Wolfgang Denk wd@denx.de wrote on 2011/01/17 23:11:59:
Dear Joakim Tjernlund,
In message OF6A573B77.9119DEB0-ONC1257814.0069A574-C1257814.007532C7@transmode.se you wrote:
It would break only if link address != load address. That is, if you want to use my new CONFIG_SYS_TRUE_PIC feature and be able to load u-boot at any address regardless of link address you would have to add LINK_OFF calls into print_83xx_arb_event() too if you want to use it.
Doesn't this add a requirement for future generic pre-relocation code to comply with, to avoid breaking your board?
Yes, but I don't mind if my board breaks from time to time. After all it isn't in u-boot so I have had to deal with quite a few breakages already. It is my hope this new feature will spread to other boards as time pass.
You mean you ask to add code that is not only highly fragile but even known to be broken for other boards than yours, and the only board that uses the feature and has been tested with it "isn't in u-boot" ?
No other board is broken. This new function is neutral to other boards. I am merely saying as my board is the first user of this new feature I expect minor breakage of my board from time to time when someone adds a new function that needs LINK_OFF to work on my board but forgets to actually add the LINK_OFF call. Once more boards uses my new feature this problem goes away.
Wolfgang, once you indicated you were interested in such feature as I have added but my first impl. had LINK_OFF calls all over the place, still you were tempted to add the feature. Now that I have reduced the LINK_OFF calls to a minimum you suddenly want to reject it even though >95% of the LINK_OFF calls are gone. Why this change of heart?
Jocke