Re: [RFA] MIPS16 FP manual call/return fixes

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Tue, 08 May 2012 16:29:44 +0200, Maciej W. Rozycki wrote:
>  I have decided to add a related_address field to bp_location; this could 
> get converted to an opaque member in the future in case some other 
> breakpoints need to carry other auxiliary information

[ Mostly unrelated to this patch:  ]

This should be C++ inherited class field; or in C it should be
bp_location-type driven union.

While I would find it cleaner with the union I do not request this as
mandatory for check-in, it could also start some new GDB coding style
discussions.


> (just as we do with some other structures).

[ Mostly unrelated to this patch:  ]

Unfortunately we do but we do not / no longer have to.


>  I have decided to keep the assertion you had concerns about after all.  

I meant that 'gdb_assert (is_gnu_ifunc);', that is gone now.


>  	      b->type = bp_gnu_ifunc_resolver;
[...]
> Is it ever possible for this breakpoint to be updated later on with
> additional locations?

I am not aware of.


> --- gdb-fsf-trunk-quilt.orig/gdb/elfread.c	2012-05-08 08:25:37.795639727 +0100
> +++ gdb-fsf-trunk-quilt/gdb/elfread.c	2012-05-08 09:42:32.075560588 +0100
> @@ -1020,21 +1020,15 @@ elf_gnu_ifunc_resolver_return_stop (stru
[...]
> +  CORE_ADDR func_func_addr;

This variable is unused now.


[...]
> --- gdb-fsf-trunk-quilt.orig/gdb/breakpoint.c	2012-05-04 19:26:24.845606346 +0100
> +++ gdb-fsf-trunk-quilt/gdb/breakpoint.c	2012-05-08 08:26:49.855603092 +0100
[...]
> @@ -6609,6 +6610,9 @@ set_breakpoint_location_function (struct
>  	      /* Create only the whole new breakpoint of this type but do not
>  		 mess more complicated breakpoints with multiple locations.  */
>  	      b->type = bp_gnu_ifunc_resolver;

Empty line before a comment according to GDB Coding Standards.

> +	      /* Remember the resolver's address for use by the return
> +	         breakpoint.  */
> +	      loc->related_address = func_addr;
>  	    }
>  	}
>  
[...]


OK for the IFUNC part from me.


Thanks,
Jan