newRPL third demo released

[Claudio L.]

(08-09-2014 11:18 PM)jwm Wrote:  Cool stuff. Part of the reason I picked up a 50g was to play with stuff like this. Is the bootloader to flash it independent of the CPU ROM or is there a chance it could be bricked if you write to the wrong spot?

The bootloader is in the first 16 kbytes of the same flash. Some chips come with write protection on the first bank, I'm not sure there's any protection in this one (I'm not willing to try...).
But there's good news: the bootloader in the 50g doesn't need to be replaced and doesn't get in the way at all. And we don't need to write to the flash either, so it's very safe. Let me explain: newRPL will not write to flash like the 50g does. The entire 2 MB of flash will be used as ROM (no user banks). If you want flash, you have SD cards.
The firmware we have is installed to the calc by the same bootloader, using the standard rom update procedures. And the calc can be restored simply by flashing a normal rom again. We never override the bootloader, nor write to the flash directly. HP's bootloader installs newRPL and HP's bootloader removes it. Nothing custom there.

(08-09-2014 11:18 PM)jwm Wrote:  It seems that since you can drop to supervisor mode on the CPU you wouldn't need to flash over the existing ROM, just write a normal HP50g program that simply takes over the entire RAM including what the OS normally uses.

I wouldn't consider "normal" a program that takes over the entire RAM and overrides the OS completely. But that idea was already executed in 2007!
So far you are describing exactly what hpgcc3 does when you run a C program, except it doesn't ever remove the original OS (but it takes every bit of ram available to give the user up to 400 kbytes), and it is capable of restoring everything back perfectly and return control to the calc OS (so you don't even need a reset to get back).

(08-09-2014 11:18 PM)jwm Wrote:  It would require a hard reboot to get back into the standard OS, but it could be started like a normal program and won't endanger the factory ROM so would be less invasive to try out.

Basically you are suggesting to write an hpgcc3 application that doesn't return.
That approach has one major problem: there's no space in ROM. HPGCC3 already had to use one user flash bank, on top of every unused bit in the rom to fit the libraries.
If you can't put your OS in ROM, it needs to take RAM to execute, and that's very scarce already.
I think your idea is great, except we've been there already. If the calc OS is there, then there's no incentive to rewrite it from scratch, and the most logic course of action is to interact with it (and you end up with something like hpgcc3).
HPGCC3 installed in ROM alongside the calculator OS, and even though installation was 100% reliable, a lot of people got nervous, and kept using hpgcc 2.0 out of fear that it might do something to their calcs.
But that was back then. The WP34 project has shown that people are not afraid to flash their devices with custom firmware anymore, so here we are to give it a shot.

Claudio