On Mon, Nov 22, 2021 at 02:56:58AM +0100, Ansuel Smith wrote:
quoted
Maybe you could keep qca8k_read and qca8k_write and make them return
regmap_read(priv->regmap, ...), this could reduce the patch's delta,
make future bugfix patches conflict less with this change, etc etc.
What do you think?
Problem is that many function will have to be moved to a separate file
(for the common stuff) and they won't have qca8k_read/write/rmw...
So converting everything to regmap would be handy as you drop the
extra functions.
But I see how reworking the read/write/rmw would massively reduce the
patch delta.
When we will have to split the code, we will have this problem again and
we will have to decide if continue using the qca8k_read/write... or drop
them and switch to regmap.
So... yes i'm stuck as if we want to save some conflicts we will have to
carry the extra function forver I think.
(Wonder if the conflict problem will just be """solved""" with the code
split as someone will have to rework the patch anyway as the function
will be located on a different file)
Yeah, well, if you have to split then you have to split. It probably
won't be pretty, with a lot of code added for v5.16 and now for v5.17,
so it hasn't had time to reach to a larger pool of users and get cleaned
of bugs which you didn't notice. But what can you do. Other than wait
for a few months for the code to stabilize (which is counterproductive
in its own ways), probably nothing.