testpmd logging

[Stephen Hemminger]

On Tue, 8 Nov 2022 11:24:53 -0500
Owen Hilyard <ohilyard at iol.unh.edu> wrote:

> On Tue, Nov 8, 2022 at 9:37 AM Honnappa Nagarahalli <
> Honnappa.Nagarahalli at arm.com> wrote:  
> 
> > + Lijuan, Owen, Juraj
> >  
> > > -----Original Message-----
> > > From: Thomas Monjalon <thomas at monjalon.net>
> > > Sent: Tuesday, November 8, 2022 7:02 AM
> > > To: Aman Singh <aman.deep.singh at intel.com>; Yuying Zhang
> > > <yuying.zhang at intel.com>
> > > Cc: david.marchand at redhat.com; dev at dpdk.org
> > > Subject: testpmd logging
> > >
> > > Hello,
> > >
> > > I see testpmd is doing log with EAL entity. Example:
> > >       RTE_LOG(ERR, EAL, "The device: %s has been added!\n"
> > > We could also discuss the log level.
> > >
> > > Please can we work on replacing these logs?  
> >  
> 
> If we want to revisit logging in testpmd, even though I know I'm invoking a
> case of second system syndrome here, I would love to see some form of
> structured logging output. Even just writing out records to a CSV file with
> proper escaping would work great, or doing a json object per log record as
> a backup. JSON might be easier since libjansson is already in DPDK, and we
> could disable the json logging flag if libjansson was not present at
> compile time.
> 
> 
> > DTS uses some of the logs from testpmd to validate the results. I am not
> > sure what all logs are used.  
> 
> 
> The old version of dts (the one with its own repository) is fairly tightly
> bound to testpmd's logging.
> 
> > We need to choose how to log (printf or registered log entity) and be  
> > > consistent.
> > > What do you think?  
> >  
> 
> Registered log entries are much better in my opinion since they provide a
> more machine-friendly format and DTS has to parse testpmd output.

My thoughts.
  1. testpmd should use its own log level, not EAL
  2. testpmd could easily have an option "--json" to format output in JSON
  3. testpmd is a mess some logs use RTE_LOG() some go to stdout, some to stderr.