On 07/09/2022 10:20, Karel Zak wrote: > On Tue, Sep 06, 2022 at 07:31:07PM +0300, Péter Ujfalusi wrote: >> On 06/09/2022 11:47, Karel Zak wrote: >>> On Tue, Sep 06, 2022 at 10:04:36AM +0300, Peter Ujfalusi wrote: >>>> lib/timeutils.c | 8 ++++++++ >>>> 1 file changed, 8 insertions(+) >>> >>> Applied, thanks. >> >> After some tinkering with a script where I needed this, would it make >> sense to change the way how one should be specifying seconds since the >> Epoch time to >> '@'<number> > > Yes, '@' seems more readable. I have no strong opinion about it, maybe > we can support both convention (with and without @). > > Anyway, we really need to add something like "INPUT TIMESTAMP" section > to the man page to describe supported formats. > >> This would match how for example systemd.time requires it [1] >> >> Basically I (or anyone) can use dmesg instead of journalctl in a script >> with the same --since=@<timestamp> parameter. >> >> 'date +%s' prints out just a number. > > It's possible to use --since=$(date +'@%s') > >> If not, then I have made a rookie mistake with the patch, which should >> be fixed (discarding the seconds we got out from the timestamp). > > Do you mean your copy & past tm.tm_sec = 0;? This should be fixed. > >> /me hides > > Come back and send a patch :-) I have sent a fix patch which requires @ as a prefix and keeps the seconds intact: https://lore.kernel.org/util-linux/20220907054141.15608-1-peter.ujfalusi@xxxxxxxxxxxxxxx/T/#u -- Péter
