Floating point printf formatting error using ARM GCC newlib-nano on STM32H753

[Gpeti]

I am working on STM32H753 (with eval board EVAL 2).

I am modifying an existing firmware built without libc and without startup files.

I need to convert floating point variables to strings, so I want to use as little as libc as possible. I choosed newlib-nano for memory size.

I compiled with `-mfpu=fpv5-d16 -mfloat-abi=hard -specs=nano.specs ...` and linked with `-mfpu=fpv5-d16 -mfloat-abi=hard -specs=nosys.specs -specs=nano.specs -u _printf_float -nostartfiles ...`

I also added the definition of symbols required by the libc and reserved a memory area for the heap.

 

I tried to write a string from a double variable with the following code (it's code for debug only, I know it may overflow the char array depending on float value):

char my_string[64];
double my_double = 1.3;
sprintf(my_string, "%.3f", my_double);

But the content of the string (checked by dumping the memory on debugger) is `1300` and not `1.300`

(i tried with different values and format specifiers)

Am I missing something ?

[Pavel A.]

What is the version of the toolchain? 

[Gpeti]

Sorry, it is arm-none-eabi-gcc 10.2.0

[Saket_Om]

Hello @Gpeti 

Are you using STM32CubeIDE?

[Gpeti]

No. I build with make and gcc and I load and debug with Lauterbach trace32