discrepancy between spectranet wiki and sources

[ZXGuesser]

I noticed there is a disagreement between some of the text on the Memory page of the Spectranet wiki, and the true behaviour.

It reads

Pages 0x01 and 0x02 are used to store a copy of the Spectrum's frame buffer when an NMI or similar event is triggered. Pages 0x1C-0x1F are used by the configuration utility when rewriting the last erase sector of the flash rom, while updating the configuration.

The latter is true, but the frame buffer is stored in pages 0x1A and 0x1B.

(that page also still has notes suggesting that the port assignments may change!)

[ZXGuesser]

This is a simple documentation fix, however it set me thinking whether it would actually be worthwhile to move all these fixed assignments to pages 0x01-0x06 and allocate from page 0x07 onwards rather than having them at the end of memory.

My thought being that if someone wanted to increase the capacity of the SRAM, these fixed blocks would then be stuck in the middle of the address space, and pagealloc currently has a hard coded magic number governing how many pages are free from LOWEST_PAGE to where it will hit the frame buffer storage.

The downside to doing this is that it would be a breaking ABI change in both the main firmware and the config and utility modules, meaning that a full update would be necessary from previous revisions, for no immediate benefit.

Either way I believe things like these fixed pages numbers ought to be defined in sysdefs.inc, and probably have the assembler calculate the available pages from them too. I'm happy to knock up a pull request doing so for either a reshuffle or the status quo.

Edit: Disregard that suggestion, I realise now that the temporary copy of flash is used in the syslib\flashwrite code so a change would break every existing module installing tool etc. too.