Wolfgang Denk wrote:
In message 474C390C.8060007@qstreams.com you wrote:
Sounds good to me. It'd be cooler if make could detect the presence/absence of <linux/mtd-user.h>, but I'm not enough of a build system whiz to know how to do that.
Where should he detect this? On the build host? Who says the target uses the same kernel / MTD version? Who says you always want to build for the same kernel / MTD version?
I think auto-detection is basicly impossible (at least without making the build depending on the knowledge of a "corresponding" Linux kernel tree, and I think Markus' new approach is more straightforward.
Right. I said it would be cool, not necessarily possible. This brings up a problem with this utility - the version of headers that it's built against is at the mercy of however the cross compiler's set up, not necessarily against the one the user really wants. Maybe we should force the user to define a path, and #error if it isn't defined. This path could be added to C_INCLUDE_PATH and then we could do all sorts of Makefile trickery. Just a thought...
- Can you also please add the other part of my patch that
creates/destroys the 'fw_setenv' symbolic link? It's a bit annoying that the documentation mentions the 'fw_setenv' utility, but it doesn't really exist without a bit of help.
- Please include changes to the relevant documentation (README + wiki)
so that people know to when and how to set MTD_VERSION. This may prevent a few 'RTFM' e-mails.
Agreed for README. The wiki cannot be "patched" like this. We can Markus only make promise that he will update the information there, too.
Sounds good. Promises are fun.
regards, Ben