[Lldb-commits] [clang] [clang-tools-extra] [lldb] [clang][TypePrinter] Unify printing of anonymous/unnamed tag types (PR #169445)

[Michael Buch via lldb-commits]

Michael137 wrote:

> * This could probably get rid of the additional bool flags, and rely on PrintingPolicy entirely.
> * Instead of the `printAnonymousTagDecl` helper, we could probably implement a `print` method for TagDecls, which takes care of the named case, and for the unnamed case, takes care of the EnumDecl and RecordDecl specific behavior, including the typedef to unnamed class.

This is mostly implemented now in the latest patchset. My only reservation about moving the printing logic into `TagDecl::printName` is that we no longer are able to easily determine whether to print the tag keyword. Previously we've done this in: https://github.com/llvm/llvm-project/blob/762a171b3d6f6eb64dc5a844fcb25caa4ece7d00/clang/lib/AST/TypePrinter.cpp#L1523-L1529

But now we have to rely on the `PrintingPolicy` telling us this, putting the burden on the callers (specifically callers of `QualType::print`). Maybe that's an OK trade-off? You can see this in the `clangd/unittests/HoverTests.cpp` changes in the latest revision. We now don't omit the tag keyword in the name itself. For the `Hover` case specifically we could fix this by setting the new `SuppressTagKeywordInAnonNames` policy here: https://github.com/llvm/llvm-project/blob/762a171b3d6f6eb64dc5a844fcb25caa4ece7d00/clang-tools-extra/clangd/Hover.cpp#L175-L183

But there's probably other places where this needs to be adjusted. Wdyt? Is that a reasonable trade-off?

https://github.com/llvm/llvm-project/pull/169445