> I was looking in the sensors program's chips.c file at the various
> print_*() routines and found the following code:
>
> if (!sensors_get_feature(*name,SENSORS_ADM1025_ALARMS,&cur))
> alarms = cur + 0.5;
> else {
> printf("ERROR: Can't get alarm data!\n");
> alarms = 0;
> }
>
> repeated in various routines for various chips. Why is 0.5 added to
> the alarm mask value?
Values are retrieved from libsensors as floating point numbers, not integers.
When converting from floating point to interger, C will simply drop the
non-interger part. Thus, 12345677.99999 will be converted to 12345677 while we
are obviously expecting 12345678. Thus it is usually considered a good
practice to add 0.5 to any floating point value before converting it to an
integer.
Note that I am not sure it is needed in our specific case. The "double" type
should be able to handle 32-bit integer values without losses, and the drivers
don't export alarm values over 32-bit as far as I know. So I guess that the "+
0.5" operation is theoretical and we don't actually need it. Feel free to drop
it in your own code unless you notice it is actually needed.
--
Jean Delvare
http://khali.linux-fr.org/
