[Bug c/108734] New: powerpc: False Detection of __atomic_*_8 Builtins

[Bug c/108734] New: powerpc: False Detection of __atomic_*_8 Builtins

[rmclure at linux dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

            Bug ID: 108734
           Summary: powerpc: False Detection of __atomic_*_8  Builtins
           Product: gcc
           Version: 11.3.0
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rmclure at linux dot ibm.com
                CC: bergner at gcc dot gnu.org
  Target Milestone: ---

Created attachment 54440
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54440&action=edit
False positive with __has_builtin

Using built-in specs.
COLLECT_GCC=powerpc-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc-cross/powerpc-linux-gnu/12/lto-wrapper
Target: powerpc-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu 12.2.0-3ubuntu1'
--with-bugurl=file:///usr/share/doc/gcc-12/README.Bugs
--enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++ --prefix=/usr
--with-gcc-major-version-only --program-suffix=-12 --enable-shared
--enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --libdir=/usr/lib --enable-nls --with-sysroot=/
--enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes
--with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-libitm
--disable-libquadmath --disable-libquadmath-support --enable-plugin
--with-system-zlib --enable-libphobos-checking=release
--without-target-system-zlib --with-libphobos-druntime-only=yes
--enable-secureplt --disable-softfloat --with-cpu=default32 --disable-softfloat
--enable-targets=powerpc-linux,powerpc64-linux --enable-multiarch
--disable-werror --with-long-double-128 --enable-multilib
--enable-checking=release --build=powerpc64le-linux-gnu
--host=powerpc64le-linux-gnu --target=powerpc-linux-gnu
--program-prefix=powerpc-linux-gnu- --includedir=/usr/powerpc-linux-gnu/include
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.2.0 (Ubuntu 12.2.0-3ubuntu1)


The attached program should compile on 32-bit systems, built with
`powerpc-linux-gnu-gcc fail.c -o fail`.

Instead, __has_builtin(__atomic_store_8) evaluates to true, and we get a linker
error.

```
/usr/lib/gcc-cross/powerpc-linux-gnu/11/../../../../powerpc-linux-gnu/bin/ld:
/tmp/ccJ9g2pV.o: in function `main':
fail.c:(.text+0x24): undefined reference to `__atomic_store_8'
collect2: error: ld returned 1 exit status
```

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|rejects-valid               |

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Link against libatomic.

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[rmclure at linux dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

--- Comment #2 from Rohan McLure <rmclure at linux dot ibm.com> ---
The issue here is less that I don't have access to the symbol __atomic_store_8,
it's that the preprocessor directive __has_builtin appears to false positive.

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=81358

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
These builtins are always available (via libatomic if neded) so testing
__has_builtin does not make sense.

If you are asking why libatomic is not automatically linked then this is a dup
of bug 81358.

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
What are you actually trying to do here?

If you're trying to provide your own definitions for the atomic built-ins then
obviously you need to know platform-specific details, including whether that
operation is already expanded without emitting a library call.

If you're trying to do something else, maybe testing the macro
__GCC_ATOMIC_LLONG_LOCK_FREE==2 can help.

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

David Edelsohn <dje at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dje at gcc dot gnu.org

--- Comment #5 from David Edelsohn <dje at gcc dot gnu.org> ---
Rohan,

I'm sorry that this is confusing, but the issue is cockpit error.

__has_builtin() does not mean that the builtin is inlined.  It only means that
GCC recognizes the builtin.  That is how __has_builtin() is documented.  In 32
bit mode, GCC emits an external reference for the builtin: 8 byte atomic
requires libatomic library, which is not linked by default (and shouldn't be).

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[rmclure at linux dot ibm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

Rohan McLure <rmclure at linux dot ibm.com> changed:

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

--- Comment #6 from Rohan McLure <rmclure at linux dot ibm.com> ---
David is right, I was assuming '__has_builtin' implied that the function would
be inlined in without needing to be obtained from libatomic. The translation
unit attached should then at least compile with no linker flags required, as if
the check fails it should define a stub implementation of __atomic_store_8
(pardon the obvious syntax error in that stub).

Sorry, my mistake. Yes that makes more sense.

[Bug c/108734] powerpc: False Detection of __atomic_*_8 Builtins

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108734

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to David Edelsohn from comment #5)
> __has_builtin() does not mean that the builtin is inlined.  It only means
> that GCC recognizes the builtin.  That is how __has_builtin() is documented.
> In 32 bit mode, GCC emits an external reference for the builtin: 8 byte
> atomic requires libatomic library, which is not linked by default (and
> shouldn't be).

And this is consistent with e.g. __has_builtin(__builtin_strlen). The fact that
GCC recognizes the token "__builtin_strlen" doesn't mean that you don't need an
extern definition of strlen for cases where the built-in isn't inlined. It's
just that for the atomic built-ins the name of the built-in is the same as the
name of the extern function that might be used.