On Sat, Jul 06, 2024 at 11:17:52AM -0600, deichert@placebonol.com wrote:
> what you suggest sounds like a really bad idea
>
> Time is an imperfect construct, with your suggestion you have to have 100% confidence that system local time is always perfect. I've been doing this a long time and can recall many instances when an issue arose because time was off.
>
> On July 6, 2024 7:01:27 AM MDT, Anon Loli <anonloli@autistici.org> wrote:
> SNIP
> >(my last email on this thread, about datetime "version check" is needed to
> >understand the meaning of this)
> >or make(1) can handle this when say building /usr/src, it can check datetime of
> >the source files (.c, .h), and then the datetime of object files (.o).. now
> >there might be a confusion between CVS repository datetime and local host
> >time, so perhaps convert everything to UTC unixtime, and perhaps CVS should
> >handle the differential between CVS and local datetime (if needed at all)
> >
> >And voila, that should handle object files so that they can't get outdated.
> >
>
That's precisely what I meant by "handling the differential", I suppose.
What do you mean that system localtime is always perfect? NTP already exists..
And I think that the local datetime should be sufficient, no matter how much
you try, the files that you get from a CVS or tar.gz should always be older
than any changes you make by hand, for example
And then when you say go and update, same should apply, but the opposite
direction, the CVS/tar.gz should always have newer datetime (assuming changes
to such and such files happened)
No comments:
Post a Comment