Re: [PATCH 1/7] event/octeontx: move eventdev octeontx test to driver

[Bruce Richardson <bruce.richardson@intel.com>]

On Wed, Dec 13, 2017 at 04:49:47PM +0530, Pavan Nikhilesh Bhagavatula wrote:
> On Wed, Dec 13, 2017 at 10:19:51AM +0000, Van Haaren, Harry wrote:
> > > -----Original Message-----
> > > From: Pavan Nikhilesh [mailto:pbhagavatula@caviumnetworks.com]
> > > Sent: Tuesday, December 12, 2017 7:27 PM
> > > To: jerin.jacob@caviumnetworks.com; Richardson, Bruce
> > > <bruce.richardson@intel.com>; Van Haaren, Harry
> > > <harry.van.haaren@intel.com>; Eads, Gage <gage.eads@intel.com>;
> > > hemant.agrawal@nxp.com; nipun.gupta@nxp.com; Ma, Liang J
> > > <liang.j.ma@intel.com>
> > > Cc: dev@dpdk.org; Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> > > Subject: [dpdk-dev] [PATCH 1/7] event/octeontx: move eventdev octeontx test
> > > to driver
> > >
> > > Move octeontx eventdev specific test (test_eventdev_octeontx.c) to
> > > driver/event/octeontx.
> >
> > <snip patch content>
> >
> > Replying to 1st patch, as no cover letter;
> >
> > Summary of patchset:
> > - Move tests for a specific Eventdev PMD into the PMD dir: drivers/event/x/x_selftest.c
> > - Enable self tests to run when passed the vdev arg "self-test=1"
> >
> >
> > A few comments on this change;
> >
> > 1) We should not lose the capability to run tests as part of the existing unit testing infrastructure. We should not fragment the testing tool - requiring multiple binaries to test a single component.
> >
> > From discussion on #IRC, it seems reasonable to call  rte_eal_vdev_init()  with "self-test=1" from the test/test/ code, and then we can continue to use the existing test infrastructure despite that the actual tests are now part of each PMD.
> >
> > 2) We should not copy/paste TEST_ASSERT macros into new test files. Abstracting the TEST_ASSERT and other macros out to a header file would solve this duplication.
> >
> 
> I initially thought of abstracting the macros but couldnt find a suitable file
> to place them in we have two options here, one is to use CFLAGS and include
> test.h directly (dirty) or have rte_assert/test in eal/common/inlcude.
> 
> Thoughts?
> 
If other device types, e.g. ethdev or cryptodev, also take the approach of
having a self_test API (something I think would be a good thing, and I
actually hacked together when working on the i40e rx and tx code), I
think we should look to have an rte_test.h header file for such macros
to avoid duplication.
At this point, moving them to an EAL include may not be worth it for
just eventdev.

/Bruce