[ovs-dev] [PATCH 1/2] Report timestamps in millisecond resolution in log messages.

Ben Pfaff blp at nicira.com
Fri Sep 13 16:07:06 UTC 2013


On Thu, Sep 12, 2013 at 06:23:22PM -0700, Paul Ingram wrote:
> On Sep 11, 2013, at 10:39 PM, Ben Pfaff <blp at nicira.com> wrote:
> 
> > It's reasonable enough.
> > 
> > I didn't really like the %. format specification for this.  It is
> > inflexible--what if someone wants only tenths of a second or wants
> > microseconds?  It is also somewhat hard to parse since it uses %
> > (naturally enough though).  Anyway I thought about it for a minute and
> > concluded that one can reasonably just replace '#' in the strftime()
> > formatted output with fractional seconds, so that "%S.###" outputs
> > milliseconds, "%S.#" outputs tenths, etc.  Also this allows for
> > formatting properly for cultures that prefer to use a comma for the
> > decimal point.
> > 
> > I made this change.
> > 
> > Some indentation was weird (I think you indented with tabs instead of
> > spaces and we must have different tab stops).  I fixed it.
> > 
> > NEWS and manpages needed an update.  I did it.
> 
> Ok, using ### works too. I like it. (I suppose something like %3. (or
> choose a format character other than '.') would be the most
> strftime-like.)

Yes, that would be a more natural strftime syntax.  I think the one
proposed here is probably OK too.  (It seems unlikely that anyone has a
desperate need for a # inside a date format.)

> I made a small change to your ### scheme to require runs of
> consecutive '#' characters and to reprint on each new run.
> 
> So given seconds and milliseconds of 12.789, the format string "%S.#
> %S.###" will produce "12.7 12.789" (instead of "12.7 12.890" where the
> 789 is spread across the two runs of '#' characters).

OK, I'm fine with that version too.

I'll apply this soon.



More information about the dev mailing list