Re: New name mangling in g++

[Jakub Jelinek <jakub at redhat dot com>]

On Mon, Jan 29, 2001 at 05:32:59PM +0100, Andreas Schwab wrote:
> Oliver Kellogg <oliver.kellogg@vs.dasa.de> writes:
> 
> |> Hello IA-64 C++ ABI experts,
> |> 
> |> I am trying to understand the name mangling scheme as 
> |> implemented in recent experimental snapshots of G++.
> |> 
> |> Given this code:
> |> 
> |> typedef struct {
> |>   int a;
> |> } stru;
> |> 
> |> class M {
> |>   public:
> |>     // [...]
> |>     void mixed (const stru &a, const stru b, stru &c, stru d,
> |>                 const stru &e, const stru f, stru &g, stru h);
> |> };
> |> 
> |> 
> |> the gcc-snapshot-20010126 mangles the method "mixed" as follows:
> |> 
> |>     _ZN1M5mixedERK4struS0_RS0_S0_S2_S0_S3_S0_
> |> 
> |> My questions are:
> |> What does the mangler treat as S1?
> |> Where is the `const' qualifier of argument b coded?
> 
> The `const' is not part of the interface.

const is part of the interface, it is encoded with the K.
I believe this is a bug in mangle.c because:
$ c++filt -s gnu-new-abi _ZN1M5mixedERK4struS0_RS0_S0_S2_S0_S3_S0_
M::mixed(const stru&, M::mixed, M::mixed&, M::mixed, const stru, M::mixed, const stru&, M::mixed)

I believe it should be mangled as
_ZN1M5mixedERK4struS2_RS1_S1_S3_S2_S4_S1_
, am just looking into mangle.c to see what's going on.

	Jakub