[v3] PR libstdc++/42408

[Ed Smith-Rowland]

Ed Smith-Rowland wrote:
> Paolo Carlini wrote:
>> Hi,
>>
>> let's templatize that constructor and that seed member, which indeed are
>> templates in the draft Walter Brown is working on. The constraining
>> should be good enough, we can tweak it later, when the iter of the
>> de-conceptified specs proceeds. Tested x86_64-linux, committed to 
>> mainline.
>>
>> Paolo.
>>
>> /////////////////
>>   
> Heh, I was just about to send out this patch for this very thing ;-)
> These should have been templates even in concept I think.
> Ignore the patch I guess.
>
> The patch I had was for what I thought were defects.  Besides the 
> template Sseq ctors and seeds I think that shuffle_order_engine should 
> compare not just the engine but also the saves engine executions that 
> are .  The requirement is that a sequence of invocations of the 
> shuffle_order_engine should return the same results.  This requires 
> equality of the rest of the state I think.
This makes no sense.  Every time the engine is called, the base engine 
is called.  Unless the base engine is trivial its state must change.  
Comparing base engines is sufficient to ensure equality of the 
shuffle_order_engine.

I'll crawl back in my hole now.
>
> I'll make a patch for just that and a test case by tomorrow.
>
> Ed
>
Ed