File:  [MW Coherent from dump] / coherent / b / lib / libc / Changes.1992
Revision 1.1.1.1 (vendor branch): download - view: text, annotated - select for diffs
Wed May 29 04:56:35 2019 UTC (7 years ago) by root
Branches: MarkWilliams, MAIN
CVS tags: relic, HEAD
coherent

libc/Changes
i8086/i386 libc source change log, begun 5/6/92

Wed May  6 14:31:23 1992 PDT
crt/Makefile
crt/i386/Makefile
crt/i386/_prof.c
crt/i8086/Makefile
crt/i8086/_prof.c
		i386 ld now writes e.g. "__end_text" rather than "etext",
		so split _prof.c into i386 and i8086 versions appropriately.
crt/i386/fdcvt.s
		Corrected bug in double -> float conversion.
gen/Makefile
gen/_addargs.c
gen/coffnlist.c
		Added _addargs() and coffnlist().
gen/i386/Makefile
gen/i386/alloca.s
		Added alloca().
gen/qsort.c
gen/shellsort.c
		Replaced qsort() and shellsort() with new sources from cef,
		with better performance.
sys/i386/brk.s
		Changed end to __end for consistency with ld.
sys/i386/mkscall.c
		Eliminated extraneous "orl %eax, %eax" in system call
		generation routine.
sys/i386/scall.s5
		Added system call poll (87).
sys/i386/msgemul.c
		Corrected by Bob H. by adding missing arguments.
sys/i386/shmemul.c
		Corrected by Bob H.
sys/i386/Makefile
sys/i386/uname.c
sys/i386/ustat.c
		Added stubs as requested by hal.

Thu May  7 10:19:55 1992 PDT
gen/Makefile
gen/_cleanup.c
		Added _cleanup() which currently does nothing,
		as requested by norm.  This is required for linking
		some ATT programs.

Thu May  7 16:03:44 1992 PDT
crt/i386/modf.c
		Corrected result for negative arguments.  For the old
		version, -1.15 returned -2.00 + .85 rather than -1 + -.15.
		The change is consistent with ANSI 4.5.4.6.

Tue May 12 15:45:46 1992 PDT
Makefile
ndir/Makefile
ndir/*.c
sys/i386/Makefile
		Added SV-style directory access routines in directory ndir.
		Modified Makefile accordingly.  Deleted getdents.o from
		system call object list in i386/Makefile, for now.

Thu May 21 14:28:58 1992 PDT
crt/i386/daddsub.s
		Promula reported a bug in floating point output,
		which was really a bug in floating point subtract.
		When the operands differed in precisely the 20th
		significant bit, the result was normalized incorrectly.

Fri May 22 13:39:32 1992 PDT
stdio/_dtefg.c
		Corrected modf() call; the 5/7/92 change above changed
		the results of modf() for negative arguments,
		this compensates for that change.

Tue May 26 13:25:29 1992 PDT
gen/ctime.c
		Added ftime(), previously missing, after modifying source
		provided by michael.

Thu May 28 16:28:29 1992 PDT
csu/i386/crts0.s
		Added runtime startoff source.

Fri May 29 13:58:04 1992 PDT
crt/i386/dmul.s
		Corrected after bug report from Promula: when rounding
		result carried past hidden bit, I decrmented the exponent
		when I should have incremented it.

Thu Jun  4 12:55:38 1992 PDT
gen/sys_err.c
		Added _I386 conditional to correspond to COH386 <errno.h>.

Tue Jun  9 16:43:42 1992 PDT
gen/Makefile
gen/lockf.c
		Added lockf.c and added lockf.o to Makefile.

Fri Jun 19 14:48:55 1992 PDT
gen/getwd.c
		Renamed getwd() to _getwd(), to avoid conflict with BSD
		getwd() which expects different arguments.
gen/Makefile
gen/getcwd.c
		Added getcwd(), modified Makefile accordingly.

Wed Jul  8 10:06:38 1992 PDT
gen/monitor.c
gen/ttyname.c
gen/ttyslot.c
gen/i386/Makefile
gen/i386/abort.c
gen/i386/signame.c
sys/i386/cerror.s
sys/i386/dup2.c
sys/i386/execve.s
sys/i386/fork.s
sys/i386/gtty.c
sys/i386/pipe.s
sys/i386/sememul.c
sys/i386/setpgrp.c
sys/i386/signal.s
sys/i386/utsemul.c
sys/i386/wait.s
		Integrated Ciaran's i386 sources into library sources.
sys/Makefile
sys/i8086/Makefile
sys/i8086/dup2.c
sys/i8086/semop.c
sys/i8086/time.c
		Moved sys/dup2.c to sys/i8086 to avoid conflict with
		the different source sys/i386/dup2.c.
		Moved sys/semop.c to sys/i8086 to avoid conflict with
		the different source sys/i386/sememul.c.
		Moved sys/time.c to sys/i8086 to avoid conflict with
		the different source sys/i386/time.c.
		Modified Makefiles accordingly.

Thu Jul  9 12:49:47 1992 PDT
stdio/fread.c
		Adjust seek position after partial read when item size > 1,
		as requested by michael.

Thu Jul 16 10:09:02 1992 PDT
crt/Makefile
crt/i386/Makefile
crt/i8086/Makefile
crt/_prof.c
		Added merged source, replacing crt/[i386|i8086]/_prof.c.
		Modified Makefiles accordingly.
crt/i386/_prof.c
crt/i8086/_prof.c
		Deleted, replaced by merged source crt/_prof.c.
crt/i386/scount.s
		Updated to work right for i386.
gen/monitor.c
		Comments added, cast added for i386.
gen/copyright.c
		Version number bumped to 4.0.

Fri Jul 17 11:22:31 1992 PDT
sys/Makefile
sys/execlep.c
sys/execl.c
sys/execle.c
sys/execlp.c
sys/execv.c
sys/execvep.c
sys/execvp.c
		Added execvep(), as requested by piggy and hal,
		a simple modification of previous execvp.c source.
		Changed execvp.c to call execvep().  Added execlep().
		Cleaned up other exec* sources slightly.

Tue Jul 21 11:58:31 1992 PDT
gen/copyright.c
		Version number bumped to 4.0.1.

Mon Jul 27 17:34:21 1992 PDT
stdio/finit.c
		Fixed obscure bug relating to errno.
		isatty() sets errno to ENOTTY (correctly) when false.
		But finit() calls isatty() to decide whether i/o should
		be buffered.  If finit() is initializing a FILE pointer
		which is a disk file, it therefore inadvertently sets errno
		to ENOTTY.  finit() now leaves errno unchanged.

Tue Jul 28 09:48:36 1992 PDT
sys/Makefile
sys/execlp.c
sys/execlpe.c
sys/execvp.c
sys/execvpe.c
		At hal's request, renamed execlep() and execvep() to
		execlpe() and execvpe(), for consistency with Borland.

Tue Aug  4 16:32:30 1992 PDT
sys/i386/Makefile
sys/i386/uname.c
sys/i386/ustat.c
		Corrected uname.c and ustat.c (they were reversed)
		and deleted the bogus utsname.c.

Tue Aug  4 19:12:41 1992 PDT
string/i386/memset.s
		Changed ambiguous "movzx" to unambiguous "movzxb" to avoid
		assembler warning message.

Thu Aug  6 10:48:02 1992 PDT
ndir/opendir.c
		At piggy's request, modified opendir() so that it stat()s
		the arg to make sure it is a directory before it opens it.
		The old code fstat()s the arg after the open, which can
		cause problems if the arg is a device.

Wed Sep  2 20:03:08 1992 PDT
		V4.0.2: recompiled from scratch after changing <stdio.h>
		_NFILES from 20 to 60, to correspond with system change by hal.

Tue Sep 22 13:29:47 1992 PDT
ctime.c
		V4.0.3: in localtime(), if DST is in effect, set isdst to 1
		unconditionally, rather than to (dstadjust != 0) as before.
		A user was intentionally setting dstadjust to 0 but still
		wanted to know that DST was in effect.

Mon Sep 28 14:54:15 1992 PDT
sys/Makefile
sys/execve.c
sys/i386/Makefile
sys/i386/_execve.s
sys/i8086/Makefile
		V4.0.4: Changed system call execve() to _execve(),
		then added new execve() which performs Unix-style
		interpretation of interpreter and arg specs on first line
		of scripts.  For example,
			#!/bin/whatever arga argb
		would run the script with /bin/whatever rather than /bin/sh,
		prepending "arga argb" to the command line arguments.

Tue Oct 27 10:01:30 1992 PST
crt/i387/rts87.s
		In _udcvt(), added code to check for overflow after conversion
		of double to 64-bit integer; return UINTMAX on overflow.
		Changed "cwchop" from 0x0FBF to 0x0F3F, resetting an unused bit.

Thu Oct 29 14:43:45 1992 PST
gen/ctime.c
		V4.1.1: modified "char tzname[2][32];" to "char *tzname[2];",
		for iBCS2 compatability, as per hal and bob.

Tue Nov  3 11:15:12 1992 PST
gen/ctime.c
		V4.1.2: modified ctime.c/isdaylight() after user
		Richard Lindner (in Australia) correctly noted that
		DST handling was broken for the Southern Hemisphere.

Fri Nov  6 14:50:41 1992 PST
stdio/Makefile
stdio/remove.c
stdio/rename.c
		V4.1.3: added ANSI-compliant functions remove() and rename().

Thu Nov 12 13:33:44 1992 PST
stdio/Makefile
stdio/fprintf.c
stdio/printf.c
stdio/sprintf.c
stdio/vprintf.c
stdio/vfprintf.c
stdio/vsprintf.c
		V4.1.4: Modified old printf.c (which defined printf(), sprintf()
		and fprintf() using static function xprintf()) into vfprintf.c.
		Added other new *printf.c using vfprintf() and modified Makefile
		accordingly.  This is still the non-ANSI printf(), though.

Mon Nov 16 09:52:33 1992 PST
gen/Makefile
gen/__assertfail.c
		V4.1.5: added __assertfail.c, defining __assertfail()
		referenced by new <assert.h>.

Mon Nov 16 15:18:02 1992 PST
gen/Makefile
gen/ftok.c
		V4.1.6: added ftok.c, defining ftok() as in SVR4,
		as requested by vlad.

Thu Nov 19 09:52:45 1992 PST
stdlib/
stdlib/Makefile
stdlib/atexit.c
		V4.1.7: added atexit.c, defining atexit() as in ANSI,
		as requested by cef.
		Added new stdlib directory and Makefile.
stdio/exit.c
		Modified to execute atexit()-specified functions.
		
Thu Dec  3 12:45:30 1992 PST
crt/i387/rts87.s
		V4.1.8: Added entry points _dpcvt() and _pdcvt() for conversions
		generated by particularly bogus source code.

Fri Dec  4 13:42:33 1992 PST
sys/i386/scall.s5
		Changed stime() entry, the flag bits previously specified
		that the arg was a pointer which should be dereferenced.
		The arg is indeed a time_t pointer, but the system expects
		the arg to be a pointer, not the value it points to.

Wed Dec  9 15:01:56 1992 PST
sys/i386/scall.s5
		V4.1.9: withdrew above stime() change after consultation
		with hal; the user interface is indeed
			int stime(time_t *timep)
		but the SV-compatible system call routine is
			int ustime(time_t time)
		so the interface routine does dereference the supplied pointer.

Wed Dec 16 11:40:07 1992 PST
gen/i386/setjmp.s
		V4.1.10: ANSI 4.6.2.1 mandates that longjmp(buf, 0) return 1,
		not 0; now it does.

Wed Dec 16 13:19:33 1992 PST
gen/ctime.c
		V4.1.11: Restored initialization to dsttimes[] deleted 11/16/92,
		after correcting the member order.

Fri Dec 18 14:21:46 1992 PST
stdio/fgets.c
stdio/gets.c
		V4.1.12: ANSI 4.9.7.2 says fgets() which encounters EOF
		without reading any characters should leave the contents
		of its array arg unchanged, the previous code stored a NUL
		character into it; similarly for gets(), ANSI 4.9.7.7.
		Now they work as required.  This fixes a bug reported
		by user Mark Sommer.

Wed Dec 23 14:28:13 1992 PST
sys/i386/shmemul.c
		V4.1.13: modified shared memory routines from vlad.

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.