Use Library Header in Fprime Submodule

[legounicycler]

Another question from the GT-2 team. We are running into the following issue:

The PolyTypes.cpp file uses macros with the form PRI___ (https://github.com/nasa/fprime/blob/devel/Fw/Types/PolyType.cpp#L617)

These are being defined by the standard library header inttypes.h included by polytypes here: https://github.com/nasa/fprime/blob/devel/Fw/Types/PolyType.cpp#L5

Arduino has an alternative definition for this file which uses double quotes instead of single quotes in its macro definitions: https://github.com/codebendercc/arduino-core-files/blob/master/v102/hardware/tools/avr/lib/avr/include/inttypes.h

The difference between single and double quotes is causing build errors.

The question is, how do we make sure components within the fprime submodule link against the Arduino header instead of the C standard library header?

Note: We are hoping to avoid modifying the submodule. We link our custom components to arduino by modifying each component's CMakeLists.txt file.

CC: @tumbleshack @SterlingPeet @teddylambert

[timcanham]

I'm a little confused by this - I looked at the Linux inttypes.h macros, and they also have double quotes.

[Joshua-Anderson]

Agreed - double quotes are the expectation. What compilation error are you seeing?

[legounicycler]

The following error shows up when running fprime-util build:

/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp: In member function 'void Fw::PolyType::toString(Fw::StringBase&, bool) const':
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:617:67: error: expected ')' before 'PRIu8'
(void) snprintf(valString, sizeof(valString), "%" PRIu8 " ", this->m_val.u8Val);
^~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:620:67: error: expected ')' before 'PRId8'
(void) snprintf(valString, sizeof(valString), "%" PRId8 " ", this->m_val.i8Val);
^~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:624:67: error: expected ')' before 'PRIu16'
(void) snprintf(valString, sizeof(valString), "%" PRIu16 " ", this->m_val.u16Val);
^~~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:627:67: error: expected ')' before 'PRId16'
(void) snprintf(valString, sizeof(valString), "%" PRId16 " ", this->m_val.i16Val);
^~~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:632:67: error: expected ')' before 'PRIu32'
(void) snprintf(valString, sizeof(valString), "%" PRIu32 " ", this->m_val.u32Val);
^~~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:635:67: error: expected ')' before 'PRId32'
(void) snprintf(valString, sizeof(valString), "%" PRId32 " ", this->m_val.i32Val);
^~~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:640:67: error: expected ')' before 'PRIu64'
(void) snprintf(valString, sizeof(valString), "%" PRIu64 " ", this->m_val.u64Val);
^~~~~~
/media/sf_SharedFolder/gt2-software/fprime/Fw/Types/PolyType.cpp:643:64: error: expected ')' before 'PRId64'
F-Prime/Fw/Types/CMakeFiles/Fw_Types.dir/build.make:230: recipe for target 'F-Prime/Fw/Types/CMakeFiles/Fw_Types.dir/PolyType.cpp.obj' failed
(void) snprintf(valString, sizeof(valString), "%" PRId64 " ", this->m_val.i64Val);
^~~~~~
make[3]: *** [F-Prime/Fw/Types/CMakeFiles/Fw_Types.dir/PolyType.cpp.obj] Error 1
CMakeFiles/Makefile2:1673: recipe for target 'F-Prime/Fw/Types/CMakeFiles/Fw_Types.dir/all' failed
CMakeFiles/Makefile2:11460: recipe for target 'Primary/Top/CMakeFiles/Primary.dir/rule' failed
make[2]: *** [F-Prime/Fw/Types/CMakeFiles/Fw_Types.dir/all] Error 2
Makefile:3251: recipe for target 'Primary' failed
make[1]: *** [Primary/Top/CMakeFiles/Primary.dir/rule] Error 2
make: *** [Primary] Error 2
[ERROR] CMake erred with return code 2

We determined the source of the error by replacing the macros with what should be their evaluated values (i.e. replace PRIu8 with 'hhu' and we receive the same compile time error. Replace PRIu8 with "hhu" and the compile time error disappears)

[timcanham]

It is almost acting like it doesn't see the macros for PRI* and is treating it as an identifier. Then it would expect the sprintf() arguments to obey a function invocation rather than the macro substitution.

[teddylambert]

Yep, it looks like seeing the macros is the issue. Traced this a little further, and looks like it has to do with the inttypes.h being included by Arduino. (For reference, I am on Arduino 1.8.10)

The Arduino version of inttypes.h is definitely the one being included, because removing/renaming this file throws an error on the #include inttypes.h line saying that it isn't found.

However, within this version of intypes.h, the PRI* macros we want are wrapped within 2 #define statements:

1. #if !defined(__cplusplus) || defined(__STDC_LIMIT_MACROS) -- This wraps all of the macros we want, so with it none of the macros get defined
2. #ifdef __avr_libc_does_not_implement_long_long_in_printf_or_scanf -- This specifically wraps around the 64 bit macros, so even if the first #ifdef is removed we still get errors on the PRI*64 macros specifically.

With both of these #ifdef statements removed, the original error disappears.

To me, it seems like having to edit this file included by the Arduino application itself is not the right way to go about things, and to instead focus on the conditions causing the #ifdef statements. Not sure about the best way to go about resolving this, but just wanted to provide an update on our debugging progress in case anyone has seen something like this before and has suggestions.

[Joshua-Anderson]

#if !defined(__cplusplus) || defined(__STDC_LIMIT_MACROS) -- This wraps all of the macros we want, so with it none of the macros get defined

It looks like __STDC_LIMIT_MACROS isn't being defined when it should be. Is there another compilation flag that causes the definition of this flag?

#ifdef __avr_libc_does_not_implement_long_long_in_printf_or_scanf -- This specifically wraps around the 64 bit macros, so even if the first #ifdef is removed we still get errors on the PRI*64 macros specifically.

That's ok... you can disable usage of 64 bit variables in F' by defining FW_HAS_F64=0

[SterlingPeet]

It is worth noting that this issue has disappeared when switching the F Prime version from devel to v2.0.1

[Joshua-Anderson]

The change introducing the use of PRI* macros for formatting polytype objects occurred after the 2.0 release.... this will be coming in the next F' release. It would be great to get to the bottom of what's causing this problem on Arduino beforehan!