Re: C++ bug with PCC_STATIC_STRUCT_RETURN

[msokolov at ivan dot Harhan dot ORG (Michael Sokolov)]

I wrote:

> I think I have found a bug in the C++ front end in the code #ifdefed on
> PCC_STATIC_STRUCT_RETURN and thus used only on the great old holy original
> ports (VAX, m68k, etc.) but not on any of the "modern" ones.
>
> This fragment in cp/semantics.c:
>
> #ifdef PCC_STATIC_STRUCT_RETURN  
>   if (!AGGR_INIT_VIA_CTOR_P (aggr_init_expr) && aggregate_value_p (type))
>     {
>       int old_ac;
>
>       flag_access_control = 0;
>       call_expr = build_aggr_init (slot, call_expr, LOOKUP_ONLYCONVERTING);
>       flag_access_control = old_ac;
>       copy_from_buffer_p = 1;
>     }
> #endif
>
> throws cc1plus into an infinite recursion of walk_tree when compiling a call to
> a function returning a structure.

OK, I have investigated it further, and here is what I found. The C++ front end
initially builds an AGGR_INIT_EXPR for every call to a function returning a
structure. Then expand_body calls simplify_aggr_init_exprs_r via walk_tree to
convert all AGGR_INIT_EXPRs into CALL_EXPRs. simplify_aggr_init_exprs_r
contains the code fragment quoted above ifdefed on PCC_STATIC_STRUCT_RETURN.
There it calls build_aggr_init to copy the struct return value from the callee-
returned static buffer into its temporary. Apparently, however, build_aggr_init
generates an AGGR_INIT_EXPR again, causing infinite recursion.

This is the point where I give up. I have no idea how to fix this. Unless a
volunteer shows up who groks the C++ front end and cares about
PCC_STATIC_STRUCT_RETURN systems, C++ will remain completely broken on VAX
4.3BSD, as well as m68k UNIXes with native PCC and all other targets that
define PCC_STATIC_STRUCT_RETURN.

-- 
Michael Sokolov
Public Service Agent
International Engineering and Science Task Force

1351 VINE AVE APT 27		Phone: +1-714-738-5409
FULLERTON CA 92833-4291 USA	(home office)

E-mail: msokolov@ivan.Harhan.ORG (ARPA TCP/SMTP)