Re: preprocessor directive indentation, cppi

[Bruce Korb]

On 02/20/11 16:58, Bruno Haible wrote:
>> it lets you distinguish conditional and unconditional
>> #include directives and #define directives based solely
>> on presence or absence of spaces after the "#".
> 
> This heuristic does not always work:
>   - sleep.c:# define WIN32_LEAN_AND_MEAN
>     is semantically an unconditional define (it is defined on all platforms
>     where it matters), like in accept.c.
>   - tsearch.c:#define CHECK_TREE(a) check_tree(a)
>     is a conditional definition. This code comes from glibc.
> 
> So in practice, I have to grep for "# *define" anyway.

OK, I've got a couple of pennies for this, too:  if you want
a CPP directive indentation validator, it should constrain itself
to validating that all directives at a particular nesting level
are at the same indentation level and that that level is not less
than its containing level.  e.g., if there is only one nesting
level in a header, it is likely the double inclusion guard
and everything ought to be hunky-dory with no indentation.
*However*, if someone chooses to indent something with a space
between the hash and directive, then everything at that level
and deeper must be indented by that much and only more if it is
nested deeper, too.

I think that's the lowest common denominator and doesn't impose
anything that neither Bruno nor I would dislike.

These would be "okay" (indentation always >= containing "#if"):

> #if
> # include
> # if
> # define
> # endif
> #endif
> 
> #if
> # include
> #endif
> #if
> #include
> #endif

This would violate "not less at same level":

> #if
> # include
> #else
> #include
> #endif

because, basically, both "includes" are in the same construct,
though not the same section of the "#if".