[Lldb-commits] [PATCH] D75929: [DebugInfo] Support DWARFv5 index sections.

[Pavel Labath via Phabricator via lldb-commits]

labath added a comment.

In D75929#1928964 <https://reviews.llvm.org/D75929#1928964>, @ikudrin wrote:

> In D75929#1926834 <https://reviews.llvm.org/D75929#1926834>, @labath wrote:
>
> > (btw, is there a test case for the "unknown column" code path?)
>
>
> Yes, it is checked in `llvm/test/DebugInfo/X86/debug-cu-index-unknown-section.s`, which was added in D75609 <https://reviews.llvm.org/D75609> and then extended in D75668 <https://reviews.llvm.org/D75668>.


Got it. Thanks.

> As for unknown columns in general, I believe they are not that important to complicate the code too much. Before D75609 <https://reviews.llvm.org/D75609>, `llvm-dwarfdump` just crashed when saw them. `dwarfdump` prints some useless (for a user) error message. An unknown column cannot be used by clients of the library because they do not know what to do with it. Dumping is the only reason to support unknown identifiers, and that should be done as simple as possible. If the current implementation seems too complicated, we can consider, for example, dropping printing raw IDs for unknown sections.

Yeah, I agree that they are not very important, but it would be a pitty to lose them. OTOH, the lazily-initialized parallel array solution seems like it's more complicated that it should be. Maybe we drop the "lazy" part, rename it to `RawColumnKinds` and always store both? It's not like that's going to waste a bunch of memory, and it will be easier to understand (I hope).


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75929/new/

https://reviews.llvm.org/D75929