[15] RFR: 8237894: CTW: C1 compilation fails with assert(x->type()->tag() == f->type()->tag()) failed: should have same type

[Aleksey Shipilev]

On 2/12/20 1:48 PM, Jamsheed C M wrote:
>> But now I have second thoughts. Do we foresee any performance troubles from not GVN'ing null
>> constants in C1, though?
> 
> i am not sure about all use case, i think only static finals are 
> affected, service loader based initialization is an use case probably, 
> but it is interface.

Ah, right. And since those are known nulls, access through them would also yield immediate NPE. I am
struggling to come up with a plausible counter-example that involves using the
known-null-object/array without tanking performance even before this patch, hm.

> these error never occurred in any other testing before , so use cases 
> were limited ?

This, unfortunately, does not tell us much. It looks like two constant nulls of different types meet
and that is how assert fails. But if we exclude all constant nulls, are we breaking the case where
two constant nulls of the same type meet. This would not fail the assert, yet have the effects on
performance. But as said above, this is probably not a problem here.

-- 
Thanks,
-Aleksey