On Wed, Aug 13, 2014 at 3:55 PM, Rowan Collins <rowan.coll...@gmail.com>
wrote:
> Ferenc Kovacs wrote (on 13/08/2014):
>
>> I think you misunderstood that part, I was thinking about a state machine
>> like this:
>>
>> 1. function doStuff(){
>> 2. switch($state){
>> 3. case OPENDOOR:
>> 4. if(!opendoor()){
>> 5. break;
>> 6. }
>> 7. $state = SITDOWN;
>> 8. case SITDOWN:
>> 9. if(!sitdown()){
>> 10. break;
>> 11. }
>> 12. $state = SIPWHISKEY;
>> 13. case SIPWHISKEY:
>> 14. sipwhiskey();
>> 15. }
>> 16. }
>>
>>
>>
>> where you modify the switch variable in one case and fall through into
>> another.
>> as I mentioned I don't have a reasonable use-case for multiple defaults,
>> but I can see some for multiple case labels in general, and I don't think
>> that it is better to have different behavior for case and default in this
>> regard.
>>
>
>
> The assignments to $state in that code do not make any difference to
> execution, control simply flows forwards whenever there is no "break".
>
> $state is only tested once, to select the initial label to jump to, so no
> amount of reassigning it, or adding duplicate labels, can ever cause a
> second jump.
>
> I'm tempted to write a PHP script that emulates the switch execution by
> building a set of if and goto statements, which might make some of this
> behaviour clearer (not just for this discussion, but for other confusion
> I've seen elsewhere), but I don't have time right now.
>
yeah, and that is a documented behavior, I've just had a momentary
brain-fart here (which is funny considering that at one point I even wrote
a codesniffer rule for generating warnings for fall-through case statements
without an explicit comment).
thanks again for pointing that out. and sorry for the noise.
ps: I still hold onto my no vote, even thought it is a small BC break, it
is still a BC break, I wouldn't consider the current patch as a bugfix(and
even thought that the spec documented this as not a valid syntax hhvm still
supported it, ableit picking the first default when there are more than
one) , I would prefer having it E_DEPRECATED first in 5.next and we can
remove it in 7.0.
--
Ferenc Kovács
@Tyr43l - http://tyrael.hu
