AVR Freaks

Take a look at the avr-libc user manual, the Program Space String Utilities and the FAQ.

Ah! Now I understand.

Um, ..... it must be one of those "seperated by a common language" issues.... I'm on the other side of the pond, so the analogy is a little lost. :lol:

Unfortunately you are quite correct. GCC, in it's current form, does not understand Harvard Architecture, or different address spaces, for any target.

Well, the one advantage I do see, is that you never have to worry about a stray data pointer pounding on your code causing your processor to go out to lunch. But, the disadvantages are having to swap around resources to the different memory addresses, and shoehorning C into Harvard Architecture. I have heard that the C Standard might possibly be updated to better handle this, and I've seen drafts to this effect, but who knows when that will be.

Then perhaps it would be best to post a bug report (technically enhancement request) at the GCC web site.

Please subscribe to the avr-gcc users list, read its archives, and
discuss this issue there. (Actually, it's already been discussed to
death, and it's best if you read Svein Seldal's article about the
memory-space implementation options, including the thread that
eventually followed.)

No, even though GCC can handle putting constant strings into R/O
memory, this will only work for von Neumann arch CPUs. For Harvard
CPUs, you cannot do this without breaking the C standards
compatibility at all (since you could not pass these pointers to the
standard functions from ), *and* GCC by now cannot handle
the different memory regions from Harvard CPUs at all (Svein considers
to add this feature, eventually, most likely in a way that will adhere
to the C standard extension Eric mentioned), so GCC later simply
wouldn't know the string is actually in ROM where it has to issue a
completely different instruction to fetch it.

What you're describing works pretty well for setups where the OS'es
MMU performs the distinction between R/O and R/W memory (as it's the
case for our typical universal computers these days), or for von
Neumann MCU architectures where the distinction is done by placing the
respective data into a completely different offset of the same (large)
address space, as it's apparently the case for the H8 but also for the
TI MSP430 MCU. In both situations, the compiler later doesn't have to
know the data are actual in ROM since it can still use the same
instructions to read them.

Oh I could think of a use: how about eating it? :) Kinda like a chocolate bunny at Easter... :wink:

A "sponge leg" :?:

How much would that two-bob watch cost now? 10 quid? :lol:

The ashtray bit though is pretty funny.

Left-ponder? :lol: