[dpdk-dev] [PATCH 1/1] eal: return true or false from lcore role check function

Thomas Monjalon thomas at monjalon.net
Fri Jan 12 19:04:31 CET 2018


12/01/2018 19:01, Carrillo, Erik G:
> From: Thomas Monjalon [mailto:thomas at monjalon.net]
> > 12/01/2018 00:09, Carrillo, Erik G:
> > > Hi Aaron,
> > >
> > > From: Aaron Conole [mailto:aconole at redhat.com]
> > > >
> > > > Hi Erik,
> > > >
> > > > Erik Gabriel Carrillo <erik.g.carrillo at intel.com> writes:
> > > >
> > > > > Update rte_lcore_has_role() so that it returns true/false instead
> > > > > of success/failure.
> > > > >
> > > > > Fixes: 78666372fa2b ("eal: add function to check lcore role")
> > > > >
> > > > > Signed-off-by: Erik Gabriel Carrillo <erik.g.carrillo at intel.com>
> > > > > ---
> > > >
> > > > I believe this breaks the published abi - Success is now 'true', and
> > > > failure is 'false';  previously success would be 0 == false.  You'll
> > > > need to invert the test, or note that the abi is breaking (since
> > > > semantically any caller will need to invert the test).
> > >
> > > Good point.  Though it seems like an API change rather than an ABI change
> > to me, would it still be handled the same way in terms of notice?  Also,  the
> > ABI policy states, "ABI breakage due to changes such as reorganizing public
> > structure fields for aesthetic or readability purposes should be avoided."
> > Perhaps I should go with an alternate patch that fixes the caller.
> > 
> > Most of the times, an API change is an ABI change.
> > Please make a deprecation notice.
> 
> Ok, thanks Thomas - will do.  Should I mark the above patch as "deferred" for the time being?

Yes, thanks
All deferred patches are set to New when starting a new release cycle.
So it should not be lost :)



More information about the dev mailing list