Created using spr 1.3.4

[skip ci]

Probably it's not important, as it should not happen often,
but if all frames are internal we should pick top, not
the bottom frame.

At the moment, lto-discard-value-names isn't handled properly by libLTO.

Created using spr 1.3.4

[skip ci]

Prepare to move this into sanitizer_common.

This test fails on power pc and solaris, moving it to AArch64 folder
because the yaml is generated for an AArch64 MachO object file.

In newer SDKs, ld64 is called ld-classic. Look for it first, as ld in
those SDKs is still missing some functionality some tests depend on.

This enables running the tests from check-llvm-tools-lto with newer SDKs
on ARM64 macOS.

When -nostdinc and -nostdinc++ are both specified and the Baremetal
toolchain is used, an unused command line argument warning for
-nostdinc++ is produced. This doesn't seem particularly meaningful as
-nostdinc++ would have been claimed/used had the check in
AddClangCXXStdlibIncludeArgs not short-circuited. So, just claim all
arguments in this check.

I believe this is consistent with what was done for the GNU toolchain
(see 6fe7de90), so hopefully this is
appropriate here as well.

When CMake on Windows is told to generate the build into a directory
whose real path has a different drive letter (e.g. due to a symlink),
the "clang/test/Lexer/case-insensitive-include-absolute.c" test fails.
That happens because because `trySimplifyPath()` in `PPDirectives.cpp`
finds out there's more than a case difference between the `#include`
path (containing `%/t`) and the real path, which prevents the diagnostic
to fire.

I thought this is only an issue on Windows due to the fact that LIT does
not drag the path to the build directory through `os.path.realpath()`
like it does on other systems (see `abs_path_preserve_drive()` in
"llvm/utils/lit/lit/util.py"). However, even after only using
`os.path.abspath()` on a Unix system, build generated into a symlinked
directory tests correctly. I assume there must be something else at
play, but I don't have the time to dig deeper.

The fix is is fairly straightforward: use the real path in the
`#include` (with `%{/t:real}`), which removes the non-case difference
and unblocks the diagnostic.

Currently, the unit tests for the BenchmarkResult struct do not check if
the register initial values can be parsed back in. This patch adds a
matcher and tests that the register initial values can be parsed
correctly. This exercises code already contained within the benchmark to
yaml infrastructure.

This method was simply a wrapper around readOrError. All users within
the llvm-exegesis code base should have been processing an actual error
rather than using the wrapper. This patch removes the wrapper and
rewrites the users (just 1) to use the readOrError method.

This allows individual object files to override the common compile
commands in
their local CMakeLists' add_object_library call.

For example, the common compile commands contain -Wall and -Wextra.
Before
this patch, the per object COMPILE_OPTIONS were prepended to these, so
that
builds of individual object files could not individually disable
specific
diagnostics from those groups explicitly.

After this patch, the per-object file compile objects are appended to
the list
of compiler flags, enabling this use case.

ARGN is a bit of cmake magic; let's be explicit in the APPEND that we're
appending the compile options.

Link: #77007

process_vm_readv is generally considered dangerous from a syscall
perspective, and is frequently blanket banned in seccomp filters such as
those in Chromium and ChromiumOS. We can get the same behaviour during
the invalid PC address case with the raw SYS_rt_sigprocmask syscall.

Testing to ensure that process_vm_readv does not appear, I ran the
output of check-unwind on an ARM64 device under strace. Previously,
bad_unwind_info in particular would use process_vm_readv, but with this
commit, it now no longer uses it:

```
strace test/Output/bad_unwind_info.pass.cpp.dir/t.tmp.exe \
  |& grep process_vm_readv
```

The libunwind unittests were also tested on ARM64 ChromeOS (Gentoo
Linux) devices.

This patch folds `switch(zext/sext(X))` into `switch(X)`.
The original motivation of this patch is to optimize a pattern found in
cvc5. For example:
```
  %bf.load.i = load i16, ptr %d_kind.i, align 8
  %bf.clear.i = and i16 %bf.load.i, 1023
  %bf.cast.i = zext nneg i16 %bf.clear.i to i32
  switch i32 %bf.cast.i, label %if.else [
    i32 335, label %if.then
    i32 303, label %if.then
  ]

if.then:                                          ; preds = %entry, %entry
  %d_children.i.i = getelementptr inbounds %"class.cvc5::internal::expr::NodeValue", ptr %0, i64 0, i32 3
  %cmp.i.i.i.i.i = icmp eq i16 %bf.clear.i, 1023
  %cond.i.i.i.i.i = select i1 %cmp.i.i.i.i.i, i32 -1, i32 %bf.cast.i
```
`%cmp.i.i.i.i.i` always evaluates to false because `%bf.clear.i` can
only be 335 or 303.
Folding `switch i32 %bf.cast.i` to `switch i16 %bf.clear.i` will help
`CVP` to handle this case.
See also
https://github.com/llvm/llvm-project/pull/76928#issuecomment-1877055722.

Compile-time impact:
http://llvm-compile-time-tracker.com/compare.php?from=7954c57124b495fbdc73674d71f2e366e4afe522&to=502b13ed34e561d995ae1f724cf06d20008bd86f&stat=instructions:u

|stage1-O3|stage1-ReleaseThinLTO|stage1-ReleaseLTO-g|stage1-O0-g|stage2-O3|stage2-O0-g|stage2-clang|
|--|--|--|--|--|--|--|
|+0.03%|+0.06%|+0.07%|+0.00%|-0.02%|-0.03%|+0.02%|

Summary:
We use the OffloadBinary to contain bundled offloading objects used to
support many images / targets at the same time. The `__tgt_device_info`
struct used to contain a pointer to this underlying binary format, which
contains information about the triple and architecture. We used to parse
this in the runtime to do image verification.

Recent changes removed the need for this to be used internally, as we
just parse it out of the ELF directly. This patch sets the pointers up
so they point to the ELF without requiring any further parsing.

Accidentally didn't commit the review feedback before merging the PR

If there is no file name in the prologue of a line table, the verifier
will try to verify the file index, which will be set to 1 by default.
This will cause the DWARF verifier to throw an error even if there is no
error.

rdar://114476503
rdar://114343624

After 0504e932

Break out and refactor AArch64 & ARM intrinsic updating.  There's a fair amount of comonality, but let's avoid continually checking the same prefixes.

…te initialization" (#76272)""

With updates the libc++ tests.

This reverts commit 2205d233 and relands
86dc6e15 and
7ab16fb5.

Original commit was reverted because of failing libc++ tests, see #76232
for the discussion.

The errors in the tests are spurious in the first place (coming from
initialization of invalid classes), so update the tests to match new
behavior that does not show those errors.

The original patch was written by @ilya-biryukov 

To fix the CI two libc++ tests are temporary disabled for clang-18.

This patch improves the value solving of select at use if the condition
is an icmp and we know the result of comparison from
`LVI->getPredicateAt`.

Compile-time impact:
http://llvm-compile-time-tracker.com/compare.php?from=7e405eb722e40c79b7726201d0f76b5dab34ba0f&to=3c315b1ddcb0ad82554b33f08b9356679fae4bb7&stat=instructions:u

|stage1-O3|stage1-ReleaseThinLTO|stage1-ReleaseLTO-g|stage1-O0-g|stage2-O3|stage2-O0-g|stage2-clang|
|--|--|--|--|--|--|--|
|-0.01%|+0.01%|-0.00%|-0.00%|-0.08%|+0.02%|-0.01%|

We can't use the script as we need to keep the shuffle mask constant pool checks, but do more than just check that a second vperm isn't generated

Summary:
This formats something according to the style, and again attempts to fix
this failing on the BE PPC test. Sorry for the spam, these commits are
the only way I can check it because the failure isn't local.

A simple clause that is permitted on a few different constructs,
  'default' takes a required parameter of either 'none' or 'present'.
  This patch implements parsing for it.

This patch canonicalizes `switch(X << C)` to `switch(X)`. If the shift
may wrap, an and instruction will be created to mask out all of the
shifted bits.
Alive2: https://alive2.llvm.org/ce/z/wSsL5y

NOTE: We can relax the one-use constraint. But I don't see any benefit
in my benchmark.

Compile-time impact:
http://llvm-compile-time-tracker.com/compare.php?from=a776740d6296520b8bde156aa3f8d9ecb32cddd9&to=6dd783b9f90ae5f258102d732953567d7e317c02&stat=instructions%3Au

|stage1-O3|stage1-ReleaseThinLTO|stage1-ReleaseLTO-g|stage1-O0-g|stage2-O3|stage2-O0-g|stage2-clang|
|--|--|--|--|--|--|--|
|-0.00%|+0.01%|-0.02%|-0.01%|+0.02%|-0.00%|+0.01%|

The LLDB expression parser relies on using the external AST source
support in LLDB. This allows us to find a class at the root namespace
level, but it wouldn't allow us to find nested classes all of the time.
When LLDB finds a class via this mechanism, it would be able to complete
this class when needed, but during completion, we wouldn't populate
nested types within this class which would prevent us from finding
contained types when needed as clang would expect them to be present if
a class was completed. When we parse a type for a class, struct or
union, we make a forward declaration to the class which can be
completed. Now when the class is completed, we also add any contained
types to the class' declaration context which now allows these types to
be found. If we have a struct that contains a struct, we will add the
forward declaration of the contained structure which can be c ompleted
later. Having this forward de...

Summary:
This test fails because AMDGPU has a check for little-endianness before
returning the architecture. This test attempts to force the type to be
considered little-endian for the purpose of this test.

[RISCV] Don't call use_nodbg_operands for physical registers in RISCVOptWInstrs hasAllNBitUsers. (#77032)

The ADDIW in the new test case was incorrectly removed due to
incorrectly following the x10 register from the return value back to the
argument. This is due to use_nodbg_operands returning every instruction
that uses a physical register regardless of the data flow.

Relaxed a bit the opcode checks to make the test less sensitive to
changes resulting in opcode numbering.

Summary:
This did not need a separate struct because it is not past the
"architecture specific" ones in the identifier. This was accidentally
left when it was added somewhere else and caused unused variable
warnings.

Summary:
This patch fixes up the `makeTriple()` interface to emit append the
operating system information when it is readily avaialble from the ELF.
The main motivation for this is so the GPU architectures can be easily
identified correctly when given and ELF. E.g. we want
`amdgpu-amd-amdhsa` as the output and not `amdgpu--`.

This required adding support for the CUDA OS/ABI, which is easily found
to be `0x33` when using `readelf`.

GCC reports an instance of -Warray-bounds in block_offset.  Reimplement
block_offset in terms of memcpy_inline which was created to avoid this
diagnostic. See the linked issue for the full trace of diagnostic.

Fixes: https://github.com/llvm/llvm-project/issues/76877

Fixes: #77071
`SubstituteDeducedTypeTransform` will transform type and it will visit
uninstantiated `ExceptionSpecInfo`, which will cause odd behavior.

Since D154036 landed (2a65d038 on July 7, 2023), I've been seeing
many (40-50) libomptarget test failures with errors like the following
on some of our test systems:

```
/auto/software/gcc/x86_64/gcc-11.1.0/lib/gcc/x86_64-pc-linux-gnu/11.1.0/../../../../include/c++/11.1.0/cstdio:99:11: error: no member named 'fpos_t' in the global namespace
   99 |   using ::fpos_t;
      |         ~~^
```

This patch fixes that and doesn't break our other test sytems. I've
looked through the glibc history and at LLVM's libc stdio.h to give me
confidence that this solution should work on other systems. Of course,
there might be use cases I've overlooked, so feedback is appreciated.