Re: kernel coding style for if ... else which cross #ifdef

["Vegard Nossum" <vegard.nossum@gmail.com>]

On Sat, May 24, 2008 at 5:45 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> Sam Ravnborg wrote:
>>
>> We should actually do as you intially suggested and alwyas
>> define CONFIG_FOO no matter if FOO is built-in or module.
>> Because we do only want to distingush between the two in rare cases.
>>
>> But that is a separate patch and lets not do the same
>> mistage with CFG_*
>>
>
> I think pretty strongly that CFG_ and CONFIG_ should be exactly parallel.
>  If you want to change the meaning of CONFIG_X in the presence of modules,
> then change CFG_X at the same time.  Making them have different meanings
> will just confuse anyone wanting to convert #ifdef CONFIG_ code into
> if(CFG_) code.
>
>> I cooked up following patch - but I have not test-build a kernel yet.
>> We may use CFG_* here and there and clash is not good.
>>
>
> I have to say I'm not very keen on the CFG_* prefix.  It doesn't have any
> inherent meaning and just looks like a redundant abbreviation of CONFIG_;
> something which actually expresses the notion that it's always a
> compile-time constant would be better.  Not that I have any particularly
> good alternatives: CONST_? CCONST_? CONFIG_X_VAL? KCONFIG_? KONFIG_?
> KCONST_?

Don't know if this is really my place, but I could not agree more with
your characterisation of the CFG_* prefix and I will make the
following suggestion:

Why not use all-lowercase config_* names? It seems elegant, and fits
in with the notion that these are to be used not as macros, but as
ordinary constants.

(The only disadvantage I can see is that they will stand out less. But
I don't know how great the disadvantage is.)

You could even go further and make them real constants, something
along the lines of:

enum config_value { no, yes, mod };

static const enum config_value config_lockdep_support = yes;

(I believe the "static const" will prevent emission of this symbol in
the object file -- though I am not certain.)

You can now also test for specific values, e.g. if
(config_scsi_wait_scan == mod), in addition to simply testing the
truth value.

That concludes my suggestion.


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036