This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.
Messages - Goplat
Pages: 1 ... 15 16 [17] 18 19 20
241
« on: December 23, 2010, 12:13:58 pm »
We should note that if OS 2.x detects a touchpad when it boots, it does some communication with it over the I2C interface, and henceforth it ignores the keypad bits that should correspond to the arrow keys, and uses the I2C to read actual x/y values instead. Maybe that initial communication switches it from "arrow keys mode" to "x/y mode" or something.
If you boot up with a clickpad, run a program like BlockDude, and switch to the touchpad, do the "arrow keys" respond then?
242
« on: December 22, 2010, 09:13:27 pm »
ExtendeD: With a fresh compile, the "swi 80003b" call in crt0.o's cache_keypad_type is getting linked into ndless_resources.tns, and the installer's swi handler doesn't handle that.
apcalc: Does this mean your touchpad is working again? Good to see the problem wasn't permanent.
243
« on: December 22, 2010, 06:15:36 pm »
Figured it out - it's a bug in nspire_emu  The thumb instruction "neg" ought to set the carry and overflow flags, but it doesn't. This messes up gcc's way of implementing the ! operator in thumb: neg r1, r0 @ r1 = -r0, carry set if r0 == 0
adc r0, r1 @ r0 is now 1 if it was 0 before, 0 otherwise
Here is a patch that hopefully fixes it: Index: cpu.c
===================================================================
--- cpu.c (revision 58)
+++ cpu.c (working copy)
@@ -861,7 +861,7 @@
case 0x6: /* SBC */ res = *dst = add(*dst, ~src, arm.cpsr_c, true); break;
case 0x7: /* ROR */ res = *dst = shift(3, *dst, src & 0xFF, true); break;
case 0x8: /* TST */ res = *dst & src; break;
- case 0x9: /* NEG */ res = *dst = -src; break;
+ case 0x9: /* NEG */ res = *dst = add(0, ~src, 1, true); break;
case 0xA: /* CMP */ res = add(*dst, ~src, 1, true); break;
case 0xB: /* CMN */ res = add(*dst, src, 0, true); break;
case 0xC: /* ORR */ res = *dst |= src; break;
244
« on: December 22, 2010, 12:25:16 pm »
I'm trying to add support for OS 1.7 back, but I am buming into a strange problem.
I have defined this function in libndls.h:
static inline BOOL is_cas_model(void) {
return *(volatile unsigned *)0x900A002C == 0x04000001;
} This requires that many of the calculator Product ID bits be zero; it will only work on 1 out of 8388608 calculators. Should be return (*(volatile unsigned *)0x900A002C & 0x04000000) != 0;
245
« on: December 20, 2010, 06:04:53 pm »
That's isalpha actually. TI's implementation of these functions is basically
int isalnum(int c) { return isdigit(c) || isalpha(c); }
int isalpha(int c) { return islower(c) || isupper(c); }
so it's very easy to confuse them if you're looking at a disassembly without symbols.
It's common practice for the standard C library to be implemented with one function per file, so the linker can omit any unused functions. That's probably what happened - whatever code used to use isalnum back in 1.7, got changed in 2.0.
246
« on: December 17, 2010, 08:19:04 pm »
Here's a possibillity: Since a negative signed number is 65536 less than its interpreted as unsigned, you can just do an unsigned multiplication and if any side is negative, subtract the other side from the high word of the result:
a b product
pos pos ab = ab
pos neg a(b-65536) = ab - 65536a
neg pos (a-65536)b = ab - 65536b
neg neg (a-65536)(b-65536) = ab - 65536(a+b)
247
« on: December 04, 2010, 06:18:34 pm »
Error at PC=1028C97C: r0 = ff000000, nav user = 11f1ee14
Yeah, it's not good to be resetting while leaving the usblink hack active, especially to a different OS version where addresses will have changed. I should fix it to clean everything up on reset. Anyway, shouldn't be a problem with the method I described.
248
« on: December 04, 2010, 03:32:54 pm »
And here comes a major problem: nspire_emu doesn't support file transfers for OS 2.x, so debugging is not easy.
Someone once suggested to transfer the file from OS 1.7 and then upgrade the OS. But I get a crash on OS startup, and anyway I wild get mad after a few debug sessions.
Here's another possible workaround, that just takes a reboot instead of a full OS reinstall: create a patched boot2 that doesn't try to load the OS but goes straight into download mode. (e.g. patch the code at offset 0x1244 from 31 06 00 eb to 15 00 00 ea). Send documents using the patched boot2 (may need to set the target folder to the empty string, it doesn't seem to like sending to Examples). Save flash, quit, and restart using the original boot2. Edit: what do you mean by getting a crash on OS startup? If you're referring to the problems in OS 2.1 CAS, here's a fix: it expects GPIO 2 to be high (indicating battery door closed, not open) Index: apb.c
===================================================================
--- apb.c (revision 56)
+++ apb.c (revision 57)
@@ -15,7 +15,7 @@
case 0x10: case 0x14: case 0x1C: case 0x20: case 0x24:
return 0;
case 0x18:
- return i == 1 ? 1 : 0;
+ return 0x0104 >> (i * 8) & 0xFF;
}
}
return bad_read_word(addr);
249
« on: November 29, 2010, 12:50:29 pm »
I used Logisim in a class a year or two ago. I made this little puzzle with it in my spare time: try to light up all the LEDs. (it's too easy, I know)
250
« on: November 28, 2010, 04:53:21 pm »
I still intend to release the source code after cleaning it up some, but I've been occupied by other things recently.
251
« on: November 24, 2010, 10:15:03 pm »
I don't think it has anything to do with USB versions. As I understand it, success or failure of Ndless installation is basically pseudorandom and depends on how much free memory you have. So plugging/unplugging USB cables, creating/deleting documents - anything that causes the OS to allocate or free memory, might change it from working to not working or vice versa.
252
« on: November 23, 2010, 06:35:29 pm »
Excellent!
One bug I noticed: After running ndless_tests.tns, the cursor moves too fast. sleep() isn't setting the timer counter reset value back properly (reading 900D0000 gets its current value, not its reset value), so it ends up speeding up the timer interrupt by a random amount.
253
« on: November 23, 2010, 02:10:51 pm »
ExtendeD: Instead of bothering with adapters and such, would it be possible to just connect the serial ports of two Nspires together in a cross-over configuration, and run a program on one to dump what it reads to a file? Seems that would be easier, to someone who has a spare.
Edit: noticed critor's comment about soldering; is a serial cable something that has to be soldered to the connector? obviously this would be a bad idea if so :p
254
« on: November 12, 2010, 12:00:24 pm »
My computer Is having some serious issues.
It shuts down after 30-60 min of being on.
There is no warning or anything. It just powers down as is if the plug was pulled. Could it be overheating? Maybe a fan is broken or something.
255
« on: November 08, 2010, 02:43:27 pm »
except that the checker for it is the boot1, which is ROM (not flash ROM either, completely unchangeable)
It actually is Flash ROM, and there is code in the diagnostics to re-flash the BOOT1 from a file read from an SD card. But there's no reason you'd ever want to flash it because 1) there would be a major chance of completely bricking your calculator, and 2) unless you have whatever hardware TI uses to connect an SD card reader, only way you could flash it is if you already have the ability to execute your own code, which would mean there's no point.
Pages: 1 ... 15 16 [17] 18 19 20
|
Show Posts