📄 36
字号:
Received: from snow-white.ee.udel.edu by whimsy.udel.edu id aa24162; 12 Nov 96 9:28 GMTReceived: from mumps.pfcs.com (mumps.pfcs.com [192.52.69.11]) by pcpsj.pfcs.com (8.6.12/8.6.9) with SMTP id EAA27052 for <stenn@whimsy.udel.edu>; Tue, 12 Nov 1996 04:27:54 -0500Received: from localhost by mumps.pfcs.com with SMTP id AA21031 (5.67b/IDA-1.5 for <stenn@whimsy.udel.edu>); Tue, 12 Nov 1996 04:27:53 -0500Prev-Resent: Tue, 12 Nov 1996 04:27:50 -0400Prev-Resent: "stenn@whimsy.udel.edu "Received: from brown.pfcs.com by mumps.pfcs.com with SMTP id AA14881 (5.67b/IDA-1.5 for <harlan@mailhost.pfcs.com>); Mon, 11 Nov 1996 18:30:57 -0500Received: from mumps.pfcs.com (mumps.pfcs.com [192.52.69.11]) by brown.pfcs.com (8.7.5/8.6.9) with SMTP id SAA01955 for <harlan@brown.pfcs.com>; Mon, 11 Nov 1996 18:30:55 -0500 (EST)Received: from pcpsj.pfcs.com by mumps.pfcs.com with SMTP id AA14877 (5.67b/IDA-1.5 for <harlan@brown.pfcs.com>); Mon, 11 Nov 1996 18:30:53 -0500Received: from postman.osf.org (uucp@localhost) by pcpsj.pfcs.com (8.6.12/8.6.9) with UUCP id SAA25700 for harlan@brown.pfcs.com; Mon, 11 Nov 1996 18:30:52 -0500Received: from postman.osf.org (postman.osf.org [130.105.1.152]) by uustar.starnet.net (8.8.2/8.8.2) with ESMTP id RAA08636 for <harlan@brown.pfcs.com>; Mon, 11 Nov 1996 17:05:15 -0600 (CST)Received: from firebird.osf.org (firebird.osf.org [130.105.3.231]) by postman.osf.org (8.7.6/8.7.3) with SMTP id SAA11057; Mon, 11 Nov 1996 18:04:44 -0500 (EST)From: Dan Murphy <dlm@osf.org>Received: by firebird.osf.org (5.65/4.7) id AA12103; Mon, 11 Nov 96 18:04:43 -0500Date: Mon, 11 Nov 96 18:04:43 -0500Message-Id: <9611112304.AA12103@firebird.osf.org>To: Mills@huey.udel.eduSubject: Re: NTP changes to uploadCc: harlan@mumps.pfcs.comResent-To: stenn@whimsy.udel.eduResent-Date: Tue, 12 Nov 1996 04:27:52 -0400Resent-Message-Id: <21029.847790872@mumps.pfcs.com>Resent-From: Harlan Stenn <Harlan.Stenn@pfcs.com> Greetings again. Per your request, I grabbed the current testing version of NTP and attempted to integrate our changes therein. The effort was successful in that I do indeed have a build which works and supports our system. Unfortunately, I could not generate what would seem to be the "right" changes in all cases to recognize and support our configuration, at least not in a few days of fooling around with it. We don't actually have all the tools installed here apparently, in particular, the ones to process the Makefile.am files. So, the patches that follow are divided into three parts. The first is a straightforward set of patches to original files. In addition, we must patch two files that are produced by the configure process. Perhaps you'll see quickly how to fix configure or configure.in to affect these same changes, but it was not obvious to me. Thus, the second part is to patch config.cache *after* configure has been run once. This fixes a few things that configure sets incorrectly. The third part is to fix config.h (after configure has been run a second time). This adds the conditional symbol for our system and also turns on REFCLOCK. I find it odd that REFCLOCK isn't on as would be consistent with including the whole bunch of reference clock drivers... So, cludgy as this may be, it does work. I tested all of the patch files below and they work correctly when applied at the root of the xntp3-5.86.8 untarred distribution. Thanks for incorporating these changes. I think this follows the requirements in the patching procedure html document. You're welcome to include my name, etc. on the copyright page if you're so inclined. dlm================================================================= #1 - This fixes config.guess to recognize our particular system. --- ../xntp3-5.86.8/./config.guess Thu Oct 10 04:28:00 1996+++ ./config.guess Thu Nov 7 16:30:59 1996@@ -299,6 +299,13 @@ hp8??:OSF1:*:*) echo hppa1.0-hp-osf exit 0 ;;+ i[345]86:OSF1:*:*)+ if [ -x /usr/sbin/sysversion ] ; then+ echo ${UNAME_MACHINE}-unknown-osf1mk+ else+ echo ${UNAME_MACHINE}-unknown-osf1+ fi+ exit 0 ;; parisc*:Lites*:*:*) echo hppa1.1-hp-lites exit 0 ;;================================================================ #2 - This causes sys/types.h to be included earlier in the sequence of include files and so avoid a problem where a typedef attempts to override a macro definition. --- ../xntp3-5.86.8/include/ntp_types.h Mon Jan 8 23:24:54 1996+++ include/ntp_types.h Fri Nov 8 13:02:37 1996@@ -8,6 +8,8 @@ #ifndef _NTP_TYPES_ #define _NTP_TYPES_ +#include <sys/types.h>+ /* * This is another naming conflict. * On NetBSD for MAC the macro "mac" is defined as 1================================================================ #3 - Two Makefiles need to be fixed to link in the library which supports nlist. Possibly this could/should be fixed in Makefile.am. --- ../xntp3-5.86.8/./xntpd/Makefile.in Tue Oct 29 21:21:30 1996+++ ./xntpd/Makefile.in Fri Nov 8 12:46:44 1996@@ -42,7 +42,7 @@ bin_PROGRAMS = xntpd INCLUDES = -I$(top_srcdir)/include -I$(srcdir) # LDADD might need RESLIB and ADJLIB-LDADD = version.o ../libntp/libntp.a @LIBPARSE@+LDADD = version.o ../libntp/libntp.a -lld @LIBPARSE@ DISTCLEANFILES = .version version.c EXTRA_DIST = xntpd.mak ETAGS_ARGS = Makefile.am --- ../xntp3-5.86.8/./util/Makefile.in Tue Oct 29 21:21:31 1996+++ ./util/Makefile.in Fri Nov 8 13:31:55 1996@@ -45,7 +45,7 @@ INCLUDES = -I$(top_srcdir)/include # LDADD might need RESLIB and ADJLIB-LDADD = ../libntp/libntp.a+LDADD = ../libntp/libntp.a -lld #EXTRA_DIST = README TAGS EXTRA_DIST = ETAGS_ARGS = Makefile.am ================================================================ #4 - The next three patches add modifications to support our very fast get time library routine. Two of the changes simply make the existing precision code believe that a 2 usec gettime isn't bogus! --- ../xntp3-5.86.8/./include/ntp_unixtime.h Sun Jun 16 02:54:23 1996+++ ./include/ntp_unixtime.h Fri Nov 8 18:01:27 1996@@ -14,6 +14,15 @@ /* gettimeofday() takes two args in BSD and only one in SYSV */ #ifndef SYS_WINNT+#ifdef SYS_IX86OSF1_MK /* use superfast lib routine */+#include <sys/timers.h>+# define GETTIMEOFDAY(a, b) \+ ( { \+ getclock( TIMEOFDAY, (struct timespec *) (a) ); \+ (a)->tv_usec /= 1000; \+ } )+# define SETTIMEOFDAY(a, b) (settimeofday(a, b))+#else /* !SYS_IX86OSF1_MK */ # ifdef SYSV_TIMEOFDAY # define GETTIMEOFDAY(a, b) (gettimeofday(a)) # define SETTIMEOFDAY(a, b) (settimeofday(a))@@ -21,6 +30,7 @@ # define GETTIMEOFDAY(a, b) (gettimeofday(a, b)) # define SETTIMEOFDAY(a, b) (settimeofday(a, b)) # endif /* SYSV_TIMEOFDAY */+#endif /* SYS_IX86OSF1_MK */ #else /* SYS_WINNT */ extern int winnt_gettimeofday P((struct timeval *)); extern int winnt_settimeofday P((struct timeval *));--- ../xntp3-5.86.8/./xntpd/ntp_proto.c Fri Sep 20 00:28:27 1996+++ ./xntpd/ntp_proto.c Fri Nov 8 18:07:09 1996@@ -2229,7 +2229,11 @@ */ #define DUSECS 1000000 /* us in a s */ #define HUSECS (1 << 20) /* approx DUSECS for shifting etc */+#ifdef SYS_IX86OSF1_MK /* actually has a clock readable in 1 usec! */+#define MINSTEP 0+#else #define MINSTEP 5 /* minimum clock increment (ys) */+#endif #define MAXSTEP 20000 /* maximum clock increment (us) */ #define MINLOOPS 5 /* minimum number of step samples */ --- ../xntp3-5.86.8/./util/precision.c Mon Oct 3 06:04:37 1994+++ ./util/precision.c Mon Nov 4 15:15:40 1996@@ -48,9 +48,13 @@ #define DUSECS 1000000 #define HUSECS (1024 * 1024)+#ifdef SYS_IX86OSF1_MK /* actually has a clock readable in 1 usec! */+#define MINSTEP 0+#else #define MINSTEP 5 /* some systems increment uS on each call */ /* Don't use "1" as some *other* process may read too*/ /*We assume no system actually *ANSWERS* in this time*/+#endif #define MAXSTEP 20000 /* maximum clock increment (us) */ #define MINLOOPS 5 /* minimum number of step samples */ #define MAXLOOPS HUSECS /* Assume precision < .1s ! */ ================================================================ ************* STOP STOP STOP STOP STOP ****************** ************* RUN CONFIGURE BEFORE APPLYING NEXT PATCH ** #5 - With the above set of patches applied, you now run configure. Unfortunately, the config results aren't quite correct, so this next patch fixes config.cache. --- ../xntp3-5.86.8/./config.cache Mon Nov 11 17:26:47 1996+++ ./config.cache Fri Nov 8 18:34:26 1996@@ -31,7 +31,7 @@ ac_cv_func_memmove=${ac_cv_func_memmove='yes'} ac_cv_func_memset=${ac_cv_func_memset='yes'} ac_cv_func_nice=${ac_cv_func_nice='yes'}-ac_cv_func_nlist=${ac_cv_func_nlist='no'}+ac_cv_func_nlist=${ac_cv_func_nlist='yes'} ac_cv_func_ntp_adjtime=${ac_cv_func_ntp_adjtime='yes'} ac_cv_func_ntp_gettime=${ac_cv_func_ntp_gettime='yes'} ac_cv_func_plock=${ac_cv_func_plock='yes'}@@ -112,7 +112,7 @@ ac_cv_struct_tm=${ac_cv_struct_tm='time.h'} ac_cv_type_s_char=${ac_cv_type_s_char='no'} ac_cv_type_signal=${ac_cv_type_signal='void'}-ac_cv_var_adjtime_is_accurate=${ac_cv_var_adjtime_is_accurate='no'}+ac_cv_var_adjtime_is_accurate=${ac_cv_var_adjtime_is_accurate='yes'} ac_cv_var_can_kmem=${ac_cv_var_can_kmem='yes'} ac_cv_var_dtr_powers_dcf77=${ac_cv_var_dtr_powers_dcf77='no'} ac_cv_var_gdt_surveying=${ac_cv_var_gdt_surveying='no'}================================================================ ******** STOP STOP STOP STOP STOP ****************** **** RUN CONFIGURE (AGAIN) BEFORE APPLYING THE NEXT PATCH **** #6 - With the above, config.h is *almost* right. However, the following patch needs to be applied. Of course, this has to be done anytime configure is run again. --- ../xntp3-5.86.8/./config.h Mon Nov 11 17:30:50 1996+++ ./config.h Mon Nov 11 13:46:20 1996@@ -1,6 +1,8 @@ /* config.h. Generated automatically by configure. */ /* config.h.in. Generated automatically from configure.in by autoheader. */ +#define SYS_IX86OSF1_MK 1+ /* Define if on AIX 3. System headers sometimes define this. We just want to avoid a redefinition error message. */@@ -58,6 +60,7 @@ #define DES 1 /* reference clock interface */+#define REFCLOCK 1 /* #undef REFCLOCK */ /* ACTS modem service */================================================================ ****** Now the normal make will produce a working osf1mk system ***** dlm
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -