Try to fix [issue](https://github.com/llvm/llvm-project/issues/68490

 )
and some extented problem. Root cause of current issue is that error
handling in instantiation of function parameter with default
initialization on sizeof or align expression. When instance an
out-of-line template member function, depth of `TemplateTypeParmDecl` in
default initialization doesn't change while depth of other template
parameter does and this will lead to some template parameter
uninstanced. Also, sometime it will leader to wrong instantiation when
it uses the template parameter of the template class.
Fix it by add template args of context. This will make
MultiLevelTemplateArgumentList::getNumLevels matching the depth of
template parameter. Testcase with some static_assert demonstrates the
template parameter has been instanced correctly.
But, the default initialization of lambda expression compiles failed
when only checking if the member function is out-of-line. We should
check the `PrimaryFunctionTemplateDecl` of the funtion if it's
out-of-line.

Co-authored-by: huqizhi <836744285@qq.com>

This allows making changes in parseArchExtension to make their way in
the command line as well, not only in target attributes.

This PR implements two missing intrinsics from F2023: ACOSD and ASIND.
The implementation breaks them down to the existing ACOS and ASIN
implementation.

llvm IR supports struct as function input, so RISCV tuple
type can just use struct of homogeneous scalable vector instead
of flatten them.

The legacy version print machine functions to a string stream, then
output the module and string in `doFinalization`. This patch break
`MIRPrintingPass` into two parts `PrintMIRPreparePass` and
`PrintMIRPass`. `PrintMIRPreparePass` output the original IR in yaml
string, `PrintMIRPass` just print the machine function, so we can avoid
the `doFinalization`.

attribute ((aligned)) should be 4 for -miamcu.

relate: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66818

This is yet another one-line patch to fix crashes on constraint
substitution.

```cpp
template <class, class> struct formatter;

template <class, class> struct basic_format_context {};

template <typename CharType>
concept has_format_function = format(basic_format_context<CharType, CharType>());

template <typename ValueType, typename CharType>
  requires has_format_function<CharType>
struct formatter<ValueType, CharType> {
  template <typename OutputIt>
  CharType format(basic_format_context<OutputIt, CharType>);
};
```

In this case, we would build up a `RecoveryExpr` for a call within a
constraint expression due to the absence of viable functions. The
heuristic algorithm attempted to find such a function inside of a
ClassTemplatePartialSpecialization, from which we started to substitute
its requires-expression, and it succeeded with a FunctionTemplate such
that

1) It has only one parameter...

This PR add support in wasm asm type checker to implement checker of
`table.grow`
Fixes: #79966.

EVL1 will be i32 and not get type legalized to i64 so we need to
manually promote it to XLenVT.

We should diagnose a const qualified member operator new but we fail to
do so and this leads to crash during debug info generation.

The fix is to diagnose this as ill-formed in the front-end.

Fixes: https://github.com/llvm/llvm-project/issues/79748

Without this change, the included test cases crash the compiler. I
believe this is fallout from the homogenous scalable struct work from a
while back; I think we just forgot to update this case.

Likely to fix https://github.com/llvm/llvm-project/issues/80463.

hlfir.declare introduce some boxes that can be later optimized away. The
OpenACC lowering is currently setting some attributes on FIR operations
to track declare variables. When the boxes are optimized away these
attributes are lost. This patch propagate OpenACC attributes from
box_addr op to the defining op of the folding result.

Add information in the operation's description about which operands must
be accessed with special getters for device_type support.

Currently, when doing e.g. "ninja check-openmp", the check-openmp target
only depends on the target "omp", which builds the library. Thus by
doing that, the separate import library "libomp.lib", which is generated
directly from a def file, never gets created, unless one does a separate
invocation first, that builds all targets.

To fix this, make the "omp" target depend on the target for the separate
import library, whenever that is created/used.

I wrote FeaturesStdExt instead of FeatureStdExt in a previous patch.

Fix a crash caused by incorrect assumptions
Reported here
https://github.com/llvm/llvm-project/pull/72644#discussion_r1469525524

* Use APFloat conversion function to avoid losing information by
converting to `double`. This would be the case with large types like
`f80` or `f128`.
* Check for potential information loss. This is intended for small
floating point types that may have values not present in larger ones
(e.g., f8m2e5fnuz and f16).
* Support folding vector constants.

Remove hardcoded string from test

rdar://120498671

The traditional promotion is known to generate wrong code.

Fixes #73805.

This is always PreservedAnalyses. Perhaps in the past there was the idea
to make some types of passes return more invalidation information, but
that hasn't happened.

Requires splitting out some analysis classes into a header so
PassManagerInternal.h can see the definition of PreservedAnalyses.

`__is_nothrow_convertible` has been used by libstdc++ previously as an
identifier.

sync_source_lists_from_cmake.py checks that every unittest in CMake
also exists in the GN build. 7b08b436 added SPRIVTests, but the GN
build doesn't include the SPIRV target. So add a dummy target for this
to placate the check.

See also VETests in the gn build, which has the same setup.

…ationFileExtensions

Deprecated since clang-tidy 17. Use the corresponding global options
instead.

Fixes #61947

---------

Co-authored-by: Carlos Gálvez <carlos.galvez@zenseact.com>

Broke sanitizer bots: https://lab.llvm.org/buildbot/#/builders/5/builds/40641

This reverts commit 67eee4a0.

This stemps from conversatin in:
https://github.com/llvm/llvm-project/pull/77457#discussion_r1457889792.
Right now Abbrev code for abbrev is combination of DIE TAG and other
attributes.
In the future it will be changed to be an index. Since DenseSet does not
preserve an order, added a sort based on abbrev code. Once change to
index is
made, it will print out abbrevs in the order they are stored.

This reverts commit bc06cd5c.

This caused the job to fail for PRs which still had an older version
of code-format-helper.py in their tree.

Reverted due to an internally discovered lld crash due to the underlying
StringMap changes, which turned out to be an existing lld bug that got
tickled by the StringMap changes. That's addressed in
dee8786f so let's have another go with
this change.

Original commit message:
    lldb was rehashing the string 3 times (once to determine which StringMap
    to use, once to query the StringMap, once to insert) on insertion (twice
    on successful lookup).

    This patch allows the lldb to benefit from hash improvements in LLVM
    (from djbHash to xxh3).

    Though further changes would be needed to cache this value to disk - we
    shouldn't rely on the StringMap::hash remaining the same in the
    future/this value should not be serialized to disk. If we want cache
    this value StringMap should take a hashing template parameter to allow
    for a fixed hash to be requested.

This reverts commit 5bc1adff.
Effectively reapplying the original 2e197602.

Reverted due to an internally discovered lld crash, which turned out to
be an existing lld bug that got tickled by this changes. That's
addressed in dee8786f so let's have
another go with this change.

Original commit message:
    Useful for lldb's const string pool, using the hash to determine which
    string map to lock and query/insert.

    Derived from https://reviews.llvm.org/D122974 by Luboš Luňák

This reverts commit f976719f.
Effectively reapplying 67c631d2.

Fixes an infinite loop.

Test copied from the non-rv64-legal-i32 test.