Re: [Tracker] REVIEW: external-crawler branch, part4

[Martyn Russell <martyn lanedo com>]

On 08/08/14 09:54, Philip Van Hoof wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> c5d068279c9db7ac45e577b18b38d0bcb62b1d06
>
> +       guint external_crawler : 1; /* TRUE if we're being feed files
> +                                    * instead of discovering them
> +                                    * ourselves */
> +       guint mtime_checking : 1;   /* TRUE if mtime checks should be done
> +                                    * during initial crawling. */
> +       guint initial_crawling : 1; /* TRUE if initial crawling should be
> +                                    * done */
>
> What is this compiler and platform specific bitfield fscking all about?

That approach is used by a few of us on the team.

> What is wrong with gboolean here? I know we are in a financial crisis
> since 2008, but bits are not that expensive that we have to do crazy
> about it.
>
> It will mess up your alignment and padding and that makes future API
> changes more difficult. You'll need to know how the bit fields will
> happen on each architecture and compiler.

Well, yes and no. If your structure is created using GSLice, it handles 
memory alignment and colouring for you.

Regarding architecture, I agree, but we've had little complaints so far.

> I think this is one of those "No you C developer. No. You should not
> use bitfields in structs. You should use int like everybody else for
> booleans (or gboolean in GLib)."

I won't discuss this much, I understand there are arguments for and 
against, you mention many above.

> In the old code bitfields were also used for this. Equally wrong imho.
>
> In general for code that is not in ie. the kernel: you should probably
> not be using bitfields. Probably never.

I think that highly depends on the platform and product you're deploying.

> For networking code it's just wrong because of padding/alignment (use
> &, ~, << and >> instead to fiddle with bits).

And byte order presumably.

Still, as said, this is a bit of a non-issue.
Thanks Philip,

--
Regards,
Martyn

Founder & Director @ Lanedo GmbH.
http://www.linkedin.com/in/martynrussell