[cfe-commits] [PATCH] Implementation quantity limits to prevent crashes

Douglas Gregor dgregor at apple.com
Wed Oct 12 09:40:08 PDT 2011


On Oct 10, 2011, at 7:47 PM, Aaron Ballman wrote:

> On Sun, Oct 9, 2011 at 12:35 PM, Douglas Gregor <dgregor at apple.com> wrote:
>>>> I'd prefer that we keep the open/close source locations within BalancedDelimiterTracker, which would simplify the interface a bit, since *every* user of BalancedDelimiterTracker needs to track this information anyway.
>>> 
>>> I've implemented this change as well -- it was a good point, and it
>>> does make things cleaner.  I had to update the signature for
>>> ParseInnerNamespace though, since that was the one use case that
>>> needed the opening brace to be passed in manually.
>> 
>> I see that you're holding on to the opening source location (which is good), but it doesn't look like you've simplified the interface at all. Why bother to still pass in the "open" and "close" addresses in these functions:
> 
> I've completed these changes as discussed, and I think the code has
> turned out to be better for them.  Thanks for the clarification
> earlier today!
> 
> I did have to modify a few parser signatures to pass around a
> BalancedDelimiterTracker, but it ended up simplifying the interfaces
> in those cases as well.


Very nice. I've committed this as r141782, with a few tiny tweaks:

+    unsigned& get(tok::TokenKind t) {
+      switch (t) {
+      default: assert(false && "Unexpected balanced token"); return ThrowAway;
+      case tok::l_brace:  return Brace;
+      case tok::l_paren:  return Paren;
+      case tok::l_square: return Square;
+      case tok::less:  return Less;
+      case tok::lesslessless:  return LLLess;
+      }
+    }

There's no need to behave well in the case where this class is abused, so I removed ThrowAway.

I also added some documentation to the new classes, so people know how to use them, and updated the "How to add an expression" section of the internals manual.

Thanks!

	- Doug



More information about the cfe-commits mailing list