PCD manual error?

akohlsmith · Joined: 25 Apr 2012 Posts: 26

I'm looking at the September 2013 PCD manual. It states that the standard C type 'int' is mapped to 'int8' for PCD.

I have "int ticks" in my code and the compiler is most definitely using 16 bits for the variable, both for storage allocation and for use. The target is PIC24F32KA304, and I am using #device ANSI as well.

Is this a manual error, a compiler option that the manual does not describe or something else?

newguy · Joined: 24 Jun 2004 Posts: 1903

It's a manual error. In PCD, int = signed int16.

akohlsmith · Joined: 25 Apr 2012 Posts: 26

Excellent, thank you for the verification.

SteveW · Joined: 27 Sep 2005 Posts: 25

I have gotten in trouble so many times with this issue with the CCS compilers that I now always clearly define all variables (i.e., "unsigned int16" or "signed int8", never simply "int").

newguy · Joined: 24 Jun 2004 Posts: 1903

Same here - now. I've only been involved with PCD for about a year but prior to that, pretty much exclusively PCH. Took code written & proven good from an 18 series PIC and migrated the same code to a dsPIC. Huge disaster. Boiled down to ints being UNsigned by default on PCH and changing to signed by default on PCD.

Now I explicitly say whether it's signed or not.

Ttelmah · Joined: 11 Mar 2010 Posts: 19236

Yes.

There are two different approaches that work.

1) Use the explicit names. Even better than the defaults, load the file 'stdint.h'.
2) Use your own explicit #type statement.

The latter is particularly useful if moving PIC16/18 code to a later chip, but if you define the sizes to be 'unusual', can lead to problems with the library code.

Best Wishes