Newlib or Picolibc libraries for LLVM may be compiled or installed from
pre-built sources independently of LLVM itself.
This means that always indicating that TOOLCHAIN_HAS_NEWLIB=OFF and
TOOLCHAIN_HAS_PICOLIBC=OFF are wrong. But it could be just as wrong to
always indicate suport for newlib or picolibc.
Some pre-built LLVM toolchains are provided with default picolibc
support, such as LLVM for Arm embedded, but can also be used with newlib
be installing newlib add-on package.
Unfortunately it's not possible to query LLVM regarding newlib or
picolibc support.
Developers have the option of `-DTOOLCHAIN_HAS_<NEWLIB|PICOLIBC>=ON`,
but this is not widely known and cumbersome to do for each build.
An indication of newlib or picolibc support is the presence of library
specific headers, so to improve current situation we check for library
specific headers, and if those are present we assume support for the
library.
This commit improves the current support for LLVM in Zephyr when
cross-compiling, especially for users of LLVM for Arm embedded.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Moving specs argument to compiler and linker properties so that the
compiler and linker in use can decide how the flags are mapped / handled
for the compiler and linker in use.
This avoids specifying `--specs=spec.picolibc` for clang which prints a
warning about an unused argument.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Zephyr is a bare metal build where standard libs are disabled.
This means that c and runtime libraries must manually be linked in.
This has generally been handled by using CMake's link libraries handling
but the issue with that is both de-duplication but also library link
order.
Standard libraries must be linked at last location to ensure symbols
are always available, however this is not optimal with
target_link_libraries() because this would ultimately require every
library to know the c library to link with, which is not desired.
Therefore, setup standard C and runtime library linking in linker
CMake files for toolchains where this is required.
This commit expands the principle introduced with toolchain abstraction,
see PR#24851.
This means that a toolchain implementation may specify standard C,
runtime, C++, etc libraries, as well as their link order.
Because a property approach is used, then Zephyr modules, such as the
Picolibc module can adjust such properties.
An optional `zephyr_linker_finalize()` macro is called at the end of
Zephyr's CMakeList process and can be used by the toolchain
implementation to define the final linker invocation.
This aligns the linker handling flow to the principle introduced in
PR#24851 and improves the flexibility and robustness of Zephyr build
system.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
The macro `toolchain_cc_nostdinc` were made obsolete with PR#24851 and
are no longer invoked.
Remove the macro implementations.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Clang uses floating-point instructions by default, even if -mfpu is not
defined. Disable using FPU when CONFIG_FPU=n.
Using floating-point instructions when FPU is not enabled generates
Usage Fault.
Signed-off-by: Dawid Niedzwiecki <dawidn@google.com>
Ensure --target and -mcpu/-mfpu/-mtune are set appropriately when building
with clang targeting arm64/aarch64.
Signed-off-by: Jonathon Penix <jpenix@quicinc.com>
Ensure --target and -march/-mabi/-mcmodel are set appropriately when
building with clang targeting RISC-V.
Signed-off-by: Jonathon Penix <jpenix@quicinc.com>
Too many times, code is pushed that includes floats that really
becomes doubles and C implicit promotion rules will want to make
floats into doubles very easily. As zephyr primarily targets
low-end process that may not have a double precision floating
point unit, enable this flag globally for now.
Signed-off-by: Ryan McClelland <ryanmcclelland@meta.com>
File gcc-m-fpu.cmake is responsible for determining what should be
passed to -mfpu option. Fortunately GCC and Clang options are compatible
so we can use the file for clang.
The list of supported -mfpu options can be found at
https://github.com/llvm/llvm-project in
llvm/include/llvm/TargetParser/ARMTargetParser.def file.
Signed-off-by: Patryk Duda <pdk@semihalf.com>
Fixes: #63771
The 'compiler/*/generic.cmake' is intended to set a C preprocessor which
can be used for devicetree preprocessing.
No C++ compiler is needed as host tool and should therefore not be set
in this file.
Remove the code to avoid setting a not required C++ compiler.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Add a a new source coverage for native builds
and new kconfig choice of COVERAGE mode to select which:
* COVERAGE_NATIVE_GCOV: what we had until now with native builds
* COVERAGE_NATIVE_SOURCE: a new LLVM source coverage mode
* COVERAGE_GCOV: the old COVERAGE_GCOV (embedded gcov data generation).
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Clang/LLVM is natively a cross-compiler, so one set of applications can
compile code to all supported targets. The default target can be changed
using '--target' option.
CMake supports this type of compilers. To change compiling target, one
should set CMAKE_C_COMPILER_TARGET accorgindly.
The '--target' option has impact on the path to clang-rt library
returned by compiler when run with '--print-libgcc-file-name' option.
Without specifying target, Clang will return path to runtime library of
the host target (e.g. x86_64-pc-linux-gnu).
Signed-off-by: Patryk Duda <pdk@semihalf.com>
Clang provides runtime library `libclang_rt.builtins.<arch>.a` but
`libgcc.a` is also supported.
The compiler can provide full path to the selected library using the
`--print-libgcc-file-name` option, for example:
```
/usr/lib64/clang/17/lib/baremetal/libclang_rt.builtins-armv7m.a
```
If `--rtlib=libgcc` option is provided, clang will also provide
appropriate path.
This patch replaces hardcoded `libgcc` with library name extracted from
path, so we are compatible with both libraries.
Signed-off-by: Patryk Duda <pdk@semihalf.com>
This property specifies the flag used to pass the linker script filename
through the compiler front end tot he linker.
For clang, we use the general purpose linker-pass through -Wl flag with -T:
-Wl,-T as clang doesn't support -T.
For gcc, we use -T directly as this keeps the picolibc specs file from
inserting the picolibc linker script as well.
If the compiler doesn't specify a value, we continue to use -Wl,-T as
before.
Signed-off-by: Keith Packard <keithp@keithp.com>
These flags were added to avoid warnings when main was declared to return
void. Now that main returns int, those warnings will flag errors.
Signed-off-by: Keith Packard <keithp@keithp.com>
Add a Kconfig option to set the compiler save-temps flag and set the GCC
implementation. This is very useful for troubleshooting macro expansion
issues, having an option allows a user to set it like any other config
option.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Some distros may provide config files for clang to change its
default behavior. We need to override that, or else developers
may be using different defaults and we will have confusing
bug reports in the future.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Unlike gcc, clang splits out the flag controlling warnings about the return
type of `main' from other warnings related to that function. Add the extra
-Wno-main-return-type flag to mask these warnings when building Zephyr
without -ffreestanding, as when using picolibc.
Signed-off-by: Keith Packard <keithp@keithp.com>
This avoids a cryptic DTC failure when compiling trying to compile
native_posix with a missing gcc or clang.
REQUIRED is available since CMake 3.18
Example with clang, cryptic error without this commit:
```
ZEPHYR_TOOLCHAIN_VARIANT=llvm west build \
-p -b native_posix samples/hello_world/
-- Found toolchain: host (clang/ld) <= this is wrong
-- Found Dtc: /usr/bin/dtc (found suitable version "1.6.1", minimum ...
-- Found BOARD.dts: zephyr/boards/posix/native_posix/native_posix.dts
CMake Error at /zephyr/cmake/modules/dts.cmake:191 (message):
command failed with return code: No such file or directory
Call Stack (most recent call first):
zephyr/cmake/modules/zephyr_default.cmake:113 (include)
zephyr/share/zephyr-package/cmake/ZephyrConfig.cmake:66 (include)
zephyr/share/zephyr-package/cmake/ZephyrConfig.cmake:92 (include_boil...
CMakeLists.txt:5 (find_package)
```
Well hidden behind the scenes, dts.cmake fails above because it invokes
`CMAKE_C_COMPILER-NOTFOUND`
With this commit:
```
ZEPHYR_TOOLCHAIN_VARIANT=llvm west build \
-p -b native_posix samples/hello_world/
-- Found toolchain: host (clang/ld) <= this is still wrong
CMake Error at zephyr/cmake/compiler/clang/generic.cmake:7 (find_program):
Could not find CMAKE_C_COMPILER using the following names: clang
```
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Clang does not have printf return value optimizations like GCC, so there's
no flag to turn them off when building against a non-standard printf
implementation (e.g., picolibc without float support).
Signed-off-by: Keith Packard <keithp@keithp.com>
Add a compiler option to not merge globals. gen_kobject_list.py
is not capable of distinguish addresses of merged objects. The script
itself does not look wrong. The dward specification says that the
attribute DW_AT_location with opcode DW_OP_addr encodes a machine
address and whose size is the size of an address on the target machine
but for merged objects the address is longer, not clear why.
Disable global merge when userspace is enabled to avoid this problem.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
When building with clang, the unittests were giving us an error:
```
error: undefined symbol: llvm_gcda_start_file
```
This seems to be from linking in `gcov` regardless of the toolchain.
It appears that clang doesn't need any special library for coverage.
With this change the following now produce identical coverage reports:
```
$ ZEPHYR_TOOLCHAIN_VARIANT=zephyr ./scripts/twister -p unit_testing \
--coverage -i -T tests/unit/intmath/
$ ZEPHYR_TOOLCHAIN_VARIANT=host ./scripts/twister -p unit_testing \
--coverage -i -T tests/unit/intmath/
$ ZEPHYR_TOOLCHAIN_VARIANT=llvm ./scripts/twister -p unit_testing \
--coverage -i --coverage-tool lcov \
--gcov-tool $(pwd)/scripts/utils/llvm-gcov.sh \
-T tests/unit/intmath/
```
Signed-off-by: Yuval Peress <peress@google.com>
This adds a choice of three different libc API buffer overflow detection
modes:
* None
* Compile-time
* Compile-time and Run-time
These correspond with the clang/gcc _FORTIFY_SOURCE modes (0/1/2).
_FORTIFY_SOURCE depends on compiler optimizations and require libc support
which the minimal C library doesn't include, so _FORTIFY_SOURCE is disabled
by default in those cases. Native tooling might also enable
_FORTIFY_SOURCE, so don't enable it by default in that case either.
Signed-off-by: Keith Packard <keithp@keithp.com>
Clang 15 added a new warning type `-Wdeprecated-non-prototype` that
warns about the functions without prototypes, which have been
deprecated since the C89 and will not work in the upcoming C2x.
This commit disables the warning because Zephyr deliberately makes use
of the functions without prototypes to allow the use of a "generic"
function pointer (notoriously in the cbprintf implementation) and
Zephyr will not move to the C2x in the foreseeable future.
Signed-off-by: Stephanos Ioannidis <root@stephanos.io>
Compilation warnings appears for C++ files, that following
options are not valid:
-ffrestanding,
-Wno-format-zero-length
-Wno-main
-fgnu89-inline
-std-gnu99
Added checks to filter out unsupported flags.
Signed-off-by: Jaroslaw Stelter <Jaroslaw.Stelter@intel.com>
Not sure why this is needed for this branch, but it pretty clearly is --
there are hundreds of set-but-unused variables in the codebase.
Signed-off-by: Keith Packard <keithp@keithp.com>
For each compiler, also set a CMAKE_GCOV var referencing the appropriate
gcov tool.
Tested with gcc and host-gcc on the ChromeOS codebase.
Signed-off-by: Jeremy Bettis <jbettis@chromium.org>
The icx compiler of oneApi will invoke clang-cl on windows,
this is not supported in zephyr now, so change to use
clang directly.
And the clang from oneApi can't recognize those cross
compiling target variables of cmake, such as
CMAKE_C_COMPILER_TARGET, so used other cmake functions
to pass them to clang.
Signed-off-by: Chen Peng1 <peng1.chen@intel.com>
When compiler results are piped through a non-terminal (e.g. ninja)
the compiler disables colour diagnostics. Using `-fdiagnostics-color`
forces the compiler to enable colour output. This flag is set for
clang and gcc when `ZEPHYR_BUILD_COLOUR_DIAGNOSTIC` environment
variable is set when a clean build is started.
Signed-off-by: Arvin Farahmand <arvinf@ip-logix.com>
The system include directory might include spaces on windows, so quote
it.
Partially part of fix for #32111
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
We do not need libgcc always, some environments do not have libgcc and
do not require it, so keep it more flexible.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Without the -Wno-typedef-redefinition option, clang complains if a
typedef gets redefined in gnu99 mode (since this is officially a C11
feature).
While new versions of GCC do not seem to issue this warning in gnu99
mode anymore. So some existing code with typedef redefined which works
well with GCC will issue this warning.
Similar to what was done in 2354f055ec.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Generate errors for implicit conversions that may alter a value, that
covers conversions implicit conversions to narrower types, different
signal, different essential types (float -> int), ... From GCC
documentation
(https://gcc.gnu.org/onlinedocs/gcc-9.3.0/gcc/Warning-Options.html)
"""
Warn for implicit conversions that may alter a value. This includes
conversions between real and integer, like abs (x) when x is double;
conversions between signed and unsigned, like unsigned ui = -1; and
conversions to smaller types, like sqrtf (M_PI). Do not warn for
explicit casts like abs ((int) x) and ui = (unsigned) -1, or if the
value is not changed by the conversion like in abs (2.0). Warnings
about conversions between signed and unsigned integers can be disabled
by using -Wno-sign-conversion.
For C++, also warn for confusing overload resolution for user-defined
conversions; and conversions that never use a type conversion
operator: conversions to void, the same type, a base class or a
reference to them. Warnings about conversions between signed and
unsigned integers are disabled by default in C++ unless
-Wsign-conversion is explicitly enabled.
"""
This partially cover rules: 10.4, 10.5, 10.6, 10.7 and 10.8
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Add implicit-fallthrough flag when CONFIG_MISRA_SANE is set.
It will require __fallthrough macro or a proper comment tells the
compiler that this happens intentionally.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Add a Kconfig option to enable compiler flags that help to enforce
some code guideline rules.
Note: As the current code base is not in compliance with the adopted
code guideline, some rules will generate warnings during the
build. This is intended to help to spot violations.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
-Wold-style-definition is not a supported option for C++ builds. To
prevent it being passed:
* the list of compiler flags to be excluded from C++ builds is moved
to be toolchain-specific;
* -Wold-style-definition is added to that list for gcc and clang;
* -Wold-style-definition is moved from zephyr_compiler_options to
zephyr_cc_option so the option checking code is executed for it.
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
