changelog

ecos为实时嵌入式操作系统

1999-04-28  Jesper Skov  <[e-mail]>
        PR 19850
 	* tests/stress_threads.c: Don't print text from alarm handler.

1999-04-27  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: 
	Use/test 'cyg_thread_delete()' function.
	Add ability to compute statistics with first datum removed which
	can show up cache effects.
	Show thread stack usage.

1999-04-26  Hugo Tyson  <[e-mail]>

	* include/kapi.h: Add missing function cyg_thread_delete();
	otherwise there is no way in the C API to re-use thread memory
	safely.  It just calls the destructor of the Cyg_Thread class.

	* src/common/kapi.cxx (cyg_thread_delete): Implement it.

1999-04-26  Hugo Tyson  <[e-mail]>

	* tests/stress_threads.c: Commit a better version for Mark since I
	happen to have it here; having addressed the main concern of my
	approval process.  It runs better in sims so long as the
	constructor execution isn't messed with, nor timeslicing nor
	priority inheritance.

1999-04-23  Mark Galassi  <[e-mail]>

	* tests/stress_threads.c (setup_death_alarm): shortened the
	simulator time by another factor of 10, hoping that it will not
	slow down nightly testing for simulator platforms too much.
	(perform_stressful_tasks): added writing of a bit pattern to the
	malloc()-ed spaces.  Also reduced stack requirements for threads.

1999-04-23  Hugo Tyson  <[e-mail]>

	* src/common/timer.cxx (Cyg_Timer::activate): we must also disable
	the alarm when resetting it so as to remove it from its queue, so
	that the enable afterwards places it on the right queue instead of
	assuming that, being enabled, it's already there.  (if not
	enabling, the behaviour is unchanged and correct) This was
	detected by uITRON tests[cx]4, with some random perms including
	enable CYGIMP_KERNEL_COUNTERS_MULTI_LIST,
	value CYGNUM_KERNEL_COUNTERS_MULTI_LIST_SIZE {8}.
	The PR is 19475.

1999-04-21  Hugo Tyson  <[e-mail]>

	* tests/bin_sem1.cxx (bin_sem1_main): Doh!  Use priorities 4 and 5
	for the two threads so that they are unique (with a bitmap
	scheduler, the threads had prios 0 and 1 resp, which causes an
	assert in the attempt to set thread 0's prio to 1, "prios not
	unique").  PR 19904.

1999-04-20  Hugo Tyson  <[e-mail]>

	* tests/stress_threads.c: applied my patch to fix stack sizes.
	And timing, in simulator; branch now matches the trunk.

1999-04-20  Jonathan Larmour  <[e-mail]>

	* tests/kcache2.c: 
	Reduce iterations for time_ilock(), time_dlock(), and test_dzero()
	if a simulated environment is detected
	Fix for PR 19853

1999-04-19  Jonathan Larmour  <[e-mail]>

	* tests/kill.cxx: Make delay ticks greater than 1 to prevent
	scheduling problems if the clock rolls over immediately
	Fix for PR 19516

1999-04-19  Hugo Tyson  <[e-mail]>

	* tests/bin_sem1.cxx (bin_sem1_main): Add priorities to the
	threads so that they run sequentially like the tester expected.
	Otherwise simple timing, and shorter timeslicing, causes failure
	to be reported; likewise if the kernel happens not to run entry0
	first!
        (another checkin)
	(bin_sem1_main): Add conditional use of set_priority on
	CYGIMP_THREAD_PRIORITY as it should be.

1999-04-14  Jesper Skov  <[e-mail]>

	* src/debug/dbg_gdb.cxx: Fixed compiler warnings, cleaned up stub
	requirements (removed FIX ME).

1999-04-14  Gary Thomas  <[e-mail]>

	* include/intr.hxx: 
	* src/intr/intr.cxx (call_pending_DSRs_inner): Rework calling of
	DSRs to allow calls from HAL code.  This will allow for proper
	use of a separate interrupt stack.

1999-04-14  Mark Galassi  <[e-mail]>

	* tests/PKGconf.mak: added an ifndef for the AEB-1 board so that
	stress_threads is only built conditionally.

	* tests/stress_threads.c: added more config smarts suggested by
	Jesper for his configurations.

1999-04-13  Jesper Skov  <[e-mail]>
        PR 19822
	* src/debug/dbg_gdb.cxx: Current thread's registers live in
	registers, not _registers.

1999-04-13  Mark Galassi  <[e-mail]>

	* tests/stress_threads.c: simple stressing of thread
	creation/deletion with some memory-intensive and alarm-based
	tasks.  This version is set so that it dies after DEATH_TIME_LIMIT
	seconds, currently set to 120.  #undef-ing DEATH_TIME_LIMIT makes
	the test run forever.  This is currently not working with the
	bitmap scheduler, probably for some simple reason, but I have not
	yet broken it down to see why.

1999-04-12  Jonathan Larmour  <[e-mail]>

	* tests/except1.cxx: 
	* tests/kexcept1.cxx: 
	Remove now unnecessary tx39-specific setting of
	CYGNUM_HAL_EXCEPTION_DATA_UNALIGNED_ACCESS VSR
	Part of cleanup for PR 19731

1999-04-09  Hugo Tyson  <[e-mail]>

	Generally: thread->to_queue_head() moves a thread to the head of
	whatever queue it is on, if relevant.  This is so that a thread
	can change priority and then put itself at the front of its new
	run queue instead of the end; otherwise it yields too, which may
	be undesirable.  Particularly for internal use of priorities in
	the uITRON compatibility layer.

	* include/thread.hxx (class Cyg_Thread): to_queue_head(): new
	function. Define Cyg_Thread::rotate_queue() for all schedulers.
	Public inheritance of Cyg_ThreadQueue_Implementation.

	* include/thread.inl (class Cyg_Thread): to_queue_head(): new
	function. Define Cyg_Thread::rotate_queue() for all schedulers.

	* include/mlqueue.hxx (class Cyg_SchedThread_Implementation):
	to_queue_head(), ...ThreadQueue...::to_head(): new functions.

	* src/sched/mlqueue.cxx (class Cyg_SchedThread_Implementation):
	to_queue_head(), ...ThreadQueue...::to_head(): new functions.

	* include/bitmap.hxx (class Cyg_SchedThread_Implementation): 
	Add empty placeholders for rotate_queue() and to_queue_head() to
	maintain commonality of interface.

1999-04-08  Jesper Skov  <[e-mail]>
        PR 19667
	* kernel/current/src/common/thread.cxx:
	Removed obsolete Cyg_Thread constructor.

	* kernel/current/tests/testaux.hxx:
	* kernel/current/tests/thread0.cxx:
	* kernel/current/tests/thread1.cxx:
	Changed to use the new Cyg_Thread constructor.

1999-04-08  Jesper Skov  <[e-mail]>

	* src/common/kapi.cxx: Minor indentation fixes.

1999-04-07  Jesper Skov  <[e-mail]>
        PR 19743
	* include/mfiximpl.inl (Cyg_Mempool_Fixed_Implementation): Moved
	argument assertions before the first use of the arguments.

1999-04-07  Hugo Tyson  <[e-mail]>

	* tests/kexcept1.c (except0_main): 
	* tests/except1.cxx (except0_main): 
	Use macro HAL_VSR_SET_TO_ECOS_HANDLER() if defined, to reset
	likely exception handler VSRs to eCos' default handlers - the
	target side equivalent to CYG_TEST_GDBCMD("handle SIGBUS nostop");
	The VSR may not be the eCos standard one in some implementations
	of cooperation with CygMon.

1999-03-31  Nick Garnett  <[e-mail]>

	* include/pkgconf/kernel.h (CYGNUM_KERNEL_COUNTERS_RTC_PERIOD):
	Fixed value of this option for TX39 66MHz to correct number.

1999-03-25  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: More adjustments to get clock latency correct.
	Some fine tuning of messages and layout to fit well within 80 columns.

1999-03-24  Nick Garnett  <[e-mail]>

	* include/pkgconf/kernel.h: Fixed an ifdef error.

1999-03-24  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx (NTEST_THREADS): Reduce # threads so test can
	build on smaller machines.  Also, only adjust timings by single
	"overhead" value instead of the more generous 2*overhead.

	* src/common/clock.cxx: Clock latency measurement defaults to 'off'.

1999-03-23  Nick Garnett  <[e-mail]>

	* tests/kcache2.c: Reduced number of loops in time_ilock(). On the
	MN10300 and TX39 this caused the test to timeout in the testing
	farm.

1999-03-23  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: 
	* src/common/clock.cxx: Add 'measure_clock_latency' boolean which
	is used to turn clock latency measurements on/off.

1999-03-23  Nick Garnett  <[e-mail]>

	* tests/tm_basic.cxx (run_all_tests): Changed initial delay from
	100 ticks to 2. This is adequate for the purpose of letting the
	clock get going, while 100 ticks was too long on simulators.

1999-03-23  Hugo Tyson  <[e-mail]>

	* tests/kcache1.c:
	* tests/kcache2.c:
	* tests/kclock1.c:
	* tests/kexcept1.c:
	* tests/kflag1.c:
	* tests/kmbox1.c:
	* tests/kmemfix1.c:
	* tests/kmemvar1.c:
	* tests/kmutex1.c:
	* tests/ksched1.c:
	* tests/ksem1.c:
	* tests/kthread0.c:
	* tests/thread0.cxx:
	Use CYGNUM_HAL_STACK_SIZE_TYPICAL instead of "4096" and include
	hal_arch.h where necessary to get it.

1999-03-22  Hugo Tyson  <[e-mail]>

	* include/thread.inl (attach_stack): 
	Fix typo in assert; missing comma.

1999-03-22  Jonathan Larmour  <[e-mail]>

	* src/debug/dbg-thread-syscall.h: Update copyright

1999-03-22  Hugo Tyson  <[e-mail]>

	* include/thread.inl:
	* src/common/thread.cxx:
	* tests/kthread1.c:
	* tests/testaux.hxx:
	* tests/thread1.cxx:
	* tests/thread_gdb.c:
	* tests/tm_basic.cxx:
	Use CYGNUM_HAL_STACK_SIZE_TYPICAL for the stack size instead of
	CYGNUM_HAL_MINIMUM_STACK_SIZE.
	Except for stack checking and the idle thread stack which uses 
	CYGNUM_HAL_STACK_SIZE_MINIMUM.

1999-03-19  Nick Garnett  <[e-mail]>

	* tests/kcache1.c:
	* tests/kcache2.c:
	Made all CYG_INTERRUPT_STATE variables register variables. The
	value of this variable has to survive cache invalidation and does
	not in some testing permutations that are unoptimized.
	Fixes PR 19165.

	* include/kapi.h: Type of cyg_tick_count_t changed to cyg_uint64.
	Response to PR 19320.

1999-03-19  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: Much better analysis of clock interrupt times.
	Print average deviation instead of variance.  Also print "confidence"
	numbers which are a measure of "how likely" a particular number would
	be relative to the mean and min of the sample set.

1999-03-17  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: Improve messages and layout.
	Better message for 'hardware clock tick' value.

	* src/common/clock.cxx: Be more careful about clock latency delta
	values that are used/kept - some hardware can yield incorrect values.

	* tests/testaux.hxx: Use HAL recommended stack size for default.

1999-03-16  Nick Garnett  <[e-mail]>

	* src/common/thread.cxx:
	Removed MIPS specific definition of idle_thread_stack[]. This is
	so it gets the benefit of the redefinition of
	CYGNUM_KERNEL_THREADS_IDLE_STACK_SIZE. This definition has also
	been made static, since the MIPS compiler would otherwise put it
	in the .data section.

	* include/intr.hxx (class Cyg_Interrupt):
	Changed implementation of Cyg_Interrupt::interrupts_enabled() to
	test the value of disable_counter, rather than the hardware.

1999-03-15  Nick Garnett  <[e-mail]>

	* tests/kcache1.c:
	* tests/kcache2.c: 
	Added interrupt disables around cache operations. If these take
	too long, an inopportune interrupt can disrupt the cache contents,
	or result in the interrupt seeing an inconsistent world.

1999-03-15  Gary Thomas  <[e-mail]>

	* tests/tm_basic.cxx: Improved and cleaned up messages overall.

1999-03-15  Jesper Skov  <[e-mail]>
        PR 19504
	* tests/tm_basic.cxx: Kill created threads after mutex, mbox and
	semaphore tests.

1999-03-12  Hugo Tyson  <[e-mail]>

	* src/debug/dbg-thread-demux.c: 
	Only try handling the SPARClite-SLEB vector correctly if
	CYG_HAL_USE_ROM_MONITOR_CYGMON ie. we are talking with CygMon.
	Otherwise the code is wrong, but also not used.  No matter.
	
1999-03-12  Hugo Tyson  <[e-mail]>

	* tests/kthread1.c: 
	* tests/thread1.cxx: 
	* tests/thread_gdb.c: 
	Deal with CYGNUM_HAL_MINIMUM_STACK_SIZE requirement.

1999-03-12  Gary Thomas  <[e-mail]>

	* include/thread.inl (attach_stack): Add check/assert for minimum
	stack size (HAL defined)

1999-03-11  Hugo Tyson  <[e-mail]>

	* src/debug/dbg-thread-demux.c (patch_dbg_syscalls): 
	Use the correct vector table slot for SPARClite-SLEB; it's not 15
	on here, it's BSP_VEC_MT_DEBUG from the hal's hal_cygm.h

1999-03-11  Nick Garnett  <[e-mail]>

	* tests/testaux.hxx:
	Move definition of inline placement operator new outside of ifdef
	for NTHREADS so that it can be used by other code.

	* tests/intr0.cxx:
	* tests/thread1.cxx:
	Make use of placement operator new rather than dynamic one in
	intr0. Removed duplicate definition from thread1.

1999-03-11  Jesper Skov  <[e-mail]>
        PR 19472
	* tests/kill.cxx: Increased delays for the synthetic target.

1999-03-11  Jesper Skov  <[e-mail]>
        PR 19468
	* tests/intr0.cxx (intr0_main): 
	* tests/kintr0.c (kintr0_main): 
	Only attach interrupt handlers to vectors not in use.
	
1999-03-10  Jesper Skov  <[e-mail]>
        PR 19368
	* src/sched/sched.cxx (start): Change the dependency for
	referencing the RTC to make its inclusion independent of scheduler
	choice.

1999-03-10  Jesper Skov  <[e-mail]>
        PR 17357
	* src/intr/intr.cxx (attach): Assert that interrupt vector attach
	will succeed.

1999-03-10  Nick Garnett  <[e-mail]>

	* include/pkgconf/kernel.h:
	Changed names used to control clock frequency on TX39 target.

1999-03-08  Gary Thomas  <[e-mail]>

	* src/sync/mutex.cxx: Allow 'Cyg_Condition_Variable::wait()'
	to be entered with the scheduler locked (and thsu exit in
	the same state).  This will allow for fully thread-safe use
	of 'Cyg_Condition_Variable::wait()'.

1999-02-26  Jonathan Larmour  <[e-mail]>

	* tests/except1.cxx (except0_main): 
	* tests/kexcept1.c (except0_main): 
	Move CYG_TEST_GDBCMD back to being before CYG_TEST_INIT

1999-02-26  Jesper Skov  <[e-mail]>
        PR 19283