| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
base_libdir may vary e.g. when using usrmerge feature
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
This was disabled for musl issue which is not related to altivec
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
| |
exporting them changes signatures of the tasks for all recipes
includeing native ones, so when meta-clang is not used it ends up
building all native recipes too. Defining these variables is good
enought for clang to see them in do_compile shell for target recipes
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
These variables are used during target package compiles in packages
which user llvm-config e.g. libclc
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
This ensures that multilib prefix is respected in dependencies and can
be handled when distro enables multilib
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
When RUNTIME=="llvm" the libcxx entry will be added in a separate if clause
right after this. We don't need it twice.
Signed-off-by: Esben Haabendal <esben.haabendal@huawei.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Workaround qemuppc illegal instruction trap on vxor instruction emulation
this instruction is emitted by clang with -mcpu=7400 on ppc32 when
compiling musl/strspn.c file
The crash shows up running qemu-ppc via gobject-introspection, so a long
unwieldy path but we know whats happening
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Enable __int128 for compiler-rt/ppc32
Drop redundant ppc patch which is fixed upstream already
Brings these fixes
* e879b2bf82ef [libcxxabi] [test] Depend on unwind only if available
* 569b773323a3 [libcxx][CI] Set Arm triples to match native clang build's default
* f7007c570a21 Lambdas are not necessarily locals. This resolves DR48250.
* eb84577cbc23 Revert "[release] Use the Bootstrapping build for building LLVM releases"
* d843bde69aab [clang][driver] Fix float128 diagnostics with glibc >= 2.32
* ba9ff030d302 [SimplifyCFG][PhaseOrdering] Defer lowering switch into an integer range comparison and branch until after at least the IPSCCP
* da3953fb0315 [NFC][PhaseOrdering] Add some tests from D119839
* 8b51e5ee0a2e [NewPM][Inliner] Make inlined calls to functions in same SCC as callee exponentially expensive
* 6a713120502a [llvm] [bindings/OCaml] Remove unused dep on ounit2
* 2de2a2bba998 [LICM][PhaseOrder] Don't speculate in LICM until after running loop rotate
* 7d1cd3380add [NFC][PhaseOrdering] Improve test coverage for D119975
* 1f4613675051 [NFC][PhaseOrdering] spurious-peeling.ll: also test -O1/-O2 results
* 36bf1a9e628d [NFC][PhaseOrdering] Precommit tests from D119965
* 67555104d23a [MachineSink] Disable if there are any irreducible cycles
* 1e4fd59253c6 [CodeGen] Add test for PR53990 (NFC)
* bfaed485756a [BOLT][CMAKE] Remove CMake 3.13.4 incompatible parameter
* 43b4544023b7 [LLD] [COFF] Order .debug_* sections at the end, to avoid leaving gaps if stripped
* e6d2aa9b0f13 [MC][WebAssembly] Fix crash when relocation addend underlows U32
* f6a4df3a0a3d [WebAssembly] Covert llvm/test/MC/WebAssembly/reloc-code.ll to asm. NFC
* 9d54fe21c67f [docs] Add RISC-V release notes for LLVM 14
* 1e3d5ccab6b6 Add cmake to source release tarballs
* 0205cc086e50 [OpenMP][FIX] Ensure custom state machine works
* 0826716786cd [Mips] support "sp" named register
* 09546e1b5103 [libc++][doc] Update the release notes.
* 8f9f84c6873e [docs] clang/docs/ReleaseNotes.rst: Add blank line
* aab33202d239 [docs] PowerPC release notes formatting/grammar fixes
* acf67b7a0dfc [docs] Tweak wording of note re: LTO on AIX
* ce1e90fc8dba [docs] Add PowerPC release notes for LLVM 14
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
This reverts commit a6cea51ab2b0923340012655ac0f13afba899e57.
|
|
|
|
|
|
|
| |
There are missing features in llvm-objdump especially for gnu-hash
sections for mips
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
do_populate_sysroot will execute the cross STRIP as part of the
processing. In non-clang builds this is present via POPULATESYSROOTDEPS
pulling in binutils, but in clang builds STRIP is set to llvm-strip which
obviously isn't part of binutils.
Set POPULATESYSROOTDEPS correctly to ensure that do_populate_sysroot has
the strip binary available.
Signed-off-by: Ross Burton <ross.burton@arm.com>
|
|
|
|
|
|
|
| |
These tools are better integrated with clang produced output consumption
pipeline, therefore use them when using clang as compiler
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: yangpeng1995 <1336781165@qq.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Commit a494bbb6ec44 ("clang: support android runtime") introduced an
unrelated change causing builds with RUNTIME="llvm" and COMPILER_RT not
set to get a dependency on virtual/${TARGET_PREFIX}compilerlibs.
This is clearly unrelated to addition of support for android runtime,
and causing unwanted depdencies to gcc-runtime for LLVM builds.
Signed-off-by: Esben Haabendal <esben.haabendal@huawei.com>
|
|
|
|
| |
Signed-off-by: Hsia-Jun(Randy) Li <randy.li@synaptics.com>
|
|
|
|
|
|
|
| |
libcxx does not build for armv5 due to atomic locks issue
and compile-rt cross build needs to be fixed for ppc32
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
We can not use nativesdk variants of libcxx and compiler-rt yet when
compiling nativesdk-clang because, it will need this compiler to build
them, so solve this catch-22, since we do not use the runtime built
during compiler builds, use libgcc/libstdc++ to pass cmake tests
during configure, this should be fine as it will be not needed for final
builds where nativesdk-clang will be used, it can still default to llvm
runtime on SDK host
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
| |
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
| |
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
| |
Use libgcc with clang-native to compile it
Fixes #465
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
klibc does not provide libssp functions e.g. __stack_chk_guard and
__stack_chk_guard which it expects from libc and compiler-rt code tend
to use these functions, so when using compiler-rt it ends up with
undefined symbols e.g.
/usr/lib/clang/13.0.0/lib/linux/libclang_rt.builtins-armhf.a(divmoddi4.c.o): in function `__divmoddi4':
divmoddi4.c:(.text+0x70): undefined reference to `__stack_chk_fail'
arm-yoe-linux-gnueabi-ld.bfd: divmoddi4.c:(.text+0x74): undefined reference to `__stack_chk_guard'
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
llvm unwind can work on rv32 and rv64 with latest clang so open up that
option
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
since toolchain-clang maybe available for native as well now, it means
we have to ensure we do not use it for native recipes _yet_
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that clang can pass right options automatically when RUNTIME =
"llvm", adjust the variables e.g. COMPILER_RT/UNWINDLIB/LIBCPLUSPLUS are
no more required to be defaulted
Disable enforcing gcc runtime for packages using gcc as its no longer
needed here but is taken care by RUNTIME variable as an aside it also
helps in building native and nativesdk llvm runtimes
Compute right dependencies based on selected RUNTIME along with
COMPILER_RT/UNWINDLIB/LIBCPLUSPLUS variables
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
This is advertised option from clang
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
| |
This ensures that it matches with CMake's understanding of this option
and when its removed then it silently removes -stdlibc=libc++ but leaves
the spurious '-' in the commandline which confuses the compiler as it
them expects input from stdin
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
There is no need for exporting CLANG_TIDY, since its not a standard
variable
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
Fixes Issue #415
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
oe-core's ccache.bbclass contains:
export CCACHE_COMPILERCHECK ?= "%compiler% -dumpspecs"
which sets a default value and allows others to easily override the
value of CCACHE_COMPILERCHECK if they wish to. Let's update
clang.bbclass to also set only the default value of CCACHE_COMPILERCHECK
too.
Signed-off-by: Mike Crowe <mac@mcrowe.com>
|
|
|
|
|
|
|
|
|
|
| |
The tune file for octeontx2 has been added into oe-core by commit
ad4f82742c6f ("tune-octeontx2.inc: Add tune for Marvell OCTEON TX2
core"). But the clang doesn't support this core yet. So remove the
-mcpu from the TUNE_CCARGS to fix the build failure for the octeontx2
core.
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
|
|
|
|
|
|
|
| |
Core uses a dumb utility called dwarfsrcfiles to grok source file names
and it assumes that .o and .a files are good old ELF files, that is not
true when using LTO in which case it becomes IR code, this tool starts
to puke, therefore disable using this tool
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
If thin-lto is used in DISTRO_FEATURES then enable thin-lto
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
lld yet does not work relaxation so accomodate for that so LTO can work
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
The -mmusl flag is GCC specific, Clang supports MUSL differently.
Prevents clang-11: error: unknown argument: '-mmusl'
Verified it does not affect the GCC toolchain.cmake build.
Signed-off-by: Leon Woestenberg <leon@sidebranch.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* issue:
the openembedded-core build procedure relies on changed default
dynamic loader path when the usrmerge distro feature turned on.
As we can find the GCC which is used as default toolchain uses
SYSTEMLIBS_DIR macro variable set to /usr/(lib|lib64) at build
time and hardcodes the binaries with that default loader path.
* solution:
change the default dynamic loader location with --dyld-prefix
clang option to conform the loader location (as GCC does) to
openembedded-core build procedure.
Signed-off-by: Oleksandr Ocheretnyi <oocheret@cisco.com>
|
|
|
|
|
|
|
| |
This helps extra tools like clang-tidy to find arch-specific macros,
headers in a cross compile environment.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
It reset variable CCACHE to disable ccache. But that doesn't effect that
CCACHE is set in an anonymous function from ccache.bbclass at last and
ccache is always not disabled with cmake.
Remove these useless piece of code.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
It fails to build compiler-rt when ccache is enabled:
| ccache: error: Failure running compiler check command: %compiler% -dumpspecs
Because clang doesn't recognize option '-dumpspecs' from default value
of CCACHE_COMPILERCHECK, override CCACHE_COMPILERCHECK for toolchain
clang in clang.bbclass.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Clang does not yet support big.LITTLE performance tunes, so use the LITTLE for tunes
see [1]
Fixes Issue #313
[1] https://reviews.llvm.org/D32076
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
| |
This option can turn Valid options into warnings and cause unintended
behaviour
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
| |
compile-rt is not fully ported for ARM big-endian therefore disable it
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|