[llvm-commits] [PATCH] Add a ValueMap type, whose keys act like WeakVHs

[Jeffrey Yasskin]

On Sat, Oct 17, 2009 at 4:05 PM, Chris Lattner <clattner at apple.com> wrote:
>
> On Oct 17, 2009, at 12:42 AM, jyasskin at gmail.com wrote:
>>
>> I guess I'll have to figure out how to fix it before submitting.
>
> I'll review it when it is fixed.

I've fixed lit.py in r84387. This patch didn't need to change.

>> Description:
>> ValueMap<Value*, T> is safe to use even when keys get RAUWed and
>> deleted
>> during its lifetime. By default the keys act like WeakVHs, but users
>> can
>> pass a third template parameter to configure how updates work and
>> whether to do anything beyond updating the map on each action.
>>
>> This type should make http://codereview.appspot.com/133043/diff/1/2
>> quite a bit simpler, but I haven't yet updated it so I could have
>> missed
>> something.
>
> Sounds nice, what happens when X->RAUW(Y) and both X and Y are already
> in the map?

Right now, you wind up with Y mapped to X's old target, and Y's old
target lost. ValueMapTest.DefaultCollisionBehavior tests for this. It
should be easy to add the ability to configure this behavior
(Config::Merge(ValueT& OldTarget, const ValueT& NewTarget), maybe with
extra parameters), but I didn't need it for my other changes.