changelog

eCos1.31版

2000-03-20  John Dallaway  <[e-mail]>

	* cdl/string.cdl,
	  cdl/stdlib.cdl,
	  cdl/stdio.cdl,
	  cdl/libc.cdl:

	Adjust documentation URLs.

2000-03-06  Jesper Skov  <[e-mail]>

	* tests/time/strftime.c (test): Adjust test to match below change.

2000-03-02  Jonathan Larmour  <[e-mail]>

	* src/time/strftime.cxx (do_format): Make sure %m counts from
	1..12 rather than 0..11. Fix for case 103300

2000-02-29  Jonathan Larmour  <[e-mail]>

	* src/stdio/output/vfnprintf.cxx: Correctly use 'n' parameter to
	restrict length. Credit due to Andrew Lunn here as well.
	Resolves 103057-CR-1

	* tests/stdio/sprintf2.c (test): 
	Add snprintf tests to verify the correct behaviour of restricting
	output length

2000-02-10  Jesper Skov  <[e-mail]>

	* tests/signal/signal2.c (main): 
	Also reclaim DATA_TLB_MISS vector if available.

2000-01-31 Simon FitzMaurice  <[e-mail]>
  * cdl/*.cdl:

  Adjust help URLs in line with new doc layout.
  
2000-01-28 Simon FitzMaurice  <[e-mail]>
  * cdl/*.cdl:

  Adjust help URLs in line with new doc layout.
  
2000-01-19  Hugo Tyson  <[e-mail]>

	* cdl/*.cdl: Add descriptions to a number of options &c which were
	lacking same, also tidied up other typos as noticed en passant.

1999-11-11  Bart Veer  <[e-mail]>

	* cdl/string.cdl:
	Switch to a single boolean configuration option for code size vs.
	speed optimization.

1999-11-03  John Dallaway  <[e-mail]>

	* cdl/libc.cdl: Define tests.

1999-10-19  Jonathan Larmour  <[e-mail]>

	* src/stdio/output/vfnprintf.cxx: We can now remove Berkeley
	advertising clause
	* src/stdlib/div.cxx: ditto
	* src/stdlib/ldiv.cxx: ditto
	* src/stdlib/strtol.cxx: ditto
	* src/stdlib/strtoul.cxx: ditto
	* src/string/strtok.cxx: ditto

1999-10-19  John Dallaway  <[e-mail]>

	* cdl/libc.cdl:

	Define CDL options for package-specific CFLAGS.

1999-10-08  Jonathan Larmour  <[e-mail]>

	* include/pkgconf/libc.h (CYGNUM_LIBC_MAIN_THREAD_PRIORITY): Change
	default to 10 to keep test farm happy
	* cdl/startup.cdl: Likewise

1999-10-07  John Dallaway  <[e-mail]>

	* cdl/stdlib.cdl, cdl/string.cdl:

	Specify radio buttons using CDL interfaces.

1999-10-06  Jonathan Larmour  <[e-mail]>

	* cdl/stdio.cdl: Change default to /dev/ttydiag as per 1999-10-01
	change below
	* cdl/startup.cdl: Add new form CDL for the startup changes as
	per 1999-10-05

1999-10-05  Jonathan Larmour  <[e-mail]>

	* include/libc.h: Add CYGSEM_LIBC_MAIN_STACK_FROM_SYSTEM to allow
	the user to supply their own stack.
	Rename CYGNUM_LIBC_MAIN_STACK_SIZE to
	CYGNUM_LIBC_MAIN_DEFAULT_STACK_SIZE and reparent under
	CYGSEM_LIBC_MAIN_STACK_FROM_SYSTEM
	Add CYGNUM_LIBC_MAIN_THREAD_PRIORITY to provide main() thread
	priority
	
	* src/support/mainthread.cxx: Add support for user-supplied stack
	Allow user to specify thread priority

	These changes fix CR 101069

1999-10-01  Jonathan Larmour  <[e-mail]>

	* include/pkgconf/libc.h (CYGDAT_LIBC_STDIO_DEFAULT_CONSOLE): Change
	default to /dev/ttydiag to get correct console-like behaviour
	Also don't preclude use if haldiag package isn't included
	(case 102065)

1999-09-15  Jesper Skov  <[e-mail]>

	* tests/signal/signal2.c (cause_fpe): Try both float and integer
	division by zero.

1999-09-10  Jesper Skov  <[e-mail]>

	* src/stdlib/malloc.cxx (cyg_libc_get_malloc_pool): Added for
	stress_threads test.

1999-09-07  Jesper Skov  <[e-mail]>
        PRs 20243, 20433
	* tests/signal/signal2.c: Disabled test3 on ARM.

1999-09-02  Hugo Tyson  <[e-mail]>

	* include/assert.h: Make the sense of NDEBUG correct; it is "set"
	when NOT debugging.  Fix for CR 101983-CR-1
	See Stroustrup, "The C++ Programming Language."

1999-08-18  Jonathan Larmour  <[e-mail]>

	* src/stdio/common/fopen.cxx (_fopen): Set buffer size 0 when not
	buffering
	Fix for case 101868

1999-08-17  John Dallaway  <[e-mail]>

	* cdl/stdlib.cdl, cdl/string.cdl:

	Implement radio buttons using "FIXME radio" hack in
	description field for now.

1999-08-12  John Dallaway  <[e-mail]>

	* cdl/stdlib.cdl:
	Take advantage of MACRONAME_ENUMVAL macros output by CDL.

1999-08-06  Jonathan Larmour  <[e-mail]>

	* tests/stdio/sprintf2.c (test): Don't rely on 2.345 being exactly
	representable in binary as a difference of 1 ulp will make it round
	the wrong way.

1999-07-16  Jonathan Larmour  <[e-mail]>

	* src/stdio/input/fgetc.cxx: 
	Include outline versions of getc() (actually just a symbol alias for
	fgetc() ) and getchar()
	Some minor clean-ups
	Add tracing/assertions

	* src/stdio/input/fputc.cxx: 
	Likewise (except applied to putc() and putchar())

	* src/stdio/common/stdioinlines.cxx (_puts):
	Call fputc() not _fputc()

	* src/clibincl/stdiosupp.hxx:
	Remove now obsolete prototypes for _fgetc() and _fputc()

	* include/stdio.h:
	Don't bother with getc()/putc() macros any more, just call
	them as functions
	Include protos for getc(), putc(), getchar() and putchar() functions

1999-06-21  Jesper Skov  <[e-mail]>

	* tests/signal/signal2.c: Disabled test 3 for SH EDK7708.

1999-06-11  Jonathan Larmour  <[e-mail]>

	* include/stdio.h: Make FILEs array types so they won't be
	given $gp relocations on MIPS targets

1999-05-27  Jesper Skov  <[e-mail]>

	* tests/time/ctime.c: Fixed compiler warning.

1999-05-25  Jonathan Larmour  <[e-mail]>

	* tests/signal/signal2.c (main): 
	Update CYGPKG_HAL_TX39 -> CYGPKG_HAL_MIPS_TX39 due to recent MIPS
	HAL changes
	Fix for PR 20116

1999-04-30  Jesper Skov  <[e-mail]>
        PR 19945/19946 workaround
	* tests/signal/signal2.c: Disabled test 3 for PowerPC sim.

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

	* include/time.inl:
	Ensure all calls to cyg_libc_time_year_is_leap() are with the full
	year, and not the direct contents of a struct tm, which are (year-1900)

1999-04-19  Jesper Skov  <[e-mail]>
        PR 19308, 19410, 19528
	* tests/time/clock.c (test): Also disable data cache which may
	affect timing when tracing/assertions are enabled.

1999-04-15  Jonathan Larmour  <[e-mail]>

	* src/clibincl/stdiosupp.hxx:
	Add prototype for new cyg_libc_stdio_flush_all_but() function

	* src/clibincl/stream.hxx:
	Don't include stdiosupp.hxx any more to prevent circular dependency,
	therefore manually provide _setvbuf prototype
	Make sure pkgconf/kernel.h is included before mutex.hxx

	* src/clibincl/stream.inl (Cyg_StdioStream::set_error()):
	Add missing unlock_me()

	* src/stdio/common/fflush.cxx:
	Separate most of code for fflush(NULL) into separate
	cyg_libc_stdio_flush_all_but() function, which has the extra
	capability of omitting one stream

	* src/stdio/common/stderr.cxx:
	* src/stdio/common/stdin.cxx:
	* src/stdio/common/stdout.cxx:
	* src/stdio/common/vsnprintf.cxx:
	* src/stdio/common/vsscanf.cxx:
	Add missing clibincl/stdiosupp.hxx include

	* src/stdio/common/stream.cxx (Cyg_StdioStream::read()):
	When reading, ensure we flush all output streams

	* src/stdio/common/stream.cxx (Cyg_StdioStream::flush_output_unlocked):
	Flush the device as well

1999-04-15  Jonathan Larmour  <[e-mail]>

	* src/stdio/common/streambuf.cxx (set_buffer): 
	* src/stdio/common/stdioinlines.cxx (_setbuf): 
	* src/clibincl/streambuf.inl: 
	* src/clibincl/streambuf.hxx: 
	* src/clibincl/stream.hxx (class Cyg_StdioStream): 
	* include/pkgconf/libc.h:
	* include/stdio.inl (setbuf): 
	Fix dumb typo: BUFSIZE -> BUFSIZ
	This fixes PR19837

1999-04-15  Jonathan Larmour  <[e-mail]>

	* src/stdio/common/stdioinlines.cxx (_setbuf): 
	* include/stdio.inl (setbuf): 
	Use _IONBF when buf==NULL, and _IOLBF otherwise.
	Fix for PR 19836

1999-04-15  John Dallaway  <[e-mail]>

	* include/pkgconf/libc.h: Tidy display string capitalization.

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

	* include/pkgconf/libc.h: 
	Clarify CDL description of CYGNUM_LIBC_MALLOC_MEMPOOL_SIZE

	* tests/signal/signal2.c: Try harder to generate an exception by
	accessing a misaligned address
	Don't handle SIGTRAP any more
	Part of fix for PRs 19731/19607

1999-04-07  Jesper Skov  <[e-mail]>
        PR 19667
	* src/support/mainthread.cxx: 
	Changed to use the new Cyg_Thread constructor.

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

	* tests/signal/signal2.c (main): Use new macro
	HAL_VSR_SET_TO_ECOS_HANDLER if defined, to wrest control from
	CygMon, for example, by reinstalling eCos' default VSRs for
	exceptions that we are provoking.

1999-03-29  Jonathan Larmour  <[e-mail]>

	* src/signal/siginit.cxx: Include stdlib.h for exit()
	Fix for PR 19720

1999-03-29  Jonathan Larmour  <[e-mail]>

	* tests/time/clock.c (test): Disable instruction cache
	Hopefully fixes PR 19308

1999-03-29  Jesper Skov  <[e-mail]>

	* tests/signal/signal2.c: Disabled test 3 for PPC as it cannot
	trap division by zero.

1999-03-26  Jonathan Larmour  <[e-mail]>

	* src/support/main.cxx (main): Invoke cyg_user_start() if no kernel
	Only sensible thing to do in our default main in a non-kernel config

1999-03-25  Jonathan Larmour  <[e-mail]>

	* tests/time/time.c: 
	Increase MAX_TIMEOUT since the sparclite gets very close to overrunning
	it even when working.
	Should return -1 iff CYGSEM_LIBC_TIME_TIME_WORKING isn't defined
	Cosmetic changes	

1999-03-25  Hugo Tyson  <[e-mail]>

	* tests/time/time.c (test): Take into account setting of
	CYGSEM_LIBC_TIME_SETTIME_WORKING ie. does cyg_libc_time_settime()
	work?  If not, check that it returns an error.
	This was exposed by the random combo in PR#19479.

1999-03-23  Jonathan Larmour  <[e-mail]>

	* include/pkgconf/libc.h:
	Rename CYGSEM_LIBC_RAND_SIMPLEST and CYGSEM_LIBC_RAND_KNUTH1 to
	CYGIMP_...

	Add another rand implementation CYGIMP_LIBC_RAND_SIMPLE1 which is
	the default

	Part of fix for PR 19562

	* include/sys/cstartup.h:
	Tidy and remove unnecessary dependencies

	* src/stdlib/rand.cxx:
	Rename algorithms from CYGSEM_... to CYGIMP_...
	Add new "simple1" algorithm

	* tests/ctype/ctype.c:
	Tidy

	* tests/stdlib/rand3.c:
	Fix so that 0..RAND_MAX is split evenly over buckets, which it wasn't
	before
	Deal with case of RAND_MAX not being divisible by NUM_BUCKETS
	exactly
	Increase TEST_LENGTH to 200000 to provide more representative sample
	Add CYG_TEST_INFO output to help ensure it doesn't time out
	Tidy

	* tests/stdlib/rand4.c:
	Only fail if more than _two_ thirds (rather than one third) match.
	This is because half would match for a perfectly random distribution
	This is the rest of the fix for PR 19562
	Cosmetic changes

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

	* src/support/mainthread.cxx (CYGNUM_LIBC_MAIN_STACK_SIZE): 
	Use CYGNUM_HAL_STACK_SIZE_TYPICAL for the stack size instead of
	CYGNUM_HAL_MINIMUM_STACK_SIZE.

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

	* tests/PKGconf.mak: Add -Wno-format when building strftime otherwise
	it generates warnings for things that we do want to test (null
	string, and %y format)

1999-03-19  Jonathan Larmour  <[e-mail]>

	* include/pkgconf/libc.h (CYGNUM_LIBC_TIME_DST_DEFAULT_STATE): 
	Put brackets back round negative number since sdf says the config
	tool is fixed now (PR 19537)

	Add new CYGSEM_LIBC_RAND_SIMPLEST and CYGSEM_LIBC_RAND_KNUTH1
	options for different rand() algorithms in a so far failed attempt
	to fix PR 19562

	* src/stdlib/rand.cxx:
	Add new optional algorithm from Knuth vol.2

	* tests/stdlib/rand4.c:
	Some tidies
	Don't run test when using simplest algorithm

	* tests/signal/signal2.c (main): 
	Don't try to make an FPE on tx39 - its possible to detect but requires
	too much HAL work to be sensible

1999-03-18  Jonathan Larmour  <[e-mail]>

	* src/signal/siginit.cxx (cyg_libc_signals_hwexcept_handler): 
	Use HAL_ENABLE_INTERRUPTS() rather than kernel to enable interrupts

1999-03-18  Mark Galassi  <[e-mail]>