Revert "[llvm][lit] Add option to run only the failed tests" (#171582)
Reverts llvm/llvm-project#158043
This was approved for earlier revisions but the tests were failing on
Windows. I pushed a speculative fix and that fixed the CI, which caused
auto-merge to merge the PR. But I'd like to have approval for the latest
revision. So reverting for now and resubmitting a new PR
[lldb][docs] Remove dead link in toctree
Fixes warning:
index.rst:144: WARNING: toctree contains reference to nonexisting document 'use/python'
There used to be such a page but the URL now redirects
to "script-driven-debugging" which is in the "Python Reference"
section.
[lldb][docs] Fix doc link syntax in bytecode formatter doc
This was flagged after the changes in 53674e2da437646591a6d20397cfd6c645dbf250.
Fixes warning:
formatterbytecode.rst:7: WARNING: 'any' reference target not found: ). To use custom data formatters, developers need to edit the global ``~/.lldbinit`
[compiler-rt][sanitizer_common] Make sanitizer_common tests work for other Apple Platforms (#150994)
This commit addresses a longstanding TODO comment, by doing the
following:
* Modifies the CMakeLists to add the new test configs
* Modifies the relevant lit file to add the required envs
* Fixes the FileCheck match in
`Darwin/symbolizer-function-offset-atos.cpp`
* XFAILs any appropriate tests
rdar://107758331
[OpenCL] Re-enable corresponding __opencl_* feature macros for OpenCL 2.0 core extensions (#171529)
Commit 423bdb2bf257e removed OpenCL 2.0 __opencl_* feature macros
(https://github.com/llvm/llvm-project/blob/fd73bdd662da/clang/lib/Headers/opencl-c-base.h#L60-L72)
from header and thus they are disabled by default for non-SPIR/SPIRV
targets now.
This PR re-enables them for OpenCL 2.0, and shall restore previous
behavior that these features macros were always defined for OpenCL 2.0
core extensions. After this PR, targets that support OpenCL 2.0 must
enable these feature macros via setSupportedOpenCLOpts API.
[lldb] Document the behaviour of IsValid for SBError (#170862)
This reverts commit d20d84fec5945fcc16aa6f63879e1458d4af9ea6.
Fixes #169788, but in a different way.
In which I changed an SBError use so that when the function succeeded,
IsValid on the SBError would be true.
This seemed to make sense but SBError acts differently to other SB
classes in this respect. For something like SBMemoryRegionInfo, if
IsValid() is false, you can't do anything with it.
However for SBError, IsValid() true only means there's some underlying
error object in there. If the SBError represents a success, there's no
need to put anything in there.
You can see this intent from a lot of its methods, many have handling
for IsValid() false.
[8 lines not shown]
[LoongArch] Add support for the ud macro instruction
This patch adds support for the `ud ui5` macro instruction. The `ui5`
operand must be inthe range `0-31`. The macro expands to:
`amswap.w $rd, $r1, $rj`
where `ui5` specifies the register number used for `$rd` in the expanded
instruction, and `$rd` is the same as `$rj`.
Relevant binutils patch:
https://sourceware.org/pipermail/binutils/2025-December/146042.html
[llvm][lit] Add option to run only the failed tests (#158043)
This patch adds a new `--filter-failed` option to `llvm-lit`, which when
set, will only run the tests that have previously failed.
[lldb][docs] Fix plaintext markers in Formatter bytecode doc (#171430)
RST uses double backtick for plain text. For single backtick it will try
to resolve it to a reference then fall back to plain text if it doesn't
resolve.
This leads to a lot of warnings like:
formatterbytecode.rst:7: WARNING: 'any' reference target not found:
~/.lldbinit
As seen on the publish-sphinx-docs bot.
[flang][docs] Do not recommend FLANG_ENABLE_WERROR=ON... (#171152)
...unless the reader is going to contribute to Flang.
Fixes #168911.
In which someone followed the getting started instructions but because
they were using GCC and it produced some warnings, those got upgraded to
errors and failed the build.
The GCC builds of anything in LLVM are always going to have some warning
of some kind, so while it would be nice to fix or silence them, let's
not inflict that on new developers.
This option defaults to `OFF` and is turned `ON` by a handful of bots,
so if the intent was to make developers more attentive to warnings, that
will still happen. Just a bit later, which that's better than putting
them off up front.
If they go on to contribute, they can switch to a Clang build and enable
the option to save themselves some post-commit hassle.
[BOLT][docs] Document commands to identify BOLT-ed binaries (#171458)
These have been useful to me in the past and I don't see any docs
referencing the `bolt_info` note either.
There is a test checking that it is emitted, so I assume it's something
we are ok with people looking for.
[clang][bytecode] Enable some placement-new tests (#171567)
They used to cause problems. Use the BYTECODE define to enable them at
least for the bytecode interpreter.
[LoongArch] Custom lowering for vector logical right shifts of integers (#171097)
After PR #169491, the DAG combiner can still recreate vector UDIV in an
illegal type even after type legalization, which is the root cause of
issue #170976.
The optimization introduced in PR #169491 is still desirable, so this
patch adds custom lowering for vector integer logical right shifts to
prevent the DAG from producing nodes with illegal types.
Fixes #170976
[compiler-rt][sanitizer] fix i386 build for Haiku (#171075)
r13 does not provide the trap err.
Co-authored-by: Jerome Duval <jerome.duval at gmail.com>
[AMDGPU] Removal of language sensitive option for _Float16 and half('e') handling (#170443)
Removing the 'e' handling for the amdgcn builtins as we decided to use
_Float16 for both HIP/C++ and OpenCL
[mlir][py] avoid crashing on None contexts in custom `get`s (#171140)
Following a series of refactorings, MLIR Python bindings would crash if
a
dialect object requiring a context defined using
mlir_attribute/type_subclass
was constructed outside of the `ir.Context` context manager. The type
caster
for `MlirContext` would try using `ir.Context.current` when the default
`None`
value was provided to the `get`, which would also just return `None`.
The
caster would then attempt to obtain the MLIR capsule for that `None`,
fail,
but access it anyway without checking, leading to a C++ assertion
failure or
segfault.
Guard against this case in nanobind adaptors. Also emit a warning to the
[13 lines not shown]