Re: [PATCH x86/mm 08/11] x86 ia32 ptrace getreg/putreg merge

[Ingo Molnar <mingo@elte.hu>]

* H. Peter Anvin <hpa@zytor.com> wrote:

> Christoph Hellwig wrote:
>> On Thu, Nov 29, 2007 at 04:00:31AM -0800, Roland McGrath wrote:
>>> +#define R32(l,q)							\
>>> +	case offsetof(struct user32, regs.l):				\
>>> +		regs->q = value; break
>>> +
>>> +#define SEG32(rs)							\
>>> +	case offsetof(struct user32, regs.rs):				\
>>> +		return set_segment_reg(child,				\
>>> +				       offsetof(struct user_regs_struct, rs), \
>>> +				       value);				\
>>> +		break
>>
>> The code would be a lot more readable if you just opencoded this in the
>> caller instead of these obsfucated macros.
>
> For better or worse, though, that is style of existing code.  I 
> personally found it easy to deal with when I went through the code 
> recently.

yep. The ptrace code has certainly lots of inconsistent style crap piled 
up during its 10 year history of only be touched with a 10 foot pole. 
I'd go for small patches that continuously improve the picture within 
the existing mechanisms than any "100% pure required" approach. With 48 
clean patches from Roland we are already on the right granularity level 
i think.

See how ptrace.c raw code quality has already increased leaps and 
bounds:

                                       errors   lines of code   errors/KLOC
 [before]
 arch/x86/kernel/ptrace_64.c               58             621          93.3
 arch/x86/kernel/ptrace_32.c               39             717          54.3

 [after]
 arch/x86/kernel/ptrace.c                  13            1135          11.4

so i'm not worried about that aspect. We are definitely "for the 
better".

	Ingo