This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA 1/6] Add initial type alignment support
Hi Tom,
On 04/24/2018 04:22 PM, Tom Tromey wrote:
> +/* * Number of bits allocated for alignment. */
> +
> +#define TYPE_ALIGN_BITS 8
> +
> /* * A ``struct type'' describes a particular instance of a type, with
> some particular qualification. */
>
> @@ -831,6 +835,14 @@ struct type
>
> struct type *chain;
>
> + /* * The alignment for this type. Zero means that the alignment was
> + not specified in the debug info. Note that this is stored in a
> + funny way: as the log base 2 (plus 1) of the alignment; so a
> + value of 1 means the alignment is 1, and a value of 9 means the
> + alignment is 256. */
> +
> + unsigned align_log2 : TYPE_ALIGN_BITS;
> +
> /* * Flags specific to this instance of the type, indicating where
> on the ring we are.
>
> @@ -841,7 +853,7 @@ struct type
> instance flags are completely inherited from the target type. No
> qualifiers can be cleared by the typedef. See also
> check_typedef. */
> - int instance_flags;
> + unsigned instance_flags : 9;
>
> /* * Length of storage for a value of this type. The value is the
> expression in host bytes of what sizeof(type) would return. This
> @@ -1292,6 +1304,26 @@ extern void allocate_gnat_aux_type (struct type *);
> so you only have to call check_typedef once. Since allocate_value
> calls check_typedef, TYPE_LENGTH (VALUE_TYPE (X)) is safe. */
> #define TYPE_LENGTH(thistype) (thistype)->length
> +
> +/* * Return the alignment of the type in bytes, or 0 if no alignment
> + was specified. */
> +#define TYPE_RAW_ALIGN(thistype) type_raw_align (thistype)
> +
> +/* * Return the alignment of the type in bytes, or 0 if no alignment
> + was specified. */
> +extern unsigned type_raw_align (struct type *);
> +
> +/* * Return the alignment of the type in bytes. Return 0 if the
> + alignment cannot be determined; but note that this makes an effort
> + to compute the alignment even it it was not specified in the debug
> + info. */
> +extern unsigned type_align (struct type *);
> +
> +/* * Set the alignment of the type. The alignment must be a power of
> + 2. Returns false if the given value does not fit in the available
> + space in struct type. */
> +extern bool set_type_align (struct type *, ULONGEST);
Are these represented in host bytes, or target addressable memory units?
See comments around type::length and type_length_units. The comments above
should be clarified in that direction.
> +/* Implement the type_align gdbarch function. */
> +
> +static ULONGEST
> +i386_type_align (struct gdbarch *gdbarch, struct type *type)
> +{
> + type = check_typedef (type);
> +
> + if (gdbarch_ptr_bit (gdbarch) == 32 &&
> + TYPE_CODE (type) == TYPE_CODE_INT && TYPE_LENGTH (type) > 4)
> + return 4;
&& on the other line.
Otherwise looks good to me, but I'd like to hear Simon's opinion.
Thanks,
Pedro Alves