On Mon, 2012-03-19 at 16:39 -0700, ron minnich wrote:
we made a decision in the early days to NOT include the chip part name in the function names. This was not a mistake or omission, it was a deliberate design choice.
The reason to name things this way is because a board is composed of a set of parts, and the partname is in the file name path. Hence, the board can be constructed of files calling functions with generic names, and the generic functions are provided by files chosen in the config. In some cases, it has proven trivial to port a mainboard to a new chipset by changing only the config to use a different chip. The fact that the function name did not include the chipname made this trivial.
I agree with this design. It works as the file-specific functions are accessed only via the chip-specific _ops structure. And those structures do have the chip name in them.
If we think that we really need the chip name in the function, ok, but it's a change in the way we designed the build process. Maybe it's a change we have to make. I am not yet convinced.
Bootblock is special, probably for simplicity the chip objects were dropped. BTW: There are boards with two super-ios.
KM
On Tue, Mar 20, 2012 at 4:58 AM, Kyösti Mälkki kyosti.malkki@gmail.com wrote:
Bootblock is special, probably for simplicity the chip objects were dropped. BTW: There are boards with two super-ios.
the objects were never used because this was usually a romcc-based stage. And, yes, there are boards with two super-ios, and have been for a long time: we always have to relax our rules :-)
thanks!
ron