jvonau at shaw.ca
Tue Jun 12 15:19:15 EDT 2012
On Tue, 2012-06-12 at 11:08 -0400, Kevin Gordon wrote:
> Disclaimer: Newbie Forth question ....:-)
> We are trying to create a consolidated unsecured update stick.
> This is not for use in the real world, but in our wild lab.
> We have learned and experimented a lot today, and everything pretty
> much humming along nicely, to a point:
> We have borrowed oodles of forth code from various sources and put
> together a nice little olpc.fth file which is sitting in the /boot
> directory on the USB stick.
> We then have an 885 .img file with the XO1 image, and an 885 .zd4 file
> with the XO1.5 image in the root of the stick.
> We hae coded it such that If we hit no game keys on the boot process,
> all machine(s) boot up normally - just like there was no stick
> inserted. This is the desired effect.
> If we hit the circle game key a bunch of stuff gets done. However,
> our attempt to select and execute fs-update when the machine is a 1.5,
> or copy-nand when the machine is a 1.0 is not working. It would
> appear that forth perhaps wants to parse the statements and fails
> because there is no command copy-nand in the 1.5 firmware or fs-update
> in the 1.0 firmware. If we use separate dedicated sticks without the
> offending command in the if statement, the machines flash beautifully.
> Basically, as it stands right now, we have working separate olpc.fth
> files for each of the machines, which isnt that inconvenient, and does
> enable us to auto-flash non-secure (security-disabled) machines
> quickly (with non-signed custom OOB builds) via game-key press rather
> than madly hitting the x key and then typing the correct command.
> So for those coming from a non-Forth background, we have hit a road
> block. Is there perhaps a way to store a 'possible' command into a
> variable then execute that 'variable' as a command, thereby perhaps
> bypassing any of the apparent syntax error checking? Unexpected
> end-of-line is the most common result from attempting to call within
> an if statement. Or, we get copy-nand? on the 1.5 or fs-update? on
> the 1.0 when the command exists in the source - whether it will
> actually get 'called' or not ,based on the variable containing the
> machine type..
> Or, is there an entirely more elegant way to accomplish this?
> Thanks for any hints. And, apologies in advance to those shaking
> their heads violently at our incompetence :-)
Have you seen?:
I'm in no way an expert, I do something something like that with
get-model?(needs updating for 1.75) and do-nand-update in our olpc.fth
script. Have a look in our git repo:
More information about the Devel