Dear Marek Vasut,
In message 201203160945.40467.marex@denx.de you wrote:
- You add code size even for systems which don't enable "bootz" support.
This is a good point, do you agree to introduce CONFIG_SUPPORT_RAW_RAMDISK to enable this feature?
Fine with me - please also document the new option.
- You change the syntax of all boot related commands where a ramdisk address may be passed (i. e. "bootm") without documenting it.
Right, this should be documented in the top-level README file? Or is there some other place?
Hm... I have no strong opinion here.
- Using this syntax with a mkimage wrapped ramdisk image will cause bad things to happen, even if you specific the correct size. The API should be more robust.
It actually won't. If you use mkimage warped ramdisk, it'll recognise it's magic numbers and simply ignore the stuff behind ":". Though I hope noone in sane mind would do that.
Argh... Passing arguments that silently get ignored under certain conditions is not something I like to see.
And are you sure hat it will work like that? [I mean, has this been tested?]
/** Check if rd_len was manually overridden, if it was,* we're loading RAW ramdisk.*/This is undocumented policy, and I dislike such a side-effect based design (if we have a size parameter, it must be a raw image).
It must not, but if it is, it'll be used as raw.
Sorry, I cannot parse this.
Best regards,
Wolfgang Denk