JDK-8160748: inconsistent types for ideal_reg

[Vladimir Kozlov]

Thank you, Kim

I set 8160748 Fix Version to jdk 10. Agree to do it in then.

Thanks,
Vladimir

On 7/5/16 4:32 PM, Kim Barrett wrote:
> [re-replying on the hotspot-compiler-dev list, as my original reply was in a related hotspot-dev thread.]
>
>> On Jul 3, 2016, at 10:54 AM, Yasumasa Suenaga <yasuenag at gmail.com> wrote:
>>
>> Hi all,
>>
>> We've worked for JDK-8160353 to resolve narrowing conversion error at GCC 6.
>> ------------
>> /home/ysuenaga/OpenJDK/hs/hotspot/src/share/vm/opto/type.cpp:101:1: error: narrowing conversion of '(uint)Node::NotAMachineReg' from 'uint {aka unsigned int}' to 'int' inside { } [-Werror=narrowing]
>> };
>> ^
>> /home/ysuenaga/OpenJDK/hs/hotspot/src/share/vm/opto/type.cpp:101:1: error: narrowing conversion of '(uint)Node::NotAMachineReg' from 'uint {aka unsigned int}' to 'int' inside { } [-Werror=narrowing]
>> /home/ysuenaga/OpenJDK/hs/hotspot/src/share/vm/opto/type.cpp:101:1: error: narrowing conversion of '(uint)Node::NotAMachineReg' from 'uint {aka unsigned int}' to 'int' inside { } [-Werror=narrowing]
>> cc1plus: all warnings being treated as errors
>> ------------
>>
>> Opcodes is defined as enum.
>> However, a part of code (e.g. NotAMachineReg) is defined as uint.
>>
>> I think Opcodes related code should consist uint type.
>>
>>  http://cr.openjdk.java.net/~ysuenaga/JDK-8160748/webrev.00/
>>
>> I've uploaded webrev for this issue.
>> I fixed Opcodes type to uint at first, and fixed many errors with GCC 6.
>>
>> I want to discuss about this.
>>
>>
>> Thanks,
>>
>> Yasumasa
>
> I’m not going to review that, but I took a very brief look and:
>
> ---------------
> src/share/vm/opto/opcodes.hpp
>  30 enum Opcodes : uint {
>
> This is not valid C++98 code.  It is using a C++11 feature, e.g. specifying the underlying type
> of an enum.
>
> ---------------
>
> My intent in separating JDK-8160748 from JDK-8160353 was to allow the remainder of the latter
> to be fixed now, and JDK-8160748  to be dealt with later, probably not in JDK 9 at all, since the
> implicit narrowing is not a problem with C++98.
>