diff options
author | Michael Buch <michaelbuch12@gmail.com> | 2025-10-17 23:27:57 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-10-17 23:27:57 +0100 |
commit | bf2d84db8e95cf0dbf782f6609b034427ab1c07d (patch) | |
tree | 3ef8d24241a725dc851f43eeae9163cb3411076f /lldb/unittests/Process/gdb-remote/GDBRemoteCommunicationClientTest.cpp | |
parent | 9522f989b06c6e7eb031758e71cd1c69755e7f32 (diff) | |
download | llvm-bf2d84db8e95cf0dbf782f6609b034427ab1c07d.zip llvm-bf2d84db8e95cf0dbf782f6609b034427ab1c07d.tar.gz llvm-bf2d84db8e95cf0dbf782f6609b034427ab1c07d.tar.bz2 |
[lldb][ObjC] Consult Objective-C runtime decl vendor when completing type (#164011)
(Note, this upstreams code that has been deployed on Apple's Swift LLDB
for many years at this point).
When a `ValueObject` computes its "complete type"
(`MaybeCalculateCompleteType`), it gives the language runtimes a chance
to override the type known to it. The current implementation of
`ObjCLanguageRuntime::GetRuntimeType`, however, didn't consult the
`AppleObjCDeclVendor` to look for types.
As demonstrated in the attached test, when we don't have debug-info for
a base class type (most commonly happens when inheriting from system
framework types) we would not be able to deduce ivars of that type.
However, the runtime knows about the ivars, so we should be able to
retrieve them.
There's still a couple of caveats for future follow-up/investigation:
1. `frame var` isn't able to access such backing ivars explicitly (even
if they do exist)
2. When compiling with `-gmodules`, LLDB gets confused about what is
correct source of information for these decls is.
rdar://162069497
Diffstat (limited to 'lldb/unittests/Process/gdb-remote/GDBRemoteCommunicationClientTest.cpp')
0 files changed, 0 insertions, 0 deletions