RFR [XS]: 8236709: struct SwitchRange in HS violates C++ One Definition Rule - was RE: struct SwitchRange and C++ One Definition Rule [-Wodr]

[David Holmes]

On 8/01/2020 6:10 pm, Baesken, Matthias wrote:
> Hello, thanks for the input .
> 
> I renamed SwitchRangeC1   to  C1SwitchRange  and  removed the unnecessary forward declaration .
> 
> New webrev :
> 
> http://cr.openjdk.java.net/~mbaesken/webrevs/8236709.1/

Seems fine.

Thanks,
David

> 
> Best regards, Matthias
> 
> 
> 
> 
>>>
>>> Hello, please review  this small fix for an issue  that I was running into  when
>> experimenting  with  gcc8  and   the -flto compiler flag .
>>> When building with those flags,  the gcc8 warns that  the  SwitchRange
>> classes  in HS code   violate the C++ One Definition Rule .
>>> So I renamed one of those 2  SwitchRange   classes .
>>>
>>> Bug/webrev :
>>>
>>> https://bugs.openjdk.java.net/browse/JDK-8236709
>>>
>>> http://cr.openjdk.java.net/~mbaesken/webrevs/8236709.0/
>>
>> I think I would prefer the C1 "namespace" disambiguator to be a prefix
>> rather than a suffix, to be consistent with usage in other parts of
>> HotSpot.  (GC code does that somewhat consistently, for example.)  I
>> don't recall seeing "namespace" suffixes used anywhere in HotSpot.
>>
>> I've occasionally considered proposing that C2 code be wrapped in a
>> namespace. It grabs a bunch of very generic global type names ("Type",
>> "Block", &etc; really!?) that can be annoying. (Though the diet for
>> precompiled.hpp helped with that.) But anything like that is probably
>> out of scope for the immediate problem.
>>
>> Giving C2 primacy of place (renaming the C1 class and letting C2 keep
>> the "good" name) seems consistent with existing practice.
>