Mailinglist Archive: zypp-devel (115 mails)

[Stanislav Visnovsky <visnov@xxxxxxx>]

Hi!

What is the reasoning behind using tribool? This seems to be the biggest 
headache for application developers so far...

Stano

Dňa Po 23. Júl 2007 17:25 Martin Vidner napísal:
> Michael is not subscribed, forwarding...
>
> ----- Forwarded message from Michael Meeks <michael.meeks@xxxxxxxxxx> -----
> Subject: tribool ...
> From: Michael Meeks <michael.meeks@xxxxxxxxxx>
> To: zypp-devel@xxxxxxxxxxxx
> Cc: Ricardo Cruz <rpmcruz@xxxxxxxxxxxxxxxxxxx>
> Organization: Novell, Inc.
> X-Mailer: Evolution 2.6.0
> Date: Mon, 23 Jul 2007 08:55:13 +0100
>
> Hi guys,
>
>         So - I learned something fun, interesting and new this morning (at the
> cost of an hour or so) - it was the joys of the tribool:
>
>         if (available_obj != NULL &&
>             !available_obj->repository().info().enabled()) {
>
>         In normal usage, by the power of sequence points, this sort of thing
> works nicely (and indeed, it used to work before the API re-hash I
> think).
>
>         Unfortunately, with a tri-bool around this becomes a mess, sequence
> points are no longer your friend:
>
>     return false
>         && !(zypp::ResObject::constPtr(NULL)->repository().info().enabled()) ? 1 :
> 0;
>
>         SEGV's immediately - try it as a return from main() ;-) it compiles to:
>
>     return tribool.operator&& (false, <segv'ing-code>)
>
>         Now the fix is 'obvious' - simply split it into 2 if's, or add a
> '(bool)' cast to what appears to be a bool anyway ;-) but it's pretty
> sickeningly non-obvious to the casual user. Also, will it survive a
> future maintainer's desire for cleaner code without "all these odd /
> redundant casts in it" ?
>
>         Urgh,
>
>                 Michael.
>
> --
>  michael.meeks@xxxxxxxxxx  <><, Pseudo Engineer, itinerant idiot
--
To unsubscribe, e-mail: zypp-devel+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: zypp-devel+help@xxxxxxxxxxxx