Forum TI-Planet.org

parrotgeek1 a écrit:out of curiosity, does it work on CX DVT? thanks for doing all of the tests

I'd like to. I'd like to be able to boot as many OSes as possible in ControlX, because it's great and because it lowers the possibilities of bricking.
And also because I'd like to tinker again with the official theme editor which was coming with one of the 3.0.0 OSes.

Unfortunately, nBoot apparently doesn't work with the CX Boot1 3.0.0.0 present on CX/Color EVT/DVT prototypes :
archives_voir.php?id=10192
I don't get the "Unlocked" string in the Firebird boot log, and the screen won't turn on.

Code: Tout sélectionner

Boot Loader Stage 1 (3.00.DEVBUILD)
Build: 2010/9/14, 16:30:29
Copyright (c) 2006-2010 Texas Instruments Incorporated
Using developer keys

Last boot progress: 65
Warning at PC=00007718: Bad write_byte: 0001c634 00

Available system memory: 33196
Checking for NAND: NAND Flash ID: Generic 1 GBit (0xA1)
SDRAM size: 32 MB
Wakeup Event: ON.
SDRAM memory test:   Pass
Clearing SDRAM...Done.
Clocks:  CPU = 132MHz   AHB = 66MHz   APB = 33MHz
Turning device off
Turning device off
Turning device off
Clearing SDRAM...Done.

Now for production calculators.
I think OS signed with the development keys (we've got some of them in versions 3.0, 3.2 and 3.6) can be installed by flashing and launching the CX Boot2 3.0.0.0 including the development keys :
archives_voir.php?id=10101
But you won't be able to boot them properly. Like every OS version older than 4.0.1, they need to be patched to prevent them from crashing in the nBoot context.
And the only Boot2 which is currently able to patch OSes before launching them is Boot2 4.0.3, which is including the production keys, and thus is going to erase such OSes.

.

I cannot for those ones, sorry.
They were shared by TI, under NDA, so they would immediately know it's coming from them.

But if it's to test the problem, it's exactly the same as with OSes 3.0.0 you can find here.

.

.

AFAICT, faking the product ID reduces the applicability of the tool, and is not even necessarily easier.
nLaunch / nLaunch CX / nLaunchy are known to kill the model / OS type checks, rather than force a model type. It shows in their source code anyway: NOPs zero out branches to error paths, 0xEA changes the condition of an instruction to "true" / "always".

critor does a large amount of testing on ControlX, which takes a lot of time; there are clear drawbacks to forking it and making the derivative a (much) less generic tool

.

I finally had time Tuesday to test many changes parrotgeek1 wanted me to test on my Nspire... and everything failed.

- I tried the first version he sent me and nlaunch didn't start at all : the OS was started normally (but maybe I mixed up the versions for not having numbered them )
- I tried the new version that can only launch 3.6 and 4.4, reading *very* carefully each line of the tutorial of nalunchy (especially the TNOC thing that I forgot for my first tests).
Result is : nlaunch starts, tries to install OS, OS starts (with the clock) and OS is rejected and the calc restarts... again and again.

One surprising thing is that when this happens, after I deleted the (fake) OS from "maintenance menu", every try of reinstalling a 3.6 TNOC-ed non CAS version on y non-CAS fails ! It reboots when pressing "OK" just after choosing the language. I had every time to reinstall full OS 3.6, then Ndless, then open "nsNandMgr" and reflash my Nand. I think I did it at least 10 times last Tuesday !!

.

.