arm64ec incorrect mangling instantiating C++ template #115231
Labels
c++
diverges-from:msvc
Does the clang frontend diverge from msvc on this issue
llvm:ir
llvm:support
release:backport
release:cherry-pick-failed
Milestone
Comments
|
@llvm/issue-subscribers-c-1 Author: Eli Friedman (efriedma-quic)
```
template <typename X> struct Wrapper {
int GetValue(void) const;
};
struct A {
int GetValue(void) const;
};
template<typename X> int Wrapper<X>::GetValue(void) const { return 3; }
template class Wrapper<A>;
```
clang produces CC @mstorsjo @cjacek @dpaoliello @steplong |
frederick-vs-ja
added
diverges-from:msvc
|
@llvm/issue-subscribers-clang-frontend Author: Eli Friedman (efriedma-quic)
```
template <typename X> struct Wrapper {
int GetValue(void) const;
};
struct A {
int GetValue(void) const;
};
template<typename X> int Wrapper<X>::GetValue(void) const { return 3; }
template class Wrapper<A>;
```
clang produces CC @mstorsjo @cjacek @dpaoliello @steplong |
|
I believe that I have the correct algorithm - PR out shortly |
dpaoliello
added a commit
that referenced
this issue
Nov 13, 2024
Arm64EC uses a special name mangling mode that adds `$$h` between the symbol name and its type. In MSVC's name mangling `@` is used to separate the name and type BUT it is also used for other purposes, such as the separator between paths in a fully qualified name. The original algorithm was quite fragile and made assumptions that didn't hold true for all MSVC mangled symbols, so instead of trying to improve this algorithm we are now using the demangler to indicate where the insertion point should be (i.e., to parse the fully-qualified name and return the current string offset). Also fixed `isArm64ECMangledFunctionName` to search for `@$$h` since the `$$h` must always be after a `@`. Fixes #115231
EugeneZelenko
added
llvm:support
llvm:ir
and removed
clang:frontend
|
/cherry-pick fa0cf3d |
|
Failed to cherry-pick: fa0cf3d https://github.com/llvm/llvm-project/actions/runs/11842908473 Please manually backport the fix and push it to your github fork. Once this is done, please create a pull request |
dpaoliello
added a commit
to dpaoliello/llvm-project
that referenced
this issue
Nov 14, 2024
Arm64EC uses a special name mangling mode that adds `$$h` between the symbol name and its type. In MSVC's name mangling `@` is used to separate the name and type BUT it is also used for other purposes, such as the separator between paths in a fully qualified name. The original algorithm was quite fragile and made assumptions that didn't hold true for all MSVC mangled symbols, so instead of trying to improve this algorithm we are now using the demangler to indicate where the insertion point should be (i.e., to parse the fully-qualified name and return the current string offset). Also fixed `isArm64ECMangledFunctionName` to search for `@$$h` since the `$$h` must always be after a `@`. Fixes llvm#115231
tru
pushed a commit
to dpaoliello/llvm-project
that referenced
this issue
Nov 15, 2024
Arm64EC uses a special name mangling mode that adds `$$h` between the symbol name and its type. In MSVC's name mangling `@` is used to separate the name and type BUT it is also used for other purposes, such as the separator between paths in a fully qualified name. The original algorithm was quite fragile and made assumptions that didn't hold true for all MSVC mangled symbols, so instead of trying to improve this algorithm we are now using the demangler to indicate where the insertion point should be (i.e., to parse the fully-qualified name and return the current string offset). Also fixed `isArm64ECMangledFunctionName` to search for `@$$h` since the `$$h` must always be after a `@`. Fixes llvm#115231
nikic
pushed a commit
to rust-lang/llvm-project
that referenced
this issue
Nov 20, 2024
Arm64EC uses a special name mangling mode that adds `$$h` between the symbol name and its type. In MSVC's name mangling `@` is used to separate the name and type BUT it is also used for other purposes, such as the separator between paths in a fully qualified name. The original algorithm was quite fragile and made assumptions that didn't hold true for all MSVC mangled symbols, so instead of trying to improve this algorithm we are now using the demangler to indicate where the insertion point should be (i.e., to parse the fully-qualified name and return the current string offset). Also fixed `isArm64ECMangledFunctionName` to search for `@$$h` since the `$$h` must always be after a `@`. Fixes llvm#115231
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
clang produces
?GetValue@$$h?$Wrapper@UA@@@@QEBAHXZ, MSVC produces?GetValue@?$Wrapper@UA@@@@$$hQEBAHXZ. I guess the algorithm in llvm::getArm64ECMangledFunctionName is wrong.CC @mstorsjo @cjacek @dpaoliello @steplong
The text was updated successfully, but these errors were encountered: