[ntp:questions] Re: VMS/UCX build?

Richard B. Gilbert rgilbert88 at comcast.net
Wed Dec 3 04:57:18 UTC 2003


Dave,

Setting the clock and reading it are documented, of course.  Slewing the 
clock is not supportted but the how-to is known and I can find it out; 
it's been used for years by people who did not want to jump the clock to 
set daylight or standard time during the dark ages when VMS kept local 
time.  And yes, the VMS sources exist and you can get listings if you 
pay big bucks and sign a license agreement.  For even bigger bucks you 
can, or once could, get machine readable sources and a set of compilers 
and other tools that might allow you to build VMS from scratch (they 
won't guarantee it).

My first problem is to get the rest of the NTP V4.2 package to compile 
and find VMS equivalents or replacements for Unixisms like "non-blocking 
I/O" in fcntl( ).  If I can get that far, I'll see if I can write code 
to dink with the clock frequency.
If I manage to come up with anything useful, I will be more than willing 
to share it.

In the mean time, I have solved, sort of, my original problem.  The UCX 
V4.2 implemenation of NTP does not translate host names to IP 
addresses!!!!!  Once I put the dotted decimal form of the server 
addresses in my ntp.conf  things started to sort of work!   I'm seeing 
excursions of plus or minus 60 to 80 milliseconds in clock offset in 
some of these systems which doesn't seem right.  Still, we are in the 
general neighborhood of the correct time which is a distinct improvement.

David L. Mills wrote:

>Richard,
>
>DTSS was built for the Enterprise community and still shows up in the
>Tru64 kernel for the Alpha. My point being that riffing through the VMS
>kernel sources (should they exist) should reveal how the clock interface
>works. Then, you should be able to wrangle systime.c in libntp to fit.
>Look for modules with names beginning "utc". Also look in the
>documentation for system calls beginning with that prefix.
>
>Dave
>
>"Richard B. Gilbert" wrote:
>  
>
>>I installed  VMS Posix 2.0 to get a shell that might run configure.
>>
>>psx> ./configure
>>gets an error as follows:  (from config.log)
>>
>>## ----------- ##
>>## Core tests. ##
>>## ----------- ##
>>
>>configure:1386: checking build system type
>>configure:1399: error: /bin/sh ./config.sub
>>  exit (1);
>>..^
>>%CC-I-IMPLICITFUNC, In this statement, the identifier "exit" is implicit
>>  ly decla
>>red as a function.
>>at line number 110 in file _ALPHA4$DKA0:[POSIX$CONT.0003E094$BFS]0004002
>>  B$BFS.;1 failed
>>
>>It did not manage to build any makefiles or config.h.  the %CC-I message
>>above is DEC C complaining gently about an undeclared function.
>>
>>I decided to try something simple.  I compiled ntpdate, got errors, and
>>started building a config.h by hand.  I got to the point where it
>>"bletched" because it could not figure out how to get thirtytwo bit
>>integers.  I tried adding:
>>"#define SIZEOF_INT = 4" to my config.h file and found that that didn't
>>help.   I didn't manage to locate anything that did define SIZEOF_INT.
>>
>>configure also complained, at the terminal, that config.sub was called
>>with too many arguments.  It didn't bother to say how many or what they
>>were.
>>
>>Harlan Stenn wrote:
>>
>>    
>>
>>>Running "configure" will generate the Makefiles from the
>>>Makefile.in files.
>>>
>>>It will also generate config.h.
>>>
>>>I suspect you will need to get some sort of "sh", and probably "sed" and
>>>a few other tools ported over first.
>>>
>>>I am happy to help, but I haven't touched VMS for a Long Time, and back
>>>when I did use it it was only briefly.
>>>
>>>I can certainly make needed changes based on your experience, however.
>>>
>>>And I'd appreciate your documenting your experience at twiki.ntp.org.
>>>
>>>H
>>>
>>>
>>>      
>>>




More information about the questions mailing list