[cfe-dev] Ownership of declarations in clang?

Chris Lattner clattner at apple.com
Mon Feb 25 15:13:46 PST 2008

On Feb 25, 2008, at 3:03 PM, Eli Friedman wrote:

> Currently, clang leaks a lot of stuff.

Yes :(, this is an open area that is very important.

> I was looking at this using
> valgrind, and then I was trying to figure out what code is supposed to
> be responsible for cleaning up the declarations generated by Sema
> after the file is finished being parsed.  I wasn't able to find any
> such code.

Ted may know better than me at this point, but I think the idea is  
that the TranslationUnit object should own (and thus destroy) the  
decls when it is torn down.

> What exactly is the expected ownership model for the declarations
> generated by Sema?  They obviously need to be around until we are
> finished parsing the file, but by then, it appears only the parser
> (specifically, the root scope) is holding onto pointers to
> declarations, and it doesn't know how to delete them.

TranslationUnit should own them.  I don't know what its current state  
is, Ted?

In addition to decls, sema leaks memory in many cases when an error  
occurs.  I implemented what I think is the best solution in  

1) on entry, the CallExpr AST node is created, which takes ownership  
of all the operands.
2) the ast node is owned by an OwningPtr<> object.
3) various checks are done, if the code is invalid and an early exit  
is taken, the Call, and all its operands, is deleted.
4) if the expr checks out, the callexpr is returned by using  
the .take() method on the OwningPtr.

I think this leads to relatively simple code and uses RAII to prevent  
obvious leaks.  Most of sema isn't following this approach yet though.

Also, I'm not sure if all the virtual dtors for the various AST nodes  
are deleting their children right now.


More information about the cfe-dev mailing list