From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <80633e2d04092101557fd556a2@mail.gmail.com>
Date: Tue, 21 Sep 2004 10:55:01 +0200
From: Christian Grothaus <christian.grothaus@gmail.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
In-Reply-To: <80633e2d0409201234781e65e6@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
References: <80633e2d0409201234781e65e6@mail.gmail.com>
Subject: [9fans] Re: jet-lagged fossil
Topicbox-Message-UUID: e5d6398e-eacd-11e9-9e20-41e7f4b1d025

I should have looked in the archives more thoroughly:
The problem with automatic snapshots has been reported before,
and a possible workaround is setting the timezone to 'GMT 0'.

I found out that it also works in any timezone, when the
real time clock is set to GMT and timesync is run without
the 'L' flag.
(Then, of course, Windows doesn't get the time correctly, but
never mind. I prefer this solution at the moment.)

But I really don't understand at the moment, why it doesn't
work. Here is the beginning of the snapEvent() function
in /sys/src/cmd/fossil/fs.c:

-----------------------------------------------------------
static void
snapEvent(void *v)
{
	Snap *s;
	u32int now, min;
	Tm tm;
	int need;

	s = v;

	now = time(0)/60;
	vtLock(s->lk);

	/*
	 * Snapshots happen every snapMinutes minutes.
	 * If we miss a snapshot (for example, because we
	 * were down), we wait for the next one.
	 */
	if(s->snapMinutes != ~0 && s->snapMinutes != 0
	&& now%s->snapMinutes==0 && now != s->lastSnap){
		if(!fsSnapshot(s->fs, nil, nil, 0))
			fprint(2, "fsSnapshot snap: %R\n");
		s->lastSnap = now;
	}
        .
        .
        .
-----------------------------------------------------------

So the setting of the timezone should have no influence on
periodic non-archival snapshots, since the time is computed
with time()?!

Christian