[PATCH] D26051: Import/update constants from the DWARF 5 public review draft document.

[David Blaikie via llvm-commits]

On Fri, Oct 28, 2016 at 8:37 AM Adrian Prantl <aprantl at apple.com> wrote:

> On Oct 27, 2016, at 8:49 PM, David Blaikie <dblaikie at gmail.com> wrote:
>
>
>
> On Thu, Oct 27, 2016 at 5:13 PM Adrian Prantl <aprantl at apple.com> wrote:
>
> aprantl created this revision.
> aprantl added reviewers: echristo, dblaikie, probinson.
> aprantl added subscribers: clayborg, llvm-commits.
> Herald added a subscriber: mehdi_amini.
>
> I noticed two incompatible changes:
>
> 1. DW_AT_dwo_id was removed because the dwo_id was moved into the CU
> header. Note that the attribute
>
>
> Well, that's a nice change... I should read the spec in more detail. (it
> would avoid the DWP tool from having to parse any DWARF,
>
>
> Yes, this would also allow us to verify the dwoid in clang for module
> debug info (we don't do it because we don't want to parse DWARF in the
> frontend).
>
> I think - though it's still nice to parse some to find the dwo_name for
> better diagnostics, at least we can wait until we've identified a problem
> before doing that anyway)
>
>
> value was reserved, so LLVM can continue to emit it without causing
> ambiguity for now.
>
> 1. The one change that could cause a problem here is that
> DW_AT_ranges_base has been replaced by DW_AT_rnglist_base.
>
>
> That's a weird one - how's a DWARF consumer meant to identify whether
> they're dealing with new or old (whether they should look in debug_ranges
> or debug_rnglists)? Just based on the version in the header? (is there
> prior art for that sort of technique?)
>
>
> The DW_AT_ranges_base attribute was never part of a DWARF standard, AFAIK
> it was only used in a proposal. For proper DWARF 4 LLVM should probably
> emit the DW_AT_GNU_ranges_base extension (which, looking at
> DwarfDebug::finalizeModuleInfo() it does) and for proper DWARF 5 use the
> standardized format with the new section.
>

Oh, right, fission thing, never got baked in the way we are using it. No
worries then.


>
> I suppose in theory our dumper could be smart enough to print the right
> attribute string depending on version - though that seems annoying...
>
>
> Looks like this is not even necessary.
>

Yep - legit. Thanks for the explanation!

Carry on, then :)

- Dave


>
> -- adrian
>
>
> - David
>
>
>
> David/Eric: Would merely renaming the constant break any existing DWARF
> (4) tooling?
>
>
> https://reviews.llvm.org/D26051
>
> Files:
>   include/llvm/Support/Dwarf.def
>   include/llvm/Support/Dwarf.h
>   lib/DebugInfo/DWARF/DWARFDebugInfoEntry.cpp
>   lib/DebugInfo/DWARF/DWARFUnit.cpp
>   lib/Support/Dwarf.cpp
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20161028/fc6c68a8/attachment.html>