 |
 |
| View previous topic :: View next topic |
| Author |
Message |
stinky
Joined: 05 Mar 2012
Posts: 99
Location: Central Illinois
|
| Non-constant variable initialization |
 Posted: Mon Jul 06, 2015 9:48 am |
 |
|
V4.141
Just got bit by a mistake. Kind of wish the compiler had thrown an error on me when I did this. Am i asking too much?
The line in question:
| Code: | | static unsigned int8 bar = savedValue(READ); |
| Code: | #include "16LF1829.h"
#device adc = 10
#define OSCILLATOR 16000000
#use delay (internal = OSCILLATOR)
#fuses INTRC_IO, NOWDT, NOPUT, MCLR
#fuses NOBROWNOUT, NOLVP, NOCPD, NODEBUG,
#fuses NOPROTECT, NOCLKOUT, NOIESO
#fuses NOFCMEN, NOWRT, STVREN
#use RS232(baud = 9600, UART1, ERRORS)
#define READ 0
unsigned int8 savedValue(unsigned int8 temp) {
static unsigned int8 foo = 2;
if(temp)
foo = temp;
return foo;
}
unsigned int8 getValue(unsigned int8 temp) {
static unsigned int8 bar = savedValue(READ);
if(temp)
bar = temp;
return bar;
}
void main(void) {
while(TRUE) {
delay_ms(1500);
printf("getValue == %u\n", getValue(READ));
}
} |
|
|
 |
Ttelmah
Joined: 11 Mar 2010
Posts: 19577
|
|
| Posted: Mon Jul 06, 2015 1:39 pm |
|
|
What's wrong?...
In the initialisation, you call 'savedValue(0)'
foo is initialised to '2', so this returns 2 and saves this in bar.
You then call getValue(0). This returns bar, which is now 2.
It works as it is coded (this in 4.141)..... |
|
|
stinky
Joined: 05 Mar 2012
Posts: 99
Location: Central Illinois
|
|
| Posted: Mon Jul 06, 2015 2:14 pm |
|
|
Well the problem is that my short test program isn't accurately recreating the real problem I'm having.
The symptom was that "getvalue(READ);" was returning unexpected results the first time it was run. When I initialize variable bar with a constant value as opposed to the return value from a function.
Grr |
|
|
temtronic
Joined: 01 Jul 2010
Posts: 9258
Location: Greensville,Ontario
|
|
| Posted: Mon Jul 06, 2015 3:09 pm |
|
|
Variables in RAM can be anything on powerup unless YOU define or declare them.
It's always a very good idea to start with KNOWN data! CCS does have #zero_ram which puts 0 (zero) into ALL RAM locations.
While it does take a wee bit of time it does allow other peripherals( like an LCD module) to get 'organized' before the PIC program accesses them.
Jay |
|
|
RF_Developer
Joined: 07 Feb 2011
Posts: 839
|
| Re: Non-constant variable initialization |
| Posted: Tue Jul 07, 2015 2:19 am |
|
|
| Quote: | Am i asking too much?
|
In my view, yes, but I can understand where you're coming from.
There's all sorts of issues of the scope and lifetime of static variables local to routines. What will happen will depend on when and how such variables get initialised.
There's also the somewhat philosphical question of whether a routine can be called and run before the program has started, or at least before user code has started? How can you run code before the code is running? In other words is it valid to call a routine during non-user initialisation? This is what I think is the step too far. I'd expect the compiler to throw an error saying it expected an expression that evaluated to a constant in that initialisation.
As it is I'd expect the statics (which in implementation terms are global variables with local scope) to be initialised in the order they occur in the source, and after any normal (i.e. non-static) global variables. As such, I'd expect your example code to "work", but not necessarily if they had been defined the other way round, and not necessarily in the same way from cold and warm starts. The reason for this would be that foo may not have been intialised by the time getValue ran. In general, in embedded code, it can also be distinctly dodgy to run code before hardware initialisation. For example, do we even know the UART will be up and running at that time? If so, would a printf work? Due to oscillator start-up arrangements, many PICs aren't even running at the correct clock speed until well into user initialisation code. Shouldn't matter for a simple variable of course, but...
This sort of thing is one reason I'm not a great fan of statics in routines. I use them from time to time, but not very often. Its all a bit clearer and better defined in object orientated languages, such as C++, where objects can be constructed at a known time and in a programmer controlled way, such as through static member functions. This example code smells a little C++-y to me, and that's may have introduced slightly inappropriate expectations of initialisation behaviour.
Regardless of how it works, and whether it works, and whether it should work, its easy to work round the issue, and that is how I tend to code anyway. That's to say don't try to be fancy, just make the variable global and initialise it expicitly during user code start-up. |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ 
Search
Register
Profile 
Log in to check your private messages 
Log in
