> From: Stephen Hemminger [mailto:step...@networkplumber.org] > Sent: Tuesday, 27 June 2023 16.49 > > On Tue, 27 Jun 2023 09:40:01 +0200 > Morten Brørup <m...@smartsharesystems.com> wrote: > > > > From: Stephen Hemminger [mailto:step...@networkplumber.org] > > > Sent: Monday, 26 June 2023 20.42 > > > To: dev@dpdk.org > > > Cc: Stephen Hemminger > > > Subject: [PATCH v4 0/5] Logging related patchs > > > > > > This patch set rebases and extends some earlier work on logging. > > > > > > Stephen Hemminger (5): > > > eal: unify logging code for FreeBsd and Linux > > > eal: turn off getopt_long error message during eal_log_level > > > eal: skip stdio on console logging > > > eal: move logging initialization earlier > > > eal: add option to put timestamp on console output > > > > Series-acked-by: Morten Brørup <m...@smartsharesystems.com> > > > > There are a few bugs, still working them out. > In initial review feedback there was desire for timestamp feature > to be global. I.e secondary process would show time since start > of primary. But this would hard to implement due to chicken/egg > issues during argument parsing during startup; so made a decision > that it wasn't worth doing.
I suppose it could be implemented with a piece of shared memory set by the primary process and read by secondary processes, holding the start time offset from a system global timer (e.g. CLOCK_MONOTONIC) and a tristate variable telling if timestamps are configured - which, if true or false, overrides any timestamp parameter given to the secondary process. Anyway, secondary process support is far from perfect, so I wouldn't worry too much about getting timestamp parameters from the primary process to the secondary process either. In short: I agree with your decision.
