[PATCH] D14697: [Verifier] Fix !dbg validation if Scope is the Subprogram

[David Blaikie via llvm-commits]

On Mon, Nov 16, 2015 at 3:28 PM, Keno Fischer via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> loladiro added a comment.
>
> I'm not sure the added complexity of walking the chain and checking
> everything would be worth it. It seems like that would just be manually
> inlining  getInlinedAtScope + doing the memoization.


Yep, something like that, I think (I forget how exactly the scope V
inlinedAt scope all plays together here)


> It seems like a reasonable assumption that the scope declared on one
> instruction would also be declared on instructions nearby, while that's not
> necessarily the case for intermediate ones (plus if it is we'll put it in
> the list when we get there).
>

Not quite following - what I'm not sure is why it's nice to cache the
immediate scope, but not intermediate ones - reasonably nested code may
have many nested scopes (big switches from tablegen, for example?) with an
immediately shared outer scope that's still nesnted within the function
scope.

Just seems a bit quirky, that's all - I /think/ we should probably just
skip caching the immediate scope (& assume the scope walk is cheap enough)
or cache all of them.


>
>
> http://reviews.llvm.org/D14697
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151116/00c2ecd5/attachment.html>