Dear Detlef Vollmann,
In message 49B97998.2080805@vollmann.ch you wrote:
+#if (CONFIG_SYS_HZ != 1000) +#error "CONFIG_SYS_HZ must be set to 1000 for this CPU" +#endif
no need please remove
As I already wrote in another message, I don't think so. All current boards that use U-Boot on that chip family need to use a value of 1000000 for CONFIG_SYS_HZ, and most of these config files are not in the public U-Boot tree.
We do not add dead code or provide other provisions for code that may exist in out-of-tree ports.
If these projects change to a current U-Boot version, they
And I can't see any reason for not having that check: it doesn't clutter the code, has no measurable effects on compile time, and absolutely no effects for the runtime.
Yes, it does clutter the code. It is dead code. We could as well add
#if 1 != 1 #error Armageddon! #endif
or similar to each and every source fine, just in case hell breaks lose.
+/* We leave some leeway for the master clock... */ +#if (AT91_MASTER_CLOCK < 85000000) || (AT91_MASTER_CLOCK > 115000000) +#error "AT91_MASTER_CLOCK must be set to 100000000 for this CPU" +#endif
why?
The old implementation simply assumes that AT91_MASTER_CLOCK has a value of 100000000, and I didn't change that. But the config file for afeb9260 sets it to a value of 89999598, and I can't change that as I don't know that board, but it's near enough to 100000000 to accept that. And I noticed that only because I had the check there in the first place, so it didn't compile with MAKEALL when I only checked for the exact value of 100000000.
That's a very obscure reasoning, me thinks. We don't understand the 89999598, so we accept 85000000...115000000 ? That's true vodoo programming.
Best regards,
Wolfgang Denk