On Thu, Dec 11, 2008 at 2:57 PM, Carl-Daniel Hailfinger < c-d.hailfinger.devel.2006@gmx.net> wrote:
On 11.12.2008 02:19, ron minnich wrote:
On Wed, Dec 10, 2008 at 5:12 PM, Peter Stuge peter@stuge.se wrote:
Carl-Daniel Hailfinger wrote:
In the past, reviews were mostly centered on coding style (not only cosmetics, but also code flow) and general sanity. While that is definitely needed, I propose another layer on top of this:
Verification of the code and comments against data sheet recommendations and documentation.
I feel very strongly that we do not need more layers.
Let me rephrase that. I do not want to hold back any commits. That would be insane.
However, whenever someone goes through in-tree code and checks the code against the data sheets and thinks that the code is OK, he/she should be free (not obliged) to improve annotation/comments and add a comment that he/she verified the code against the data sheets.
IMO, if we do this, we need to also require the datasheet revision/release date and if any update notes or errata are taken into account.
-Corey
There are problems anyway. What if the doco are known to be wrong, due to an NDA, and you can't even say "the doco is wrong".
If the NDA is so strict that you can't even say that the docs are wrong, how are you preventing erroneous "bugfixes" from being committed? I honestly have no idea how to solve that problem and it exists regardless of whether my RFC is accepted or not.
Sorry, I vote with peter.
I understand that because my original RFC implied things I didn't want to suggest. How about the new text I proposed above?
Regards, Carl-Daniel
-- coreboot mailing list: coreboot@coreboot.org http://www.coreboot.org/mailman/listinfo/coreboot