This is a follow-up to
https://github.com/swiftlang/llvm-project/pull/12317#discussion_r2850297229
Per that discussion, given that deserializers *can* fail given a corrupt
PDB, it's preferable to handle the error instead of crashing.
This specific change is limited to "easy" changes (read: I have high
confidence in their correctness). The ideal end state is funneling all
errors to a few central places in `SymbolFileNativePDB`.
lldb had three preprocessor defines for logging,
LLDB_LOG - formatv style argument
LLDB_LOGF - printf style argument
LLDB_LOGV - formatv style argument, only when verbose enabled
If you weren't looking at Log.h and the definition of these three, and
wanted to log something with formatv, it was easy to use LLDB_LOGV by
accident. We just had a situation where an important log statement
wasn't logging and it turned out to be this. This is fragile if you
aren't looking at the header directly, so I'd like to make this more
explicit. My proposal:
LLDB_LOG - formatv style argument
LLDB_LOG_VERBOSE - formatv style argument, only when verbose enabled
LLDB_LOGF - printf style argument
LLDB_LOGF_VERBOSE - printf style argument, only when verbose enabled
The new fouth one is to remove several places where we do `if (log &&
log->GetVerbose()) LLDB_LOGF (...)` in the sources today, and make both
styles consistent.
This PR implements that change, mechanically changing all LLDB_LOGV's to
LLDB_LOG_VERBOSE.
It also updates many of the `if (log && log->GetVerbose()) LLDB_LOGF`'s.
Some uses of this conditional expression do extra calculations in
addition to logging, and so those were left as-is so we're not doing
throwaway work when running without verbose logging.
There were many instances throughout lldb where callers are still doing
`if (log) LLDB_LOG*(...)`, a remnant of when all calls were to the `Log`
object's `Printf()` method, and you had to check if your local Log*
pointer was non-nullptr before calling the method. I removed those,
again keeping ones where work for logging is done in the block of code.
The code changes are all mechanical and uninteresting, but the question
of whether this naming change is widely agreed on is maybe worth
discussing.
Add a LoadCompilationPrefixMap() helper in SymbolFile::FindPlugin that
walks up from the symbol file's directory looking for a
compilation-prefix-map.json file. When found, each key→value entry is
applied to the module's source path mapping list, allowing LLDB to
resolve source file paths that were rewritten by -fdebug-prefix-map at
build time without requiring manual `settings set target.source-map`.
The JSON file format maps fake paths (as written into debug info) back
to their real on-disk counterparts:
{ "/fake/srcdir": "/real/srcdir" }
Directory results are cached so the filesystem is walked at most once
per unique directory across all modules loaded in a session.
Also apply the module's source path remappings in
SymbolFileDWARFDebugMap::ParseCompileUnitAtIndex when constructing
compile units from N_SO stabs. This mirrors what MakeAbsoluteAndRemap
does for the dSYM case so that fake paths baked into the debug map are
transparently resolved to real paths.
rdar://84824567
Assisted-By: Claude
Minimal infrastructure for a the SymbolLocator plugin that fetches debug
info from Microsoft SymStore repositories. This can work cross-platform
and for various debug info formats in principle, but the current plan is
focussed on PE/COFF on Windows with debug info in PDB files. Once we
have a stable first version, we'd like to add features like download,
environment variables, caching and progress feedback for users.
SymbolVendorPECOFF was tailored towards DWARF debug info so far. I added
code to load the PDB path from the executable (it only checked
gnu_debuglink so far) and not bail out if DWARF sections are missing, so
that in the PDB case we still call AddSymbolFileRepresentation() in the
very end of CreateInstance().
The API test in this patch mocks the directory layout from SymStore, so
it doesn't depend on SymStore.exe from the Windows SDK. It runs on all
platforms that link debug info in a PDB file, which is still just
Windows, but it could be cross-platform in principle.
-----
Relands with minor fixes: API tests create mocked SymStore in the test's
build directory. One log instruction was moved. One more object
access goes through module in SymbolFile.
In #183172, I added support for reading `S_REGREL32_INDIR` records. This
adds support in LLDB. The record is emitted by MSVC for structured
bindings and in the compiler generated coroutine stubs.
It describes a location at `*(Register + Offset) + OffsetInUdt`.
Equivalent to
```
DW_OP_breg{reg} {Offset}
DW_OP_deref
DW_OP_plus_uconst {OffsetInUdt}
```
LLVM doesn't create this record - it only creates `S_LOCAL`s for local
variables. We'll probably need `S_DEFRANGE_REGISTER_REL_INDIR` for this
- should be simple to figure out the structure, but I haven't seen that
record emitted yet.
Use the same format string in DWARFCompileUnit.cpp as we do everywhere
else to report the error from GetAttributeAddressRanges. I noticed the
inconsistency when looking at our system log:
```
2026-03-12 05:36:48.600914 -0700 0x2e66: DIE has no address range information
2026-03-12 05:36:48.600950 -0700 0x2e8c: DIE has no address range information
2026-03-12 05:36:48.601068 -0700 DIE(0x1dc2): DIE has no address range information
2026-03-12 05:36:48.631000 -0700 DIE(0x80f4): DIE has no address range information
```
This makes the constructions of string errors more concise and more
consistent, mainly by removing the `inconvertibleErrorCode()`.
Additional changes replace `createStringError(formatv(...), ...)` with
`createStringErrorV(...)`.
Assisted-by: Claude
When a class indirectly inherits from a class with virtual bases, it
will get an `LF_IVBCLASS` record in its fieldlist, even though it
doesn't directly inherit that class.
In the following example, `UserUser` inherits from `User`, which
virtually inherits from `VBase`:
```cpp
struct User : public virtual VBase {};
struct UserUser : public User {};
```
For this we get
```
0x1015 | LF_FIELDLIST [size = 72]
- LF_BCLASS
type = 0x1002 (-> 0x102A), offset = 0, attrs = public
- LF_IVBCLASS
base = 0x1003, vbptr = 0x1005, vbptr offset = 0, vtable index = 1
attrs = public (...)
0x1016 | LF_STRUCTURE [size = 48] `UserUser`
unique name: `.?AUUserUser@@`
vtable: <no type>, base list: <no type>, field list: 0x1015
options: has ctor / dtor | has unique name | overloaded operator | overloaded operator=, sizeof 16
0x1029 | LF_FIELDLIST [size = 56]
- LF_VBCLASS
base = 0x1003, vbptr = 0x1005, vbptr offset = 0, vtable index = 1
attrs = public (...)
0x102A | LF_STRUCTURE [size = 40] `User`
unique name: `.?AUUser@@`
vtable: <no type>, base list: <no type>, field list: 0x1029
options: has ctor / dtor | has unique name | overloaded operator | overloaded operator=, sizeof 16
```
If I understand correctly, then `LF_IVBCLASS` indicates that if this
class (e.g. `UserUser`) is created as the most derived object, it will
host the class (e.g. `VBase`).
The VS debugger actually shows this as a separate field. LLDB on the
other hand doesn't, so I removed it.
---------
Co-authored-by: Zequan Wu <zequanwu@google.com>
In https://github.com/llvm/llvm-project/pull/182956 we stopped using
access specifiers from DWARF for most Clang decls we create. This
parameter no longer needs to be modified and we can make it a
`const-ref`.
This patch removes all our manual adjustments to the access control
specifiers of Clang decls we create from DWARF.
This has led to occasional subtle bugs in the past (the latest being
https://github.com/llvm/llvm-project/issues/171913) and it's ultimately
redundant because Clang already has provisions for LLDB to bypass access
control for C++ and Objective-C. Access control doesn't affect name
lookup so really we're doing a lot of bookkeeping for not much benefit.
The only "feature" that relied on this was that `type lookup <foo>`
would print the access specifier in the output structure layout. I'm not
convinced that's worth keeping the infrastructure in place for (but
happy to be convinced otherwise).
I'd rather lean fully into the Clang access control bypass instead.
Note, i still kept the `AccessType` parameters to the various
`TypeSystemClang` APIs to reduce the size of the diff. A follow-up NFC
change will remove those parameters and adjust all the call-sites.
This was originally introduced to support kalimba DSPs featuring 24-bit
bytes by f03e6d84 and also c928de3e, but the kalimba support was mostly
removed by f8819bd5. This change removes the rest of the support, which
was far from complete.
LLVM added support for emitting the tagging offset of a variable for
hwasan/memtag-stack using the DW_AT_LLVM_tag_offset attribute in
dabd262. This patch teaches LLDB to read this attribute.
In #168245, I attempted to dump the available settings to Markdown. That
required a full build of LLDB. However, to build the docs, only the swig
wrappers should need to be compiled. The comment was that we should be
able to use the definitions from the TableGen files.
Currently, the property definitions in don't have information about the
path where they will be available. They only contain a `Definition`
which groups properties, so they can be added to
`OptionValueProperties`.
With this PR, I'm adding the path for each property definition. For
example, `symbols.enable-external-lookup` would have `Name =
enable-external-lookup, Path = symbols`. In LLDB itself, we don't need
this path, we only need it for the documentation. To avoid mismatches
between the actual path and the declared one, I added a debug-only check
when a property group is added to a parent
(`OptionValueProperties::AppendProperty`).
The TableGen emitter for the properties now additionally emits
`g_{definition}_properties_def`, which includes both the array of
properties and the expected path. This constant has to be used to
initialize a `OptionValueProperties`.
I couldn't test this for everything (e.g. IntelPT or ProcessKDP), but
the necessary changes are simple: (1) set the `Path` in the TableGen
file, (2) update `initialize` to use `_def`.
This is part of a patch series to clean up the
`TypeSystemClang::IsFloatingPointType` API. Currently the API is a bit
of a foot-gun because it returns `true` for both Complex floats and
vector types whose element types are floats, but most call-sites
probably don't handle these correctly. The former aligns with the
`clang::Type::isFloatingType` API, but the latter doesn't. This specific
implementation choice will be addressed in a separate patch. This patch
adds a new `CompilerType::IsRealFloatingPointType` API which clients can
use to query about non-complex floats (named after the similarly named
`clang::Type::isRealFloatingType`).
This allows us to clean up some of the callers which only wanted to
handle non-complex floats. I cleaned those up as part of this patch.
Wherever we checked for `is_float && !is_complex && !is_vector_type` i
just replaced it with the new API.
On encountering complex/vector floats, some of the ABI plugins would set
an error but then immediately overwrite the error to something else. I
just removed those branches entirely, since it was essentially dead
code.
Some of the diff is large because of un-indentation when we remove
if-guards, so I would recommend turning off white-space diff when
reviewing.
Removes duplicated code by simply calling into the base-class.
Motivation is just that I stumbled upon this when making changes in the
surrounding area.
LLDB's DWARF parser didn't support parsing DW_FORM_GNU_ref_alt and
DW_FORM_GNU_strp_alt forms which would cause any file loaded by LLDB to
fail to parse any DWARF. Added support for parsing this information
only, not for actually finding the debug info reference to an alternate
file or a string in an alternate file. These extensions are used by DWZ
files which are present in some linux distros, so it will be good for
LLDB to just be able to parse these without emitting an error like:
(lldb) b bar
warning: (arm64) /tmp/a.out unsupported DW_FORM values: 0x1f20 0x1f21
Currently we forcefully complete C++ types if we can't find their
definition for layout purposes. This ensures that we at least don't
crash in Clang when laying out the type. The definition is required for
types of members/array elements/base classes for the purposes of
calculating their layout. This is also true for Obj-C types, but we
haven't been forcefully completing those.
The test-case that's being un-XFAILed in this patch demonstrates a case
where not completing the super-class forcefully causes a clang crash.
rdar://168440264
Uncommenting `ENABLE_DEBUG_PRINTF` wasn't compiling on top-of-tree. This code may have been useful for debugging locally in the past but it doesn't seem compelling to keep this in the source-tree. Especially because it doesn't compile, indicating it's most likely not used.
This patch removes `DEBUG_PRINTF` entirely.
Introducing `llvm::createStringErrorV` caused a `0.5%` compile-time
regression because it's an inline function in a core header. This moves
the API to a new header to prevent including this function in files that
don't need it.
Also includes the header in the source files that have been using
`createStringErrorV` (which currently is just LLDB).
This changes `PdbAstBuilder` to a language-neutral abstract interface
and moves all of its functionality to the `PdbAstBuilderClang` derived
class.
All Clang-specific methods with external callers are now public methods
on `PdbAstBuilderClang`. `TypeSystemClang` and `UdtRecordCompleter` use
`PdbAstBuilderClang` directly.
Did my best to clean up includes and unused methods.
RFC for context:
https://discourse.llvm.org/t/rfc-lldb-make-pdbastbuilder-language-agnostic/89117
Currently, `SymbolFileNativePDB` calls several `PdbAstBuilder` methods
for side-effects to ensure the AST is populated.
This change adds new void-returning methods for `SymbolFileNativePDB` to
use as a hook instead, so that it doesn't depend on Clang-specific parts
of `PdbAstBuilder`'s interface.
This is part of the work to allow language-agnostic `PdbAstBuilder` (see
RFC:
https://discourse.llvm.org/t/rfc-lldb-make-pdbastbuilder-language-agnostic/89117)
This change implements part of
https://discourse.llvm.org/t/rfc-lldb-make-pdbastbuilder-language-agnostic/89117,
the goal of which is make `PdbAstBuilder` language-agnostic.
It adds several new interface methods to `PdbAstBuilder` which allow
`SymbolFileNativePDB` to interact with it without using Clang-specific
types. These new methods would make up much of the interface of a future
abstract `PdbAstBuilder`. `SymbolFileNativePDB` is updated to use these.
The old Clang-specific methods were renamed if used by
`UdtRecordCompleter` or internally, or removed where possible.
The list table header size was always computed assuming `DWARF32`, even
when parsing `DWARF64` units. This results in an incorrect header size
for `DWARF64` and can cause mis-parsing of list tables.
Fix this by using the Dwarf Format from the form parameters instead of
always assuming `DWARF32` when computing the `DWARF` header size.
Typedef/using declarations in structs and classes were not created with
the native PDB plugin. The following would only create `Foo` and
`Foo::Bar`:
```cpp
struct Foo {
struct Bar {};
using Baz = Bar;
using Int = int;
};
```
With this PR, they're created. One complication is that typedefs and
nested types show up identical. The example from above gives:
```
0x1006 | LF_FIELDLIST [size = 40, hash = 0x2E844]
- LF_NESTTYPE [name = `Bar`, parent = 0x1002]
- LF_NESTTYPE [name = `Baz`, parent = 0x1002]
- LF_NESTTYPE [name = `Int`, parent = 0x0074 (int)]
```
To distinguish nested types and typedefs, we check if the parent of a
type is equal to the current one (`parent(0x1002) == 0x1006`) and if the
basename matches the nested type name.
We used to search for constants using the name we parsed. For C++, this
would mean using the demangled struct name (from the unique name). This
name is not always equal to the one used for the struct's name by the
compiler. For example:
```
0x105E | LF_STRUCTURE [size = 120, hash = 0xF38F] ``anonymous namespace'::Anonymous<A::B::C<void> >::D`
unique name: `.?AUD@?$Anonymous@U?$C@X@B@A@@@?A0x8C295248@@`
```
We would use the unique name and get to `(anonymous
namespace)::Anonymous<struct A::B::C<void>>::D`. Then, when finding the
constant in the field list, we'd search for `(anonymous
namespace)::Anonymous<struct A::B::C<void>>::D::StaticMember`. This
wouldn't yield any results, because the constant will use the demangled
name as given by the compiler.
With this PR, we use the struct's name as given in the PDB and append
the member name.
This reverts commit
54a4da9df6.
MSVC supports an extension allowing to delete an array of objects via
pointer whose static type doesn't match its dynamic type. This is done
via generation of special destructors - vector deleting destructors.
MSVC's virtual tables always contain a pointer to the vector deleting
destructor for classes with virtual destructors, so not having this
extension implemented causes clang to generate code that is not
compatible with the code generated by MSVC, because clang always puts a
pointer to a scalar deleting destructor to the vtable. As a bonus the
deletion of an array of polymorphic object will work just like it does
with MSVC - no memory leaks and correct destructors are called.
This patch will cause clang to emit code that is compatible with code
produced by MSVC but not compatible with code produced with clang of
older versions, so the new behavior can be disabled via passing
-fclang-abi-compat=21 (or lower).
Fixes https://github.com/llvm/llvm-project/issues/19772
The ObjectFile plugin interface accepts an optional DataBufferSP
argument. If the caller has the contents of the binary, it can provide
this in that DataBufferSP. The ObjectFile subclasses in their
CreateInstance methods will fill in the DataBufferSP with the actual
binary contents if it is not set.
ObjectFile base class creates an ivar DataExtractor from the
DataBufferSP passed in.
My next patch will be a caller that creates a VirtualDataExtractor with
the binary data, and needs to pass that in to the ObjectFile plugin,
instead of the bag-of-bytes DataBufferSP. It builds on the previous
patch changing ObjectFile's ivar from DataExtractor to DataExtractorSP
so I could pass in a subclass in the shared ptr. And it will be using
the VirtualDataExtractor that Jonas added in
https://github.com/llvm/llvm-project/pull/168802
No behavior is changed by the patch; we're simply moving the creation of
the DataExtractor to the caller, instead of a DataBuffer that is
immediately used to set up the ObjectFile DataExtractor. The patch is a
bit complicated because all of the ObjectFile subclasses have to
initialize their DataExtractor to pass in to the base class.
I ran the testsuite on macOS and on AArch64 Ubutnu. (btw David, I ran it
under qemu on my M4 mac with SME-no-SVE again, Ubuntu 25.10, checked
lshw(1) cpu capabilities, and qemu doesn't seem to be virtualizing the
SME, that explains why the testsuite passes)
rdar://148939795
---------
Co-authored-by: Jonas Devlieghere <jonas@devlieghere.com>
Some months ago, the LookupInfo constructor logic was refactored to not
depend on language specific logic, and use languages plugins instead. In
this refactor, when the language type is unknown, a single LookupInfo
object will handle multiple languages. This doesn't work well, as
multiple languages might want to configure the LookupInfo object in
different ways. For example, different languages might want to set the
m_lookup_name differently from each other, but the previous
implementation would pick the first name a language provided, and
effectively ignored every other language. Other fields of the LookupInfo
object are also configured in incompatible ways.
This approach doesn't seem to be a problem upstream, since only the
C++/Objective-C language plugins are available, but it broke downstream
on the Swift fork, as adding Swift to the list of default languages when
the language type is unknown breaks C++ tests.
This patch makes it so instead of building a single LookupInfo object
for multiple languages, one LookupInfo object is built per language
instead.
rdar://159531216
Depends on:
* https://github.com/llvm/llvm-project/pull/170132
Clang gained the `-gtemplate-alias` not too long ago, which emits C++
alias templates as `DW_TAG_template_alias` (instead of
`DW_TAG_typedef`). The main difference is that `DW_TAG_template_alias`
has `DW_TAG_template_XXX` children. The flag was not enabled by default
because consumers (mainly LLDB) didn't know how to handle it. This patch
adds rudimentary support for debugging with `DW_TAG_template_alias`.
This patch simply creates the same kind of `TypedefDecl` as we do for
`DW_TAG_typedef`. The more complete solution would be to create a
`TypeAliasTemplateDecl` and associated `TypeAliasDecl`. But that would
require DWARF to carry generic template information, but currently each
`DW_TAG_template_alias` represents a concrete instantiation. We could
probably hack up some working AST representation that includes the
template parameters, but I currently don't see a compelling reason to.
All we need is the `DW_AT_name` and the `DW_AT_type` that the typedef
refers to.
rdar://137499401
We almost had RTTI support for `DWARFASTParserClang`, but because
`classof` was protected, using `llvm::cast`/etc. on it would fail to
compile with:
```
llvm/include/llvm/Support/Casting.h:64:57: error: 'classof' is a protected member of 'DWARFASTParserClang'
64 | static inline bool doit(const From &Val) { return To::classof(&Val); }
| ^
llvm/include/llvm/Support/Casting.h:110:32: note: in instantiation of member function 'llvm::isa_impl<DWARFASTParserClang, lldb_private::plugin::dwarf::DWARFASTParser>::doit' requested here
110 | return isa_impl<To, From>::doit(*Val);
```
This patch makes `classof` public and turns `static_cast`s of
`DWARFASTParserClang` into `llvm::cast`s.
When a PDB is loaded through `target symbols add <pdb-path>`, its
`m_objectfile_sp` is an `ObjectFilePDB` instead of `ObjectFilePECOFF`
(the debugged module). In both the native and DIA plugin, some paths
assumed that `m_objectfile_sp` is the debugged module. With this PR,
they go through `m_objfile_sp->GetModule()->GetObjectFile()`.
For the DIA plugin, this lead to an assertion failure
(https://github.com/llvm/llvm-project/issues/169628#issuecomment-3582555277)
and for both plugins, it meant that the symbol table wasn't loaded.
This patch fixes and eliminates the possibility of SupportFileSP ever
being nullptr. The support file was originally treated like a value
type, but became a polymorphic type and therefore has to be stored and
passed around as a pointer.
To avoid having all the callers check the validity of the pointer, I
introduced the invariant that SupportFileSP is never null and always
default constructed. However, without enforcement at the type level,
that's fragile and indeed, we already identified two crashes where
someone accidentally broke that invariant.
This PR introduces a NonNullSharedPtr to prevent that. NonNullSharedPtr
is a smart pointer wrapper around std::shared_ptr that guarantees the
pointer is never null. If default-constructed, it creates a
default-constructed instance of the contained type. Note that I'm using
private inheritance because you shouldn't inherit from standard library
classes due to the lack of virtual destructor. So while the new
abstraction looks like a `std::shared_ptr`, it is in fact **not** a
shared pointer. Given that our destructor is trivial, we could use
public inheritance, but currently there's no need for it.
rdar://164989579
Currently LLDB's `ParseRustVariantPart` generates the following
`CXXRecordDecl` for a Rust enum
```rust
enum AA {
A(u8)
}
```
```
CXXRecordDecl 0x5555568d5970 <<invalid sloc>> <invalid sloc> struct AA
|-CXXRecordDecl 0x5555568d5ab0 <<invalid sloc>> <invalid sloc> union test_issue::AA$Inner definition
| |-CXXRecordDecl 0x5555568d5d18 <<invalid sloc>> <invalid sloc> struct A$Variant definition
| | |-DefinitionData pass_in_registers aggregate standard_layout trivially_copyable trivial
| | | `-Destructor simple irrelevant trivial needs_implicit
| | `-FieldDecl 0x555555a77880 <<invalid sloc>> <invalid sloc> value 'test_issue::AA::A'
| `-FieldDecl 0x555555a778f0 <<invalid sloc>> <invalid sloc> $variant$ 'test_issue::AA::test_issue::AA$Inner::A$Variant'
|-CXXRecordDecl 0x5555568d5c48 <<invalid sloc>> <invalid sloc> struct A definition
| `-FieldDecl 0x555555a777e0 <<invalid sloc>> <invalid sloc> __0 'unsigned char'
`-FieldDecl 0x555555a77960 <<invalid sloc>> <invalid sloc> $variants$ 'test_issue::AA::test_issue::AA$Inner'
```
While when the Rust enum type name is the same as its variant name, the
generated `CXXRecordDecl` becomes the following – there's a circular
reference between `struct A$Variant` and `struct A`, causing #163048.
```rust
enum A {
A(u8)
}
```
```
CXXRecordDecl 0x5555568d5760 <<invalid sloc>> <invalid sloc> struct A
|-CXXRecordDecl 0x5555568d58a0 <<invalid sloc>> <invalid sloc> union test_issue::A$Inner definition
| |-CXXRecordDecl 0x5555568d5a38 <<invalid sloc>> <invalid sloc> struct A$Variant definition
| | `-FieldDecl 0x5555568d5b70 <<invalid sloc>> <invalid sloc> value 'test_issue::A' <---- bug here
| `-FieldDecl 0x5555568d5be0 <<invalid sloc>> <invalid sloc> $variant$ 'test_issue::A::test_issue::A$Inner::A$Variant'
`-FieldDecl 0x5555568d5c50 <<invalid sloc>> <invalid sloc> $variants$ 'test_issue::A::test_issue::A$Inner'
```
The problem was caused by `GetUniqueTypeNameAndDeclaration` not
returning the correct qualified name for DWARF DIE `test_issue::A::A`,
instead, it returned `A`. This caused `ParseStructureLikeDIE` to find
the wrong type `test_issue::A` and returned early.
The failure in `GetUniqueTypeNameAndDeclaration` appears to stem from a
language check that returns early unless the language is C++. I changed
it so Rust follows the C++ path rather than returning. I’m not entirely
sure this is the right approach — Rust’s qualified name rules look
similar, but not identical? Alternatively, we could add a Rust-specific
implementation that forms qualified names according to Rust's rules.
This patch is updating the reading capabilities of the LLDB DWARF parser
for a llvm-dwp patch https://github.com/llvm/llvm-project/pull/167457
that will emit .dwp files where the compile units are DWARF32 and the
.debug_str_offsets tables will be emitted as DWARF64 to allow .debug_str
sections that exceed 4GB in size.
MSVC supports an extension allowing to delete an array of objects via
pointer whose static type doesn't match its dynamic type. This is done
via generation of special destructors - vector deleting destructors.
MSVC's virtual tables always contain a pointer to the vector deleting
destructor for classes with virtual destructors, so not having this
extension implemented causes clang to generate code that is not
compatible with the code generated by MSVC, because clang always puts a
pointer to a scalar deleting destructor to the vtable. As a bonus the
deletion of an array of polymorphic object will work just like it does
with MSVC - no memory leaks and correct destructors are called.
This patch will cause clang to emit code that is compatible with code
produced by MSVC but not compatible with code produced with clang of
older versions, so the new behavior can be disabled via passing
-fclang-abi-compat=21 (or lower).
This is yet another attempt to land vector deleting destructors support
originally implemented by
https://github.com/llvm/llvm-project/pull/133451.
This PR contains fixes for issues reported in the original PR as well as
fixes for issues related to operator delete[] search reported in several
issues like
https://github.com/llvm/llvm-project/pull/133950#issuecomment-2787510484https://github.com/llvm/llvm-project/issues/134265
Fixes https://github.com/llvm/llvm-project/issues/19772
