On 10/09/2015 02:28 AM, Benoît Thébaudeau wrote:
Dear Wolfgang,
On 08/10/2015 23:29, Wolfgang Denk wrote:
Dear Stephen,
In message 56167DB6.3000508@wwwdotorg.org you wrote:
What's the expected correct behavior here?
After successful loading the data to memory, load_addr should be set correctly, for all commands. In the error case, the value of load_addr is undefined.
Is this documented anywhere? If not, I'm not convinced that there's a contract to be followed; it "just happens" that some filesystem-related commands work(ed) that way (and as Benoît pointed out, apparently some don't irrespective of the mentioned patch).
I'm afraid it's not documented, but it is what I would consider a sane and consistent behaviour. If we intend to implement POLA [1] (and I very much think we should), this is how U-Boot should behave.
[1] https://en.wikipedia.org/wiki/Principle_of_least_astonishment
I'm not certain that this would be the least astonishing behavior. When I read the documentation, I rather expect the loadaddr environment variable to be used whenever the address is omitted in a command invocation. Moreover, one may have to read/load several data pieces before booting, and the last loaded piece would not necessarily be the one containing the kernel to be booted. This should at least be documented.
Another approach would be to compel users to pass an address for all commands. Implicit behaviors are always dangerous, all the more if they are undocumented. But of course, this would break some existing configurations.
I tend to agree with all of the above; U-Boot's implicit/automatic/hidden/undocumented usage of variables that I didn't specify on the command-line, and setting of variables as a side-effect of executing commands, has always been quite astonishing (rather than the opposite of astonishing) to me:-(