Re: [PATCH 2/3] eal: uninline rte_str_to_size

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

On Mon, Aug 22, 2022 at 09:24:47AM +0200, Morten Brørup wrote:
> > From: Dmitry Kozlyuk [mailto:dmitry.kozliuk@gmail.com]
> > Sent: Sunday, 21 August 2022 22.50
> > To: dev@dpdk.org
> > Cc: Dmitry Kozlyuk; Ray Kinsella
> > Subject: [PATCH 2/3] eal: uninline rte_str_to_size
> > 
> > There is no reason for rte_str_to_size() to be inline.
> > Move the implementation out of <rte_common.h>.
> > Export it as a stable ABI because it always has been public.
> > 
> > Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
> 
> Acked-by: Morten Brørup <mb@smartsharesystems.com>
> 
> > ---
> > Now <rte_common.h> doesn't need to #include <ctypes.h> and <stdlib.h>,
> > but removing them breaks some DPDK code, may break user code too.
> > I'm not sure what is the compatibility policy in this regard.
> > If such a breakage is allowed, I'd remove includes and fix DPDK code.
> > 
> 
> The question I'm asking myself here is: Do we want rte_common.h to include common headers like these, just so we don't need to include them elsewhere? I think not.
> 
> I'm in favor of the principle of keeping it clean: Remove them from rte_common.h, and deal with the consequences.
> 
> If we keep them, we will forget why they are there, and some day in the future, someone will ask what these unused headers are doing in <rte_common.h>.
> 
+1
Since removing headers is a build-time issue only and not runtime, I think
we should just remove them.

/Bruce