to the instance of a
 	device after it was deleted. Instead add and remove cap's from the
 	cap db as they are created and deleted.  This ensuring that a
 	capibility is only used during the lifetime of the corresponding
 	object.
	
	* cap.h, cap.c: Correct cap type - was signed32 should be
 	signed_cell.
	
	* cap.c (cap_add, cap_remove): New methods for cap object that
 	allow the explicit addition and removal of internal objects that
 	the cap knows about.
	
	* cap.c (cap_init): Rewrite.  Verify that the only objects
 	remaining in the cap data base are those that were entered first.
  	Thse objects will be the permenant ones.
	* device.c (device_init_address): Remember to initialize the cap
 	database.