On Wed, Oct 24, 2007 at 12:08:58AM +0200, Uwe Hermann wrote:
I think it's pretty clear that I totally dislike duplicated code
Me too, but it's not worth the effort to fix that in v2.
The bigger problem is more recent boards.
..
worse they contain a _lot_ of duplicated or near-duplicated code.
Yes, likely because of copy-paste development.
Maybe we don't even have to fully merge the boards in v2, some simple cleanups in the current code base would also be helpful, I think.
Disagree. Not for v2.
get_bus_conf.c, irq_tables.c, mptable.c, resourcemap.c etc.
..
create patches to fix _that_ in v2, but I'd really appreciate it if v3 would have that kind of information as a simple config value part of the dts.
This is indeed my idea for v3 too.
There are just 2-3 places where you have to add your board in the #elif.
On the long run (in v3 maybe)
How long is "long" ?
No idea. As short as possible, of course, but just like anyone else I have no release date.
Understood. What I was getting at was that it may be quicker to just forget v2 and work with new shinyness in v3.
I would much rather see this effort go into v3 and a bonus is that
Sure, I'm willing to also provide patches to eliminate any uselessly duplicated code from v3.
My point was that there is little if any duplicate code in v3 and that the v3 design is intended to minimize it - working on v3 is much more interesting, since it solves many of the problems in v2.
many of the boards supported by v2 would be working also with v3.
Can you elaborate? How does eliminating duplicated code help or not help in supporting v2 boards in v3? Those issues are not really related IMO.
That's right. But I meant that moving one v2 port to v3 (part of which is structuring the code a bit better) should make it very easy to add new boards with similar hardware - just like we want.
And, as Stefan pointed out, putting effort into v2 means v3 becomes stale.
I still expect people to work on v2 to some degree since it's much easier to do copy-paste-development there, but I don't think we should encourage any efforts in v2 except as a testbed now that v3 is starting to run on hardware.
//Peter