On Mon, Sep 14, 2020 at 09:00:19AM +0100, Yang, SteveX wrote:
> Hi Bruce,
>
> > -----Original Message-----
> > From: Bruce Richardson <bruce.richard...@intel.com>
> > Sent: Thursday, September 10, 2020 5:24 PM
> > To: Yang, SteveX <stevex.y...@intel.com>
> > Cc: dev@dpdk.org; Lu, Wenzhuo <wenzhuo...@intel.com>; Xing, Beilei
> > <beilei.x...@intel.com>; Yang, Qiming <qiming.y...@intel.com>
> > Subject: Re: [dpdk-dev] [PATCH v1] app/test-pmd: fix meson build failed
> > when used latency stats lib
> >
> > On Thu, Sep 10, 2020 at 02:20:18AM +0000, SteveX Yang wrote:
> > > meson build cannot find the header rte_latencystats.h when build DPDK
> > > first time or never installed DPDK lib after build via meson/ninja.
> > >
> > > Because the corresponding header directory isn't included after
> > > enabled RTE_LIBRTE_LATENCY_STATS flag.
> > >
> > > Add the header file location and link library to meson.build of
> > > test-pmd
> > >
> > > Signed-off-by: SteveX Yang <stevex.y...@intel.com>
> > > ---
> >
> > As before, I think we have a mismatch in old vs new macro names which is
> > why this wasn't caught before. Therefore I suggest we just change testpmd
> > to use the new macro to keep things standardized. [Other alternatives if we
> > want to keep compatibility are to rename the latencystats directory to have
> > an underscore in it, or to override the name (which I'd rather not go for)]
>
> Got it, I can change the old macro to new one RTE_LIBRTE_LATENCYSTATS.
>
> >
> > > app/test-pmd/meson.build | 4 ++++
> > > 1 file changed, 4 insertions(+)
> > >
> > > diff --git a/app/test-pmd/meson.build b/app/test-pmd/meson.build index
> > > db0ff02eb..6ee7af750 100644
> > > --- a/app/test-pmd/meson.build
> > > +++ b/app/test-pmd/meson.build
> > > @@ -9,6 +9,10 @@ if dpdk_conf.has('RTE_LIBRTE_BOND_PMD')
> > > cflags += '-I' + meson.source_root() + '/drivers/net/bonding'
> > > endif
> > >
> > > +if dpdk_conf.has('RTE_LIBRTE_LATENCYSTATS')
> > > +cflags += '-I' + meson.source_root() + '/lib/librte_latencystats'
> > > +endif
> > > +
> >
> > Rather than working with long cflags, just add latencystats as a dependency
> > of testpmd.
>
> If I add the latencystats as a dependency, that means the flag will be
> **always enabled** no matter if the RTE_LIBRTE_LATENCY_STATS flag is set from
> compiler cmdline.
> Is it expected behavior for test-pmd app?
>
Not sure what you mean here, but so long as latency stats is available it
should be a dependency of testpmd, so that the functionality is available.
/Bruce
