Re: [PATCH v1 1/4] lib/vsprintf: Print time64_t in human readable format

[Andy Shevchenko]

On Mon, Sep 30, 2019 at 10:08:09PM +0200, Alexandre Belloni wrote:
> On 26/07/2019 16:20:37+0300, Andy Shevchenko wrote:
> > On Thu, Jan 10, 2019 at 10:58:58PM +0100, Alexandre Belloni wrote:
> > > On 08/01/2019 16:25:28+0100, Petr Mladek wrote:
> > > > On Fri 2019-01-04 21:30:06, Andy Shevchenko wrote:
> > > > > There are users which print time and date represented by content of
> > > > > time64_t type in human readable format.
> > > > > 
> > > > > Instead of open coding that each time introduce %ptT[dt][r] specifier.
> > > > > 
> > > > > Few test cases for %ptT specifier has been added as well.
> > 
> > > > > +void time64_to_rtc_time(time64_t time, struct rtc_time *rtc_time)
> > > > > +{
> > > > > +#ifdef CONFIG_RTC_LIB
> > > > > +	rtc_time64_to_tm(time, rtc_time);
> > 
> > > > I wonder if the conversion between struct tm and rtc_time
> > > > might be useful in general.
> > > >
> > > > It might make sense to de-duplicate time64_to_tm() and
> > > > rtc_time64_to_tm() implementations.
> > 
> > > Looking at 57f1f0874f42, this seemed to be the plan at the time
> > > time_to_tm was introduced but this was never done. Seeing that tm and
> > > rtc_time are quite similar, we could probably always use time64_to_tm as
> > > it is more accurate than rtc_time64_to_tm as the latter assumes a
> > > specific year range.
> > 
> > So, do I understand correctly that dropping #ifdef along with
> > rtc_time64_to_tm() call is sufficient for now?
> > 
> 
> I'd be fine with that.

Good, thanks! I'll send v2 soon.

> > > Maybe be rtc_str should take a struct tm instead of an rtc_time so
> > > time64_to_rtc_time always uses time64_to_tm.
> > 
> > Because this one, while sounding plausible, maybe too invasive on current
> > state of affairs.
> 
> Well, if the kernel struct tm had an int tm_year instead of long
> tm_year, then you could simply cast a struct rtc_time to a struct tm.

I don't think so. It will be error prone from endianess prospective on
64-bit platforms.

> I'm not sure was was the rationale to have a long, especially since
> userspace has an int.

Yeah, this is strange, I guess we simple may long -> int in kernel's struct tm.

-- 
With Best Regards,
Andy Shevchenko