[PATCH] D16041: Change vfs::FileSystem to be managed with std::shared_ptr

Benjamin Kramer via cfe-commits cfe-commits at lists.llvm.org
Mon Jan 11 14:13:54 PST 2016


On Mon, Jan 11, 2016 at 8:08 PM, Owen Anderson <resistor at mac.com> wrote:
>
>> On Jan 11, 2016, at 8:25 AM, David Blaikie <dblaikie at gmail.com> wrote:
>>
>>
>>
>> On Sun, Jan 10, 2016 at 11:42 PM, Owen Anderson via cfe-commits <cfe-commits at lists.llvm.org> wrote:
>> resistor created this revision.
>> resistor added reviewers: chandlerc, bkramer, klimek.
>> resistor added a subscriber: cfe-commits.
>> resistor set the repository for this revision to rL LLVM.
>> Herald added a subscriber: klimek.
>>
>> Managing it with IntrusiveRefCntPtr caused the virtual destructor not to be called properly.
>>
>> Regardless of the broader discussion on this patch, I'm confused by why this ^ would be the case. What is it that IntrusiveRefCntPtr is doing that's causing problems with destruction? (& I'm all for changing this to non-intrusive smart pointers over intrusive ones anyway, but I'd still like to understand the extra motivation here)
>>
>
> ThreadSafeRefCountedBase, which classes must inherit from in order to use IntrusiveRefCntPtr, does not handle virtual destructors properly.  For the non-thread safe version, there is RefCountBaseVPTR which solves this problem.  An alternative solution would have been to add a corresponding ThreadSafeRefCountedBaseVPTR, but IMO at that point one might as well use shared_ptr since it’s 90% of the way there.

I find this surprising. ThreadSafeRefCountedBase<FileSystem> calls
"delete static_cast<const FileSystem *>(this)". As FileSystem has a
virtual dtor, polymorphic deletion should just work. Am I missing
something?

- Ben


More information about the cfe-commits mailing list