[solved] coreutils - date ( misleading documentation? )

[paziu]

Hello Everybody,

I am trying to find out more about the syntax of the date command - I looked in the man pages from 2002, 2010, and several I found on the net.
The problem I got might seem to look trivial to some, but it cost me some explanations, and of course admitting to the mistake.
For years I have been under impression, that only "date -s" changes the time, this is what I learned in the 90s on SCO.
For one or another reason I issued "date nnnnnnnnn" command using a string format that I wanted to test if it will "spit out" the correct date, before I decide to set it. Unfortunately it changed the year to 2030 - 10 seconds after the fact I re-set it to the correct date, but than I found out that "-s" is not needed since the system changed the time, and it caused a "bit" of a problem on the production database system.
There is no trace on any man page, that says that the time can be changed without providing any switches, options, etc.
Of course, "mea culpa" but on the other hand it seems like 'date', like on the AT&T version, should be required to provide the "-s" to imply the date change, otherwise either error out, or >>>display<<< the date reflecting the date string without changing the time....

Well, I might be completely wrong, in any case I would like to hear some opinions from you... not opinions about making such mistake, but about the "destructive" feature of date without "-s" option, and what caused that Linux 'date' ( coreutils ) is so different from some Unix systems.

Thank you everybody, have a good day,

Mike

[eccerr0r]

Oh...this is trying to recall back neurons from more than 20 years...

From what I remember, my old TRS-Xenix box did not use -s ... It had to be a valid date, else it would not set the date. I guess this is similar to what I've been using with Linux for a while.

I so rarely use 'date' anymore to set time, mostly do it in BIOS if I can/have to reboot the machine. If I have network access, I usually use rdate or ntp to set the date, which is a lot more accurate than carefully watching the seconds...
_________________
Core-i7-2700K@4.1GHz/8GB RAM/180GB SSD/Intel HD3000 graphics
What the heck am I advocating?

[paziu]

Thanks for your reply, although it does not present much value to me,

I understand, 20 years ago is in the "past", but lack or improper documentation for something as basic as "date" is "present".
I am not asking what way of setting time is better, there is situations where NTP is not available, or you have 1000 users on the system, so going in the bios to set the time during the day would cost you your job.

Thanks again,

Mike

[khayyam]

[John R. Graham]

[paziu]

Thank you khayyam,

The display option would be the one I would rather expect to be the default behavior, instead of set.
In my case, I wanted to validate a date string, which did not include the year at the end - but '30' for the seconds... so the year changed to 2030.

Using "-d" is a very good idea, thank you.

I will still "stick" to my opinion, that the man page should include the default behavior if no option is provided... I am quite sure I am not the first and not the last who did this mistake....


Thank you khayyam - "-d" is now in my blood stream.

Mike

[paziu]

Thank you John,

[khayyam]

[eccerr0r]

and don't run with scissors as root unless you want to cut something...
(also all these lineages of Un*x, some were different... BSD releases, SysV, POSIX,...) I mean, come on, why does ps have different options on different flavors?