[TYPO3-core] RFC #10454: t3lib_div::makeInstanceClassName() allows to bypass checks for t3lib_Singleton

[Jeff Segars]

Ingo Renner wrote:
> Jeff Segars wrote:
> 
> Hi Jeff,
> 
>>> well to me this particular exception is pretty clear: language 
>>> actually doesn't have a constructor - at least not an explicit one...
>>>
>>> If the message solely resulted from that, than I think it's ok. Could 
>>> you test it with a different class again, a class that has a 
>>> constructor, but doesn't expect arguments...
>>
>> I agree the message is clear; my only (very small) problem is that it 
>> differs from happens with the when I create an object directly with 
>> "new" and do not get a fatal error.
>>
>> After some more testing, the fatal error only occurs when there's not 
>> a constructor at all.  If there is a constructor with no arguments but 
>> arguments are unexpectedly passed to the constructor, it works fine.
> 
> So this is a rather rare case as it seems to me. As 
> t3lib_div::makeInstance() is a quit central method and Reflection isn't 
> the fastest I'd like to suggest to ignore these cases for now. We can 
> (and of course should) address this issue when it really becomes an 
> often occuring error. I'd also argue that this is a bug with reflection 
> if the behavior is different from just using new.
> 
> I'll attach your patch to the BT issue, so that it doesn't get lost in 
> case we need it.
> 
> 
> all the best
> Ingo
> 

Fair enough :) I had the same concerns about speed so like you said, we 
can always add the additional checks if it proves to be a problem.

Thanks,
Jeff