Update to Clang 19 and other changes by mkwcat · Pull Request #12 · DotKuribo/llvm-project

mkwcat · 2025-12-30T13:21:24Z

Welp, @JoshuaMKW said he doesn't mind how many commits there are so uhh here you go!!!!!

List of additions that I can remember:

Rebased on Clang 19.
Support for CodeWarrior PTMFs (pointer-to-member-functions). I'm not sure comparing them works but storing them and calling them works fine, and their size matches now. It does not call the __ptmf_scall, etc. helper functions though, it just loads and calls the pointer directly.
Use of CodeWarrior's __register_global_object instead of the regular atexit or __cxa_atexit. I think this may require -fno-use-cxa-atexit but I don't quite remember.
A new clang::external function attribute that takes a single string argument address, that:
- Creates an empty definition in the .external.$address section called __clang_external_$mangled, and;
- When applied to a declaration, creates a weak, alias definition for the function, aliased to the the empty external symbol.
- When applied to a definition, emits the function as well as a two-pointer struct in the section replace_array named __clang_external_replaced_$mangled, which contains pointers to the empty definition and the function definition in that order.
- As you could probably guess, yes, this attribute was designed for a hyper-specific use case within my own project.

Bugs in base Clang (I think) that I fixed:

Fixed a crash with using the gnu::naked attribute on C++ constructors.
Fixed the gnu::alias attribute not working on definitions defined inside a C++ class body.
Fixed a crash upon attempting to mangle a C++ lambda definition in an extern "C" context (this might be a Kuribo-specific issue, I'm not sure).

This enables "-rc" suffix in release branches. (cherry picked from commit 25efb74)

) This patch refactors the handling of reduction to eliminate layering violations. * Introduced `getReductionIntrinsicID` in LoopUtils.h for mapping recurrence kinds to llvm.vector.reduce.* intrinsic IDs. * Updated `VectorBuilder::createSimpleTargetReduction` to accept llvm.vector.reduce.* intrinsic directly. * New function `VPIntrinsic::getForIntrinsic` for mapping intrinsic ID to the same functional VP intrinsic ID. (cherry picked from commit 6d12b3f)

Since llvm#99276 has been landed, the dependency has become redundant. This reverts commit aa94a43. (llvmorg-19-init-17718-gaa94a43178e1) (cherry picked from commit 5bf0859)

``` SanitizerCommon-ubsan-sparc-Linux :: Linux/signal_line.cpp ``` currently `FAIL`s on Linux/sparc64 (32 and 64-bit) for `n == 2`. Instead of the expected `SIGSEGV`, the test dies with `SIGBUS`. `strace` reveals that this is due to a unaligned access: ``` --- SIGBUS {si_signo=SIGBUS, si_code=BUS_ADRALN, si_addr=0x1} --- ``` which is to be expected on a strict-alignment target like SPARC. Fixed by changing the invalid pointer to be better aligned. Tested on `sparc64-unknown-linux-gnu` and `x86_64-pc-linux-gnu`. (cherry picked from commit 94394ca)

llvm#101011) …parc64 ``` SanitizerCommon-Unit :: ./Sanitizer-sparc-Test/SanitizerCommon/InternalMmapWithOffset ``` `FAIL`s on 32-bit Linux/sparc64: ``` projects/compiler-rt/lib/sanitizer_common/tests/./Sanitizer-sparc-Test --gtest_filter=SanitizerCommon.InternalMmapWithOffset -- compiler-rt/lib/sanitizer_common/tests/sanitizer_libc_test.cpp:335: Failure Expected equality of these values: 'A' Which is: 'A' (65, 0x41) p[0] Which is: '\0' ``` It turns out the `pgoffset` arg to `mmap2` is passed incorrectly in this case, unlike the 64-bit test. The caller, `MapWritableFileToMemory`, passes an `u64` arg, while `mmap2` expects an `off_t`. This patch casts the arg accordingly. Tested on `sparc64-unknown-linux-gnu` and `x86_64-pc-linux-gnu`. (cherry picked from commit 1c25f2c)

…g SME attrs" (llvm#100991) (llvm#100996) Test `aarch64-sme-inline-streaming-attrs.c` caused some buildbot failures, because the test was missing a `REQUIRES: aarch64-registered target`. This was because we've demoted the error to a warning, which then resulted in a different error message, because Clang can't actually CodeGen the IR. (cherry picked from commit 389679d)

…lvm#92103) Also changes the behaviour of `__builtin_is_layout_compatible` None of the historic nor the current definition of layout-compatible classes mention anything about base classes (other than implicitly through being standard-layout) and are defined in terms of members, not direct members.

…gisters. (llvm#94628) On PowerPC there are 128 bit VSX registers. These registers are half overlapped with 64 bit floating point registers (FPR). The 64 bit half of the VXS register that does not overlap with the FPR does not overlap with any other register class. The FPR are the only subregisters of the VSX registers but they do not fully cover the 128 bit super register. This leads to incorrect lane masks being created. This patch adds phony registers for the other half of the VSX registers in order to fully cover them and to make sure that the lane masks are not the same for the VSX and the floating point register. (cherry picked from commit 53c37f3)

…script (llvm#101647) Current CSA logic does not expect `LazyCompoundValKind` as array index. This may happen if array is used as subscript to another, in case of bitcast to integer type. Catch such cases and return `UnknownVal`, since CSA cannot model array -> int casts. Closes llvm#94496 (cherry picked from commit d96569e)

In terms of bug catching capability, `_FORTIFY_SOURCE` does not perform as well as some dynamic instrumentation tools. When a sanitizer is used, generally `_FORTIFY_SOURCE` should be disabled since sanitizer runtime does not implement most `*_chk` functions. Using `_FORTIFY_SOURCE` will regress error checking (asan/hwasan/tsan) or cause false positives (msan). `*printf_chk` are the most pronounced `_chk` interceptors for uninstrumented DSOes (https://reviews.llvm.org/D40951). glibc 2.40 introduced `pass_object_info` style fortified source for some functions ([1]). `fprintf` will be mangled as `_ZL7fprintfP8_IO_FILEU17pass_object_size1PKcz`, which has no associated interceptor, leading to printf-fortify-5.c failure. Just disable the test. Fix llvm#100877 [1]: https://sourceware.org/pipermail/libc-alpha/2024-February/154531.html Pull Request: llvm#101566 (cherry picked from commit bbdccf4)

This reverts commit b933517. (cherry picked from commit dd7a4c3)

(cherry picked from commit 8f39502)

…1820) llvm#98489 resurrected an [old patch](https://reviews.llvm.org/D10833) that was adding new libclang functions. That PR got merged with old `LLVM_13` symbol versions for new functions. This patch fixes this oversight. (cherry picked from commit 2bae7ae)

``` UBSan-Standalone-sparc :: TestCases/Misc/Linux/diag-stacktrace.cpp ``` `FAIL`s on 32 and 64-bit Linux/sparc64 (and on Solaris/sparcv9, too: the test isn't Linux-specific at all). With `UBSAN_OPTIONS=fast_unwind_on_fatal=1`, the stack trace shows a duplicate innermost frame: ``` compiler-rt/test/ubsan/TestCases/Misc/Linux/diag-stacktrace.cpp:14:31: runtime error: execution reached the end of a value-returning function without returning a value #0 0x7003a708 in f() compiler-rt/test/ubsan/TestCases/Misc/Linux/diag-stacktrace.cpp:14:35 DotKuribo#1 0x7003a708 in f() compiler-rt/test/ubsan/TestCases/Misc/Linux/diag-stacktrace.cpp:14:35 DotKuribo#2 0x7003a714 in g() compiler-rt/test/ubsan/TestCases/Misc/Linux/diag-stacktrace.cpp:17:38 ``` which isn't seen with `fast_unwind_on_fatal=0`. This turns out to be another fallout from fixing `__builtin_return_address`/`__builtin_extract_return_addr` on SPARC. In `sanitizer_stacktrace_sparc.cpp` (`BufferedStackTrace::UnwindFast`) the `pc` arg is the return address, while `pc1` from the stack frame (`fr_savpc`) is the address of the `call` insn, leading to a double entry for the innermost frame in `trace_buffer[]`. This patch fixes this by moving the adjustment before all uses. Tested on `sparc64-unknown-linux-gnu` and `sparcv9-sun-solaris2.11` (with the `ubsan/TestCases/Misc/Linux` tests enabled). (cherry picked from commit 3368a32)

Crash seems to be caused by the check function not handling inline namespaces correctly for some instances. Changed how the Replacer is got from the MatchResult now which should alleviate any potential issues Fixes llvm#100406 (cherry picked from commit 0762db6)

This adds release binary builds for the 4 platforms currently supported by the free GitHub Action runners: * Linux x86_64 * Windows x86_64 * Mac x86_64 * Mac AArch64 The test stages for these are known to fail, but the creating and upoading of the release binaries should pass. (cherry picked from commit 247251a)

(cherry picked from commit 18dee70)

…ranch (llvm#100805) If we pull from the release tag, then if there is a bug in one of the actions on the release tag, then we can never do a build for that tag. Pulling from main will allows us to fix bugs in the actions we use to build the releases. (cherry picked from commit b0860b2)

Introduced in b0860b2. (cherry picked from commit d41f565)

…t workspace (llvm#100845) Otherwise, the checkout step will override them. (cherry picked from commit 41003ff)

…acos-14 (llvm#101774) python3 wasn't able to see modules installed by pip, so we need to use the setup-python action to ensure that the default pip and python3 both use the same prefix. See actions/runner-images#10385 (cherry picked from commit 59476c9)

…tion (llvm#101569) The old version in the llvm/actions repo stopped working after the version variables were moved out of llvm/CMakeLists.txt. Composite actions are more simple and don't require javascript, which is why I reimplemented it as a composite action. This will fix the failing abi checks on the release branch. (cherry picked from commit 14837af)

Fixes llvm#101859. If we have at least 2 ranges, we have to try to merge the last and first ones to handle the wrap range. (cherry picked from commit 4377656)

…lvm#101877) (cherry picked from commit 3c483b8)

…m#101828) For functions with private linkage, pick IMAGE_SYM_CLASS_STATIC rather than IMAGE_SYM_CLASS_EXTERNAL; GlobalValue::isInternalLinkage() only checks for InternalLinkage, while GlobalValue::isLocalLinkage() checks for both InternalLinkage and PrivateLinkage. This matches what the AArch64 target does, since commit 3406934. This activates a preexisting fix for the AArch64 target from 1e7f592, for the ARM target as well. When a relocation points at a symbol, one usually can convey an offset to the symbol by encoding it as an immediate in the instruction. However, for the ARM and AArch64 branch instructions, the immediate stored in the instruction is ignored by MS link.exe (and lld-link matches this aspect). (It would be simple to extend lld-link to support it - but such object files would be incompatible with MS link.exe.) This was worked around by 1e7f592 by emitting symbols into the object file symbol table, for temporary symbols that otherwise would have been omitted, if they have the class IMAGE_SYM_CLASS_STATIC, in order to avoid needing an offset in the relocated instruction. This change gives the symbols generated from functions with the IR level "private" linkage the right class, to activate that workaround. This fixes llvm#100101, fixing code generation for coroutines for Windows on ARM. After the change in f786881, coroutines generate a function with private linkage, and calls to this function were previously broken for this target. (cherry picked from commit 8dd065d)

… system. (llvm#101392) Fixes build errors on some SDKs. rdar://132607572 (cherry picked from commit 3a4c7cc)

llvm#100532) …sparc64 Two ioctl tests `FAIL` on Linux/sparc64 (both 32 and 64-bit): ``` SanitizerCommon-Unit :: ./Sanitizer-sparc-Test/SanitizerIoctl/KVM_GET_LAPIC SanitizerCommon-Unit :: ./Sanitizer-sparc-Test/SanitizerIoctl/KVM_GET_MP_STATE ``` like ``` compiler-rt/lib/sanitizer_common/tests/./Sanitizer-sparc-Test --gtest_filter=SanitizerIoctl.KVM_GET_LAPIC -- compiler-rt/lib/sanitizer_common/tests/sanitizer_ioctl_test.cpp:91: Failure Value of: res Actual: false Expected: true compiler-rt/lib/sanitizer_common/tests/sanitizer_ioctl_test.cpp:92: Failure Expected equality of these values: ioctl_desc::WRITE Which is: 2 desc.type Which is: 1 ``` The problem is that Linux/sparc64, like Linux/mips, uses a different layout for the `ioctl` `request` arg than most other Linux targets as can be seen in `sanitizer_platform_limits_posix.h` (`IOC_*`). Therefore, this patch makes the tests use the correct one. Tested on `sparc64-unknown-linux-gnu` and `x86_64-pc-linux-gnu`. (cherry picked from commit 9eefe06)

…m#100534) ``` SanitizerCommon-Unit :: ./Sanitizer-sparc-Test/SanitizerCommon/StartSubprocessTest ``` and every single test using the `llvm-symbolizer` `FAIL` on Linux/sparc64 in a very weird way: when using `StartSubprocess`, there's a call to `internal_fork`, but we never reach `internal_execve`. `internal_fork` is implemented using `syscall(SYS_clone)`. The calling convention of that syscall already varies considerably between targets, but as documented in `clone(2)`, SPARC again is widely different. Instead of trying to match `glibc` here, this patch just calls `__fork`. Tested on `sparc64-unknown-linux-gnu` and `x86_64-pc-linux-gnu`. (cherry picked from commit 1c53b90)

The CI has been a complete mess for the past week, and the only thing preventing it from being back is the Clang tidy checks. Disable them (as a total hack) to get CI back. (cherry picked from commit 78c6506)

) `llvm::ConstantFP::get(llvm::LLVMContext&, APFloat(float))` always returns a f32 constant. Fix llvm#107054.

…lvm#106949) Cache negative search result from getStoreMergeCandidates() so that mergeConsecutiveStores() does not iterate quadratically over a potentially long sequence of unmergeable stores. (cherry picked from commit 8f77d37)

…lvm#107432) After llvm#92205, LoongArch ISel selects `div.w` for `trunc i64 (sdiv i64 3202030857, (sext i32 X to i64)) to i32`. It is incorrect since `3202030857` is not a signed 32-bit constant. It will produce wrong result when `X == 2`: https://alive2.llvm.org/ce/z/pzfGZZ This patch adds additional `sexti32` checks to operands of `PatGprGpr_32`. Alive2 proof: https://alive2.llvm.org/ce/z/AkH5Mp Fix llvm#107414. (cherry picked from commit a111f91)

…107971) If all incoming values of `div.d` are sign-extended and all users only use the lower 32 bits, then convert them to W versions. Fixes: llvm#107946 (cherry picked from commit 0f47e3a)

In the test case, the BECount of the second loop uses %load, but we only have an LCSSA phi node for %add, so that is what gets invalidated. Use the forgetLcssaPhiWithNewPredecessor() API instead, which will invalidate the roots of the expression instead. Fixes llvm#109333. (cherry picked from commit 5bcc82d)

) This allows clang-scan-deps to work correctly when using cross compilers with names like <triple>-clang. (cherry picked from commit 87e1104)

This fixes building in this configuration after 87e1104. (cherry picked from commit aa34657)

This allows the clang driver to know which tool is meant to be executed, which allows the clang driver to load the right clang config files, and allows clang to find colocated sysroots. This makes sure that doing `clang-scan-deps -- <tool> ...` looks up things in the same way as if one just would execute `<tool> ...`, when `<tool>` isn't an absolute or relative path. (cherry picked from commit a26ec54)

Since a26ec54, we expand the executable name to an absolute path, if it isn't already one, if found in path. This broke a couple tests in some environments; when the clang workdir resides in a path under e.g. /opt. Tests that only use a tool name like "clang-cl" would get expanded to the absolute path in the build tree. The loop for finding the last "-o" like option for clang-cl command lines would inspect all arguments, including Args[0] which is the executable name itself. As an /opt path matches Arg.starts_with("/o"), this would get detected as an object file output name in cases where there was no other explicit output argument. Thus, this fixes those tests in workdirs under e.g. /opt. (cherry picked from commit cead904)

(cherry picked from commit 63267ca)

In ExpandFPLibCall, an assumption is made that all floating point libcalls that take integer arguments use unsigned integers. In the case of ldexp and frexp, this assumption is incorrect, leading to miscompilation and subsequent target-dependent incorrect operation. Indicate that ldexp and frexp utilize signed arguments in ExpandFPLibCall. Fixes llvm#108904 Signed-off-by: Timothy Pearson <tpearson@solidsilicon.com> (cherry picked from commit 90c1474)

…rbose_trap support (llvm#110161) We should have been checking against 1700, not 17000, which was a typo. (cherry picked from commit 1eba879)

Large scratch offset with one on highest bit selected as negative, negative offset has same binary representation in 16 bits as large unsigned offset. (cherry picked from commit e9d12a6)

…lvm#110256) Use i32 for offset instead of i16, this way it does not get interpreted as negative 16 bit offset. (cherry picked from commit 83fe851)

Until llvm#103056 lands or another more appropriate check can be found. This test fails on Ubuntu Focal where zdump is built with 32 bit time_t but passes on Ubuntu Jammy where zdump is built with 64 bit time_t. Marking it unsupported means Linaro can upgrade its bots to Ubuntu Jammy without getting an unexpected pass. (cherry picked from commit 6f6422f)

This fixes the test on Arm 32 bit Ubuntu Jammy where time_t is 32 bit. (cherry picked from commit cdd608b)

…lvm#90967) Multiple invocations of the pass could interfere with eachother, preventing some undefs being initialised. I found it very difficult to create a unit test for this due to it being dependent on particular allocations of a previous function. However, the bug can be observed here: https://godbolt.org/z/7xnMo41Gv with the creation of the illegal instruction `vnsrl.wi v9, v8, 0`

…ssing NoCD8 (llvm#109579) Promoted KMOV* was encoded with CD8 incorrectly, see https://godbolt.org/z/cax513hG1 (cherry picked from commit 0d334d8)

This was mistakely changed by llvm#109579, which doesn't match with other EVEX decoding. (cherry picked from commit 70529b2)

chapuni and others added 30 commits August 4, 2024 11:20

[Bazel] Use PACKAGE_VERSION for version string.

4bf04b2

This enables "-rc" suffix in release branches. (cherry picked from commit 25efb74)

Revert "[llvm][Bazel] Adapt to 4eb30cf"

0ee4442

Since llvm#99276 has been landed, the dependency has become redundant. This reverts commit aa94a43. (llvmorg-19-init-17718-gaa94a43178e1) (cherry picked from commit 5bf0859)

Revert "[AIX] Turn on #pragma mc_func check by default (llvm#101336)"

9be2c0e

This reverts commit b933517. (cherry picked from commit dd7a4c3)

[AMDGPU] Include <cstdint> in AMDGPUMCTargetDesc (llvm#101766)

51b4c66

(cherry picked from commit 8f39502)

workflows: Remove left over debugging step from release-binaries job

e28632e

(cherry picked from commit 18dee70)

workflow/release-binaries: Fix typo

bb42925

Introduced in b0860b2. (cherry picked from commit d41f565)

workflows/release-binaries: Fetch composite actions outside of defaul…

5657751

…t workspace (llvm#100845) Otherwise, the checkout step will override them. (cherry picked from commit 41003ff)

[Metadata] Try to merge the first and last ranges. (llvm#101860)

1069d16

Fixes llvm#101859. If we have at least 2 ranges, we have to try to merge the last and first ones to handle the wrap range. (cherry picked from commit 4377656)

InferAddressSpaces: Fix mishandling stores of pointers to themselves (l…

0648412

…lvm#101877) (cherry picked from commit 3c483b8)

Forward declare OSSpinLockLock on MacOS since it's not shipped on the…

2193e4f

… system. (llvm#101392) Fixes build errors on some SDKs. rdar://132607572 (cherry picked from commit 3a4c7cc)

ReleaseNotes: lld/ELF: mention CREL

2834433

Bump version to 19.1.0-rc2

d033ae1

ldionne and others added 30 commits September 26, 2024 16:00

[libc++] Disable the clang-tidy checks to get CI back (llvm#109989)

8a25c60

The CI has been a complete mess for the past week, and the only thing preventing it from being back is the Clang tidy checks. Disable them (as a total hack) to get CI back. (cherry picked from commit 78c6506)

[Clang][CodeGen] Fix type for atomic float incdec operators (llvm#107075

149bfdd

) `llvm::ConstantFP::get(llvm::LLVMContext&, APFloat(float))` always returns a f32 constant. Fix llvm#107054.

[LoongArch] Eliminate the redundant sign extension of division (llvm#…

9905852

…107971) If all incoming values of `div.d` are sign-extended and all users only use the lower 32 bits, then convert them to W versions. Fixes: llvm#107946 (cherry picked from commit 0f47e3a)

[clang-scan-deps] Infer the target from the executable name (llvm#108189

7d1f206

) This allows clang-scan-deps to work correctly when using cross compilers with names like <triple>-clang. (cherry picked from commit 87e1104)

[clang-scan-deps] Fix builds with BUILD_SHARED_LIBS=ON

a0fc8a2

This fixes building in this configuration after 87e1104. (cherry picked from commit aa34657)

[LoongArch] Fix the assertion for atomic store with 'ptr' type

b3734d9

(cherry picked from commit 63267ca)

[libc++] Fix AppleClang version number when checking for __builtin_ve…

53010fc

…rbose_trap support (llvm#110161) We should have been checking against 1700, not 17000, which was a typo. (cherry picked from commit 1eba879)

AMDGPU: Add test for 16 bit unsigned scratch offsets (llvm#110255)

03d1337

Large scratch offset with one on highest bit selected as negative, negative offset has same binary representation in 16 bits as large unsigned offset. (cherry picked from commit e9d12a6)

AMDGPU: Fix inst-selection of large scratch offsets with sgpr base (l…

962edd3

…lvm#110256) Use i32 for offset instead of i16, this way it does not get interpreted as negative 16 bit offset. (cherry picked from commit 83fe851)

Bump version to 19.1.1

fae11d4

[libcxx][test] Use smaller time range for 32 bit time_t (llvm#104762)

23eadbd

This fixes the test on Arm 32 bit Ubuntu Jammy where time_t is 32 bit. (cherry picked from commit cdd608b)

[X86][APX] Fix wrong encoding of promoted KMOV instructions due to mi…

edd018e

…ssing NoCD8 (llvm#109579) Promoted KMOV* was encoded with CD8 incorrectly, see https://godbolt.org/z/cax513hG1 (cherry picked from commit 0d334d8)

[X86][APX] Do not emit {evex} prefix for memory variant (llvm#109759)

d401987

This was mistakely changed by llvm#109579, which doesn't match with other EVEX decoding. (cherry picked from commit 70529b2)

Bump version to 19.1.2

d5498c3

Add support for CW PTMFs and global objects

4dc98cf

Merge branch 'main' into release-19-with-cw

a03545e

Update TargetCXXABI.def

79c9db8

Fix various build errors

01523ea

Fix can't mangle LambdaSig

a2a576c

Add "external" attribute

401c109

Fix crash with naked attr on C++ constructor

7f635c2

[Clang][CodeGen] Emit C++ constructor aliases in class body

b9b2fb6

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update to Clang 19 and other changes#12

Update to Clang 19 and other changes#12
mkwcat wants to merge 10000 commits intoDotKuribo:mainfrom
mkwcat:release-19-with-cw

mkwcat commented Dec 30, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Conversation

mkwcat commented Dec 30, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants