Re: [Prec/RFA] Remove macros for record size in i386-linux-tdep.c

[Pedro Alves <pedro@codesourcery.com>]

A Saturday 23 May 2009 13:12:54, Hui Zhu escreveu:
> On Sat, May 23, 2009 at 19:59, Hui Zhu <teawater@gmail.com> wrote:
> > On Sat, May 23, 2009 at 19:12, Pedro Alves <pedro@codesourcery.com> wrote:
> >> On Saturday 23 May 2009 05:00:22, Hui Zhu wrote:
> >>> 2009-05-23  Hui Zhu  <teawater@gmail.com>
> >>>
> >>>         * i386-linux-tdep.c (I386_LINUX_RECORD_SIZE_*,
> >>>         I386_LINUX_RECORD_IOCTL_*,
> >>>         I386_LINUX_RECORD_FCNTL_*) Removed.
> >>>         (i386_linux_init_abi): Change size of record from macros to
> >>>         numbers.
> >>
> >> Okay.
> >>
> >>>         * i386-linux-tdep.c (i386_linux_init_abi): Change
> >>>         i386_linux_record_tdep.size_char from 8 to 1.
> >>
> >> You didn't mention this, so I take it it was going to be a separate patch.
> >> Why do you need size_char at all?  size of char is always 1 by definition.
> >> Does the kernel define this as variable somewhere also?
> >
> > OK.  I make a new patch remove it.
> > Please help me review it.
> >
> > Thanks,
> > Hui
> >
> > 2009-05-23  Hui Zhu  <teawater@gmail.com>
> >
> >        * i386-linux-tdep.c (I386_LINUX_RECORD_SIZE_*,
> >        I386_LINUX_RECORD_IOCTL_*,
> >        I386_LINUX_RECORD_FCNTL_*) Removed.
> >        (i386_linux_init_abi): Change size of record from macros to
> >        numbers.
> >
> >        * i386-linux-tdep.c (i386_linux_init_abi): Remove size_char.
> >        * linux-record.c (record_linux_system_call): Ditto.
> >        * linux-record.h (linux_record_tdep): Ditto.

This is OK, although the changelog wants to record different unrelated
changes: either commit this in two parts, or merge the (i386_linux_init_abi)
entries.

> About char part.  I am not very sure about it.  This is from gdb:
> /* Number of bits in a char or unsigned char for the target machine.
>    Just like CHAR_BIT in <limits.h> but describes the target machine.  */
> #if !defined (TARGET_CHAR_BIT)
> #define TARGET_CHAR_BIT 8
> #endif
> 
> So, maybe we can keep it.

No, this is bits in char, while size_char wants bytes per char,
which is always 1.  Even if we were to make GDB behave
for TARGET_CHAR_BIT != 8, linux-record would not be the place to
store that info.

-- 
Pedro Alves