This is the mail archive of the mailing list for the GDB project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH 2/8] Add new gdbarch method, unconditional_branch_address

* Kevin Buettner <> [2015-08-19 00:00:02 -0700]:

> diff --git a/gdb/gdbarch.h b/gdb/gdbarch.h
> index c1e2c1a..1770960 100644
> --- a/gdb/gdbarch.h
> +++ b/gdb/gdbarch.h
> @@ -924,7 +924,7 @@ extern void set_gdbarch_max_insn_length (struct gdbarch *gdbarch, ULONGEST max_i
>     If your architecture doesn't need to adjust instructions before
>     single-stepping them, consider using simple_displaced_step_copy_insn
>     here.
> -
> +  
>     If the instruction cannot execute out of line, return NULL.  The
>     core falls back to stepping past the instruction in-line instead in
>     that case. */
> @@ -1478,6 +1478,16 @@ typedef int (gdbarch_addressable_memory_unit_size_ftype) (struct gdbarch *gdbarc
>  extern int gdbarch_addressable_memory_unit_size (struct gdbarch *gdbarch);
>  extern void set_gdbarch_addressable_memory_unit_size (struct gdbarch *gdbarch, gdbarch_addressable_memory_unit_size_ftype *addressable_memory_unit_size);
> +/* Examine instruction at PC.  If instruction at PC is an unconditional
> +   branch, return the address to which control is transferred when the
> +   branch is taken.  Return 0 when this method is not implemented by
> +   architecture, PC refers to an invalid address, or instruction at PC
> +   is not an unconditional branch. */
> +
> +typedef CORE_ADDR (gdbarch_unconditional_branch_address_ftype) (struct gdbarch *gdbarch, CORE_ADDR pc);
> +extern CORE_ADDR gdbarch_unconditional_branch_address (struct gdbarch *gdbarch, CORE_ADDR pc);
> +extern void set_gdbarch_unconditional_branch_address (struct gdbarch *gdbarch, gdbarch_unconditional_branch_address_ftype *unconditional_branch_address);

Personally I'm not a fan of overloading the return values on these
functions, especially when the return value is an address, on some
targets 0 is a valid address.  I know there are lots of other places
where we use 0 as a special address in gdb, so this would be just one
more... but...

How would you feel about changing the function so that it returned a
bool and placed the address into a CORE_ADDRESS passed by pointer?

Just a thought,

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]