error: conversion from 'double' to 'float' may change value [-Werror=float-conversion]

[superphosphate]

Description
I encountered a compilation error when integrating cJSON into my C++ project via CMake's FetchContent. The build fails at line 607 in cjson.c with the following error:

error: conversion from 'double' to 'float' may change value [-Werror=float-conversion]

Environment

• cJSON Version: v1.7.19
• Compiler: gcc version 15.1.0 (x86_64-win32-seh-rev0, Built by MinGW-Builds project)
• Build System: cmake version 4.0.2 (CMake; JetBrains IDE bundle; build 1)
• OS: Windows 11

Steps to Reproduce

1. Configure CMake to include cJSON via FetchContent:

   include(FetchContent)  
   FetchContent_Declare(  
     cjson  
     # actualy I downloaded form release and use `SOURSE_DIR` to include it
     GIT_REPOSITORY https://github.com/DaveGamble/cJSON.git  
     GIT_TAG v1.7.19
   )  
   FetchContent_MakeAvailable(cjson)  
   target_link_libraries(my_target cjson)  

2. Enable strict warnings (e.g., -Werror=float-conversion).
3. Build the project.

Error Log

"D:\software\CLion 2024.3.5\bin\cmake\win\x64\bin\cmake.exe" --build D:\projects\prj-terminal\cmake-build-debug --target prj_terminal -j 22
[1/3] Building C object _deps/cjson-build/CMakeFiles/cjson.dir/cJSON.c.obj
FAILED: _deps/cjson-build/CMakeFiles/cjson.dir/cJSON.c.obj 
D:\software\mingw64\bin\cc.exe -DCJSON_API_VISIBILITY -DCJSON_EXPORT_SYMBOLS -DENABLE_LOCALES -Dcjson_EXPORTS  -std=c89 -pedantic -Wall -Wextra -Werror -Wstrict-prototypes -Wwrite-strings -Wshadow -Winit-self -Wcast-align -Wformat=2 -Wmissing-prototypes -Wstrict-overflow=2 -Wcast-qual -Wundef -Wswitch-default -Wconversion -Wc++-compat -fstack-protector-strong -Wdouble-promotion -Wparentheses -Wformat-overflow -Wunused-macros -Wmissing-variable-declarations -Wswitch-enum -fvisibility=hidden -g -std=gnu11 -fdiagnostics-color=always -MD -MT _deps/cjson-build/CMakeFiles/cjson.dir/cJSON.c.obj -MF _deps\cjson-build\CMakeFiles\cjson.dir\cJSON.c.obj.d -o _deps/cjson-build/CMakeFiles/cjson.dir/cJSON.c.obj -c D:/projects/prj-terminal/third_party/cJSON-1.7.19/cJSON.c
In file included from D:/projects/prj-terminal/third_party/cJSON-1.7.19/cJSON.c:42:
D:/projects/prj-terminal/third_party/cJSON-1.7.19/cJSON.c: In function 'print_number':
D:/projects/prj-terminal/third_party/cJSON-1.7.19/cJSON.c:607:15: error: conversion from 'double' to 'float' may change value [-Werror=float-conversion]
  607 |     if (isnan(d) || isinf(d))
      |               ^
D:/projects/prj-terminal/third_party/cJSON-1.7.19/cJSON.c:607:27: error: conversion from 'double' to 'float' may change value [-Werror=float-conversion]
  607 |     if (isnan(d) || isinf(d))
      |                           ^
cc1.exe: all warnings being treated as errors
ninja: build stopped: subcommand failed.

Proposed Solution
The error occurs due to an implicit double-to-float conversion.
I asked ai, and it told me that to resolve this, consider:

1. Explicitly casting the result to float (if precision loss is acceptable).
2. Using -Wno-float-conversion for cJSON (not ideal for strict compliance).
3. Refactoring the calculation to avoid precision loss warnings.

Additional Context

• This issue arises when compiling with strict warning flags (e.g., -Werror).
• The problematic code is in the parse_hex4 function (line 607 in latest master).

I wander if there is a way to solve this without remove the flag Werror=float-conversion

---

I'm still learning Cpp and Cmake. So if you would be willing to help me, I would greatly appreciate it. Thank you for reading my issue. Let me know if you need more details!

[coveyjorjet]

Hi @superphosphate,

I am not a maintainer of this project but I propose to do this edit in the cjson.c line 607 in your version and try again.

#if defined(__GNUC__)
    if (__builtin_isnan(d) || __builtin_isinf(d))
#else
    if (isnan((double)d) || isinf((double)d))
#endif

I hope this fix your problem, if it did please message here so that one of us can send a pull request for this change.

[superphosphate]

Thank you for your wonderful suggestion! I tried your fix and it worked perfectly. The issue is now resolved in my environment. I'd appreciate it if you could submit a pull request with this change because my network condition is kind of bad and I'm sometimes unable to connect to the Github server.
Anyway, thank you again for you comment!

[GitMensch]

The issue with the approach above:

• this does not work for old C90 GCC compilers - the current version compiles with GCC 2.95.2!
• those builtins are possibly not available for other compilers defining the GCC macro