[Lldb-commits] [PATCH] D130796: [LLDB][NativePDB] Switch to use DWARFLocationList.

Zequan Wu via Phabricator via lldb-commits lldb-commits at lists.llvm.org
Wed Aug 3 14:06:12 PDT 2022


zequanwu added a comment.

In D130796#3696351 <https://reviews.llvm.org/D130796#3696351>, @labath wrote:

> In D130796#3692243 <https://reviews.llvm.org/D130796#3692243>, @zequanwu wrote:
>
>> In D130796#3691227 <https://reviews.llvm.org/D130796#3691227>, @zequanwu wrote:
>>
>>>> If you find yourself needing to do extra work to work its limitations, we should fix that algorithm instead.
>>>
>>> That makes sense. I'll work on fixing `RangeVectorData`.
>>
>> After thinking it again, not just `RangeVectorData` need to be written but also `RangeVector`, although they are similar. It requires rewrite the whole file `RangeMap.h` and all other places uses its APIs.
>
> Why is that? I don't think we need to change the API -- just the way it is implemented. And the RangeVectorData class already contains the upper_bound argument, which should make it possible to find the element containing a value quickly <https://en.wikipedia.org/wiki/Interval_tree#Java_example:_Searching_a_point_or_an_interval_in_the_tree>. That basically means taking the algorithm in FindEntryIndexesThatContain, but making it stop at the first element found.
>
> And you can just ignore the RangeVector class. That problem (two separate classes) can be handled separately and holistically.
>
> Alternatively, if we don't want to/need to support overlaps, we could switch to some completely different class, which implements searches correctly. llvm::IntervalMap, for example.

Oh, I didn't notice that FindEntryIndexesThatContain exist until now. This function indicates that it allows multiple overlapping ranges exist. I originally thought that it doesn't make sense to have overlapping ranges in variable locations and it must be a bug in debug info generation. But now what you are saying at here <https://github.com/llvm/llvm-project/issues/56616#issuecomment-1193951602>, "it means all of them are valid", makes sense to me. FindEntryIndexThatContains does the exact thing you mentioned.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D130796/new/

https://reviews.llvm.org/D130796



More information about the lldb-commits mailing list