[PATCH] D133499: [clang]: Add DeclContext::dumpDecl() in order to conveniently dump an AST from a DeclContext.

Shafik Yaghmour via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Fri Sep 9 10:28:48 PDT 2022


shafik added inline comments.


================
Comment at: clang/lib/AST/ASTDumper.cpp:205
+  if (const Decl *D = dyn_cast<Decl>(this))
+    D->dump();
+}
----------------
aaron.ballman wrote:
> shafik wrote:
> > aaron.ballman wrote:
> > > erichkeane wrote:
> > > > aaron.ballman wrote:
> > > > > erichkeane wrote:
> > > > > > One thing to note is that the 'else' case here is a little uninformative.  See https://clang.llvm.org/doxygen/DeclBase_8cpp_source.html#l00915 for some similar logic here (though not sure we should be emulating that).
> > > > > > 
> > > > > > More, I wonder if there is SOME message here that should be dumped for 'else'.
> > > > > Looking at what inherits from `DeclContext`, is there use of it which is *not* a `Decl`? I couldn't find a use where it's not also a `Decl`.
> > > > I've DEFINITELY run into it in the debugger before, but I have no idea WHAT case that is. It is sometimes just "DeclContext is an invalid pointer" kinda thing, so it might be worth-while to have SOME output besides "print nothing", particularly when debugging.
> > > IIRC, the case this comes up in is when the object is only partially constructed, and so I agree that having an `else` clause here would be useful -- because this interface is predominately used from a debugger, it has to deal with special "impossible" situations a bit more carefully.
> > So looking at the other dump member functions, all of them seem to assume we have a valid `DeclContext` and so they do not have any else either.
> > 
> > So maybe we want to think about making the rest a bit smarter as well?
> > So maybe we want to think about making the rest a bit smarter as well?
> 
> What do you have in mind?
> 
> I think it's plausible to be in a partially constructed state in terms of when ctors are called, but I'm far less worried about people calling a dump method in the middle of the member inits happening.
> > So maybe we want to think about making the rest a bit smarter as well?
> 
> What do you have in mind?
> 
> I think it's plausible to be in a partially constructed state in terms of when ctors are called, but I'm far less worried about people calling a dump method in the middle of the member inits happening.

Apologies, what I meant was if we are going to add an `else` here then we should be consistent and add it in the other dump functions as well. 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D133499/new/

https://reviews.llvm.org/D133499



More information about the cfe-commits mailing list