Re: [PATCH v3 1/6] ilog2: create truly constant version for sparse

[Martin Wilck <mwilck@suse.com>]

On Tue, 2018-04-17 at 17:07 -0700, Linus Torvalds wrote:
> On Tue, Apr 17, 2018 at 4:35 PM, Martin Wilck <mwilck@suse.com>
> wrote:
> > Sparse emits errors about ilog2() in array indices because of the
> > use of
> > __ilog2_32() and __ilog2_64(),
> 
> If sparse warns about it, then presumably gcc with -Wvla warns about
> it too?

No, it doesn't (gcc 7.3.0). -> https://paste.opensuse.org/27471594
It doesn't even warn on an expression like this:

  #define SIZE (1<<10)
  static int foo[ilog2(SIZE)];

sparse 0.5.2 doesn't warn about that either. It emits "error: bad
integer constant expression" only if ilog2 is used in an array
initializer, like this:

  #define SIZE (1<<10)
  #define SUBS (1<<5)
  static int foo [ilog2(SIZE)] = {
          [ilog2(SUBS)] = 0,
  };

So maybe I was wrong, and this is actually a false positive in sparse.

> So I suspect that what you'd want is
> 
>   #define ilog2(n) \
>         __builtin_choose_expr(__is_constexpr(n), \
>                 const_ilog2(n), \
>                 __builtin_choose_expr(sizeof(n) <= 4, \
>                         __ilog2_u32(n), \
>                         __ilog2_u64(n)))
> 
> or something. Hmm?

Do you want me to convert the patch to your approach anyway?
Or should I throw this away and report to sparse?

Regards and thanks,
Martin

PS: apologies to all recipients for the broken cc list in my post.
-- 
Dr. Martin Wilck <mwilck@suse.com>, Tel. +49 (0)911 74053 2107
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)