[Dwarf-discuss] Proposal to encode Changed Function Signatures in dwarf

[Frank Ch. Eigler]

Hi -

> > Thanks. Now let us say I will not try to add those "changed function
> > signatures" to dwarf
> > and actually rely on dwarf locations to somehow recover parameters.
> 
> 
> That isn't possible, so far as I can see - DWARF only describes how to
> access the parameters from inside the function, not how they were passed on
> the call.

At the moment of entry to a function, those are the same thing.

> (eg: an unoptimized build will probably describe the parameters
> as being in stack memory even though the values were passed in registers -
> because the prologue moves the values from registers to memory, and the
> DWARF locations only have to be valid after the prologue)

That sounds like a quality of implementation issue.  Higher quality
compilers accurately note the location of parameters, as they change
within and after the prologue and other points in the function.  It
is a location LIST after all.


> > I actually want to ask how to identify whether the return type is
> > changed or not in dwarf.
> > [...]
> https://dwarfstd.org/issues/221105.1.html in DWARFv6 might help provide
> some of this information & is closer to an ABI-encoding than we have for
> parameters.

This will be lovely.


- FChE