Wolfgang Denk wrote: > Dear Scott Wood, > > In message <20091019212409.ga31...@loki.buserror.net> you wrote: >> The override keyword is needed for make to take our version over the one >> specified on the command line, and remove it from the list of command line >> overrides that are passed to submakes. IMHO, the combination of "export" >> and "override" ought to do this automatically, but oh well. > > I have to admit that I don't see the problem. When I explicitly ask > for a specific ARCH setting on the command line (versus using some > setting I inherited from the envrionment, eventually even unaware of > the current value), then I do want to use that. So the current code > seems to do what I would expect from it. > > Maybe my expectations are whacky, though. > > I tend to NAK this one.
For the benefit of those who weren't on the IRC channel when we discussed this, I'll restate my objection. I think you're reading too much into the difference between setting ARCH with an environment variable and setting it on the command line. The number of users who are going to be confused by this (at least one, since I was) is greater than the number of users who would have done something useful by truly forcing ARCH=powerpc (zero without other changes, since all it gets you is a failed build). The way it is now, it looks as if ARCH=powerpc is accepted as an alias for ARCH=ppc. If this is not intended to be the case, and it's only intended to be accepted as an alias if you use one specific method of passing arguments (which again I recommend against, as it's confusing and not actually useful), then can we at least add a comment to the part of the makefile that does the rewriting clarifying this? Printing a warning if ARCH is still powerpc probably wouldn't hurt either, even in the absence of any rewriting. -Scott _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot