• State Accepted Answer
  • +1 person also asked this people also asked this
  • Replies 1 reply
  • Subscribers 10 subscribers
  • Views 232 views
  • Users 0 members are here
Options
Related

Different binary outputs with GCC on windows and linux

__mo
Offline

Hello the ARM Community,

Context:
We have an STM32 project we build using the arm-none-eabi 10.3-2021.07 toolchain.
The development/debug is mainly done on Windows.
Then we came to setup a continuous integration system (on Linux server/workers), using the linux build of the same toolchain version.
Both toolchains came from developer.arm.com/.../downloads.

Problem:
Using the same source, same cmake file, same linker script, the binary outputs are 99% identical, except for a few functions that are not compiled the same way.
The disassembling shows the differences are quite light.
Both binaries seem to run flawlessly (as far as I can tell), but from a validation point of view I find it worrisome as if differences can occur some in the future could be less innocuous, and I would prefer that both binaries be the same (apart from the build date, etc., though are reproducible build should even be feasible).

Questions:

1. Does anybody stumbled in the same kind of problem?
2. Is there a known way to avoid this difference? (except running the windows toolchain on linux or vice versa)

All suggestion or welcome.
Have a nice day!

Parents
  • +1 Offline

    Solved the issue.

    Compiled function differences came from... a wrongly synchronized source file. (yeah, no comment)

    But, there was also a difference in a .data section.
    A global array of struct, sized implicitely by its initializers was producing a different output depending of the compilation platform.

    Considering a structure like that:
    typedef struct {
    char * str;
    func1 pfunc1; //function pointers
    func2 pfunc2; //
    func3 pfunc3; //
    func4 pfunc4; //
    } mystruct_t;

    The global array

    mystruct_t g_list[] = {
      {"test", myfunc1, myfunc2, myfunc3, myfunc4},
      {NULL, NULL, NULL, NULL, NULL}
    };

    was not compiled/linked the same way (size mismatch) than

    mystruct_t g_list[2] = {
      {"test", myfunc1, myfunc2, myfunc3, myfunc4},
      {NULL, NULL, NULL, NULL, NULL}
    };

    This time addresses and dissassembled codes match when cross-compiled on windows and linux.

Reply
  • +1 Offline

    Solved the issue.

    Compiled function differences came from... a wrongly synchronized source file. (yeah, no comment)

    But, there was also a difference in a .data section.
    A global array of struct, sized implicitely by its initializers was producing a different output depending of the compilation platform.

    Considering a structure like that:
    typedef struct {
    char * str;
    func1 pfunc1; //function pointers
    func2 pfunc2; //
    func3 pfunc3; //
    func4 pfunc4; //
    } mystruct_t;

    The global array

    mystruct_t g_list[] = {
      {"test", myfunc1, myfunc2, myfunc3, myfunc4},
      {NULL, NULL, NULL, NULL, NULL}
    };

    was not compiled/linked the same way (size mismatch) than

    mystruct_t g_list[2] = {
      {"test", myfunc1, myfunc2, myfunc3, myfunc4},
      {NULL, NULL, NULL, NULL, NULL}
    };

    This time addresses and dissassembled codes match when cross-compiled on windows and linux.

Children
No data