On Thu, 21 Mar 2024 10:49:06 -0700 Tyler Retzlaff <roret...@linux.microsoft.com> wrote:
> On Thu, Mar 21, 2024 at 10:16:36AM -0700, Stephen Hemminger wrote: > > On Thu, 21 Mar 2024 10:11:31 -0700 > > Tyler Retzlaff <roret...@linux.microsoft.com> wrote: > > > > > > > > > > > > > > +Console timestamp > > > > +~~~~~~~~~~~~~~~~~ > > > > + > > > > +On Linux and FreeBSD, an optional timestamp can be added before each > > > > +message by adding the ``--log-timestamp`` option. > > > > > > i don't think Windows should be excluded from the feature, in general > > > core libraries that get features should cover all platforms. > > > > The problem is that current Linux/FreeBSD log code depends on fopencookie() > > to work. I could not find a way to do that on Windows. > > but that looks like it is part of general log initialization. surely > when we are talking about logging a timestamp to a console we're just > talking about a standard stream and a format string for Windows? > > i'll try to study more to gain a better understanding. > > thanks What the code does is take what rte_log created and add more data before actual printing it. The other option is doing this at rte_log() but that will break the syslog side. I am not sure that having log go directly to syslog makes sense anyway on a modern Linux system. If DPDK is being used by a daemon, that daemon will be a systemd service. And systemd has simple way to make stderr go to syslog. So if we drop the syslog support, then the whole thing gets more generic and simpler.
