Bug in DTT for temperatures below zero degress !!!
Hi there,
there's a terrible bug in the current U-Boot CVS concerning DDT:
For temperature below zero the function dtt_get_temp() (at least for lm75) returns wrong values!
...
DTT: 1 is 235 C LCD: Hitachi TX14D11 5,7" 320x240 TFT (235?C) WDT: disabled
This is because of the use of an "int" although the lm75 only returns two byte ("signed short").
This trivial fix is needed:
Index: u-boot/dtt/lm75.c
===================================================================
RCS file: /cvsroot/u-boot/u-boot/dtt/lm75.c,v
retrieving revision 1.2
dtt/lm75.c 15 Jun 2005 08:44:21 -0000 renamed from dtt/lm75.c 27 Jun 2003 21:32:38 -0000 1.2 +1 -1
@@ -174,7 +170,7 @@ int dtt_get_temp(int sensor) { - return (dtt_read(sensor, DTT_READ_TEMP) / 256); + return ( ((signed short) dtt_read(sensor, DTT_READ_TEMP)) / 256); } /* dtt_get_temp() */ #endif /* CONFIG_DTT_LM75 */
Then the output is correct:
...
DTT: 1 is -28 C LCD: Hitachi TX14D11 5,7" 320x240 TFT (-28?C) CCFL: disabled (too cold!) WDT: disabled
I am afraid this is an issues on all DTT in U-Boot. Some even only use "uchar".
Comments?
--
Steven
Quoted 6 lines · In message <[email protected]> you wrote:
there's a terrible bug in the current U-Boot CVS concerning DDT:
For temperature below zero the function
dtt_get_temp()(at least for lm75) returns wrong values!
Quoted 7 lines
...
DTT: 1 is 235 C LCD: Hitachi TX14D11 5,7" 320x240 TFT (235?C) WDT: disabledThis is because of the use of an "int" although the lm75 only returns two byte ("signed short").
Ummm...
Quoted 1 line
This trivial fix is needed:
Wait a moment...
dtt/lm75.c 15 Jun 2005 08:44:21 -0000 line 173
Quoted 8 lines
dtt/lm75.c 15 Jun 2005 08:44:21 -0000 renamed from dtt/lm75.c 27 Jun 2003 21:32:38 -0000 1.2 +1
-1@@ -174,7 +170,7 @@ int dtt_get_temp(int sensor) { - return (dtt_read(sensor, DTT_READ_TEMP) / 256); + return ( ((signed short) dtt_read(sensor, DTT_READ_TEMP)) / 256);
This is not the right way to fix it. I reject this patch.
dtt_read() is declared to return "int", so no such cast should be needed, i. e. you fix the symptoms instead of the cause.
I think the patch must be done in dtt_read(); IMO the line
return ((int)((short)data[1] + (((short)data[0]) << 8)));
is wrong; it should read
return ((short)((short)data[1] + (((short)data[0]) << 8)));
or, even better,
short rc_s; ... rc_s = (short)data[1] + (((short)data[0]) << 8); return (rc_s);
And the same problem is with dlen == 1 - it should read
return ((signed char)data[0]);
Do you agree with this assessment? And can you please confirm that my suggested patch fixes your problem, too?
Best regards,
Wolfgang Denk
Signature · 5 lines
Software Engineering: Embedded and Realtime Systems, Embedded Linux Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: [email protected] Of all possible committee reactions to any given agenda item, the reaction that will occur is the one which will liberate the greatest amount of hot air. -- Thomas L. Martin
Quoted 7 lines · Wolfgang Denk
dtt/lm75.c 15 Jun 2005 08:44:21 -0000 line 173
Quoted 2 lines
dtt/lm75.c 15 Jun 2005 08:44:21 -0000 renamed from dtt/lm75.c 27 Jun 2003 21:32:38 -0000 1.2 +1
-1- return (dtt_read(sensor, DTT_READ_TEMP) / 256); + return ( ((signed short) dtt_read(sensor, DTT_READ_TEMP)) / 256);This is not the right way to fix it. I reject this patch.
dtt_read()is declared to return "int", so no such cast should be needed, i. e. you fix the symptoms instead of the cause.
You're probably right. But at the moment I have no access to suitable hardware and I am off for a two weeks vacations today.
I promise to test this when I am back.
Thansk.
--
Steven