[ntp:questions] Re: New NTP version for Windows with GUI setup available
mayer at ntp.isc.org
Thu Jan 5 04:05:54 UTC 2006
David J Taylor wrote:
> Martin Burnicki wrote:
>>David J Taylor wrote:
>>>Martin Burnicki wrote:
>>>>There are still some pieces of code in the NTP package where this
>>>>extra handling for Windows has not yet been implemented and thus
>>>>needs some cleanup.
>>>Just a small point, but it would be nice if ntpd used the event
>>>number ID on Windows as well. At the moment, all events have an
>>>event ID of 3, whereas there could be different numbers for (say)
>>>startup, shutdown, time resets, change of sync source,
>>>initialisation messages etc. The Windows event viewer allows you to
>>>filter by event ID.
>>You're right, that would be a good idea for the Windows version.
>>However, this would require an additional bunch of changes since
>>every log message had to be assigned a number, and the whole logging
>>stuff which is used under a bunch of operating systems had to be
>>modified or rewritten.
>>Remember, most of the core developers are Unix people ;-))
> OK, Martin, but Danny has already said that he has two different logging
> modules for network and non-network errors, so that would be a start,
> providing two numbers. I had rather assumed that UNIX might have
> supported event IDs as well (I'm sure VMS did!). Add a third category for
> Windows-specific events, and we've got three different event IDs without a
> lot of effort.
You misunderstood me. It's just a different name so that one can call
GetLastError() and the other WSAGetLastError(). Otherwise they are
identical. The ID's are provided by the first argument: LOG_INFO,
LOG_ERROR, etc and they are just mapped to the ID.
VMS works very differently and has 3 parts: the facility (such as
SYSTEM), the error category (S, E, I, W, etc.) and the id of the
specific message. The message facility in VMS is interesting. One day
I'll tell you about some of the April Fool's Day messages that were
created some other time,
> Might be a good start - I don't think /every/ event needs a different ID
> (although it might be nice).
More information about the questions