[Bug c/104248] New: armel: C11 atomics requires to be linked with libatomic.a explicitly

[Bug c/104248] New: armel: C11 atomics requires to be linked with libatomic.a explicitly

[mathieu.malaterre at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104248

            Bug ID: 104248
           Summary: armel: C11 atomics requires to be linked with
                    libatomic.a explicitly
           Product: gcc
           Version: 11.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mathieu.malaterre at gmail dot com
  Target Milestone: ---

Consider the following c11 code (lib + executable):

```
::::::::::::::
foo.c
::::::::::::::
_Atomic(long long) ll;

int foo(void)
{
        ++ll;
        return 42;
}
::::::::::::::
prog.c
::::::::::::::
int foo(void);

int main(int argc, char* argv[])
{
  return foo();
}
```

On Debian armel arch, most built system will fail to compile the above. The
error reported is:

```
/usr/bin/ld: libfoo.a(foo.c.o): in function `foo':
foo.c:(.text+0x40): undefined reference to `__atomic_fetch_add_8'
```

One need to carefully pass `atomic` library on the compilation line. So if you
are lucky this is just:

LDFLAGS=-latomic ...

some other time this more complex as in this above case where a static library
is build, thus one needs to:

```
cc -rdynamic prog.o -o prog libfoo.a -Wl,-Bstatic -latomic -Wl,-Bdynamic
```

---

I see that on some other arch (riscv), the spec file for gcc has been updated
to pass automatically (?) the proper flag: `--as-needed -latomic` [1]

Could someone from gcc/arm team please describe what is the reason for not
doing something equivalent for the spec file in armel case ? Thanks for your
time.

[1] https://github.com/riscv-collab/riscv-gcc/issues/12#issuecomment-276587351

[Bug c/104248] armel: C11 atomics requires to be linked with libatomic.a explicitly

[mathieu.malaterre at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104248

--- Comment #1 from Mathieu Malaterre <mathieu.malaterre at gmail dot com> ---
In case that help, cmake instructions for the test case is:

```
project(p C)
set(CMAKE_C_STANDARD 11)
add_library(foo STATIC foo.c)
add_executable(prog prog.c)
target_link_libraries(prog foo)
```

[Bug driver/104248] armel: C11 atomics requires to be linked with libatomic.a explicitly

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104248

--- Comment #2 from Andreas Schwab <schwab@linux-m68k.org> ---
This is a general problem with architectures not implementing all atomic
operations inline.

[Bug driver/104248] armel: C11 atomics requires to be linked with libatomic.a explicitly

[mathieu.malaterre at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104248

--- Comment #3 from Mathieu Malaterre <mathieu.malaterre at gmail dot com> ---
@Andreas the specific issue I am raising is the difference between the approach
of riscv vs armel. The riscv team is taking the responsability for putting the
missing `-latomic`, while armel expect the user to understand how c11 atomics
are implemented on a particular architecture and fix the link line at the build
system level (eg. cmake / meson ...).

In summary: what is the risk (if any) to update the gcc spec file on armel ?

[Bug driver/104248] armel: C11 atomics requires to be linked with libatomic.a explicitly

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104248

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |DUPLICATE
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Dup of bug 81358.

Note riscv atomics are messed up too(there is a bug report for that).

*** This bug has been marked as a duplicate of bug 81358 ***