[LLVMdev] debug metadata incomplete for array arguments to functions?

[Renato Golin]

On 15 July 2011 12:18, Eli Bendersky <eliben at gmail.com> wrote:
> Do you imply that DWARF doesn't support this construct, so the "cause is
> lost"?

Absolutely not!

I was just giving you some hints on how to investigate *how* this
construct needs to be represented in IR/Dwarf.

Using other tool-chains to produce Dwarf, that is both correct and
reflects the code as is, will help you trace back how the IR looks
like.

I was also suggesting, as Devang also said, that you start by changing
the IR by hand, before try more adventurous changes in Clang.

*If* it's not possible to represent that in IR (I have no idea), then
LLVM needs to be changed first.

Only when you're sure that there is a clear representation of your
problem in IR that you can go on changing Clang.

cheers,
--renato