Portable, Device-Oriented stdio C Runtime Library

The purpose of this library is to allow a device (such as a serial port) to work with stdio functions like printf and scanf. It defines the C FILE structure to have references to user-written functions to read and write a byte at a time. This has been a feature of AVR-LIBC, the standard stdio C runtime library for the 8-bit AVR microcontroller family. The library presented here started from AVR-LIBC soure code, cloned from https://github.com/stevenj/avr-libc3 in April 2020.

Functions Provided

The following standard C runtime functions are included. This is basically the normal set except for functions to open, close, or position a file.

formatted out	formatted in	simple out	simple in
fprintf	fscanf	fputc	fgetc
printf	scanf	fputs	fgets
snprintf	sscanf	fwrite	fread
sprintf	vfscanf	putc	getc
vfprintf	vscanf	putchar	getchar
vprintf	vsscanf	puts	gets
vsnprintf			ungetc
vsprintf

In addition, there are some functions and macros to create and initialize the C FILE structure. The user must provide read and/or write functions, which have the following signature:

void _fdev_put(void *, char);	// write a char
int  _fdev_get(void *);		// read a char or EOF

The void * is for user data, which can be anything but is the same for both functions in a given FILE. This signature is compatible with C++ method calls where the first argument is the this pointer.

To leverage the conversions from string to floating-point required by the scanf family, implementations of strtod, strtof, and atof have also been included.

Math Requirements

This library can be compiled into several different flavors depending on the level of math support needed. At the low end is support for integers up to size long, and no floating-point. At the other end is support for long long integers and both float and double math.

Support is set separately for integer and floating-point math. The options are defined in stdio_private.h:

// Values for INT_MATH_LEVEL
#define INT_MATH_MIN		0
#define INT_MATH_LONG_LONG	1

// Values for FP_MATH_LEVEL
#define FP_MATH_NONE    0
#define FP_MATH_FLT     1
#define FP_MATH_DBL     2
#define FP_MATH_FLT_DBL 3

It is expected that the values for INT_MATH_LEVEL and FP_MATH_LEVEL would be set on the compiler command line, but they can also be set by editting the file.

Note that FP_MATH_FLT_DBL only applies to the scanf family. scanf can potentially be asked to read in a float and/or a double, so the math for whichever will be used must be present.

The printf family all use variadic arguments, which means that float is promoted to double. This library supports a non-stardard hack in which only float support is present. In order to pass a float without promoting it to double, the bits are passed in a 32-bit integer. This can be done using a union of float and int32_t. The compiler may produce a warning that the type passed does not match the type specified by the format string.

Platforms

The code is written entirely in C (with an optimization for ARM of a single line of inline assmbly language). It has been shown to work on 32-bit and 64-bit Windows as well as ARM Cortex-M0+. It is intended that it can be compiled for any platform with a standard C compiler. Optimizations originally present for the 8-bit AVR have generally been removed.

This project was developed in Microchip Studio targeting ARM Cortex-M0+ microcontrollers. Fully compiled and ready-to-link libraries in every combination of math support are included for this target. The Microchip Studio project files are also included.