RFR (S): 8214791: Rename vm_operations_g1* files to g1VMOperations*
coleen.phillimore at oracle.com
coleen.phillimore at oracle.com
Wed Dec 5 15:21:31 UTC 2018
Thanks for asking!
On 12/5/18 10:05 AM, Thomas Schatzl wrote:
> Hi,
>
> On Wed, 2018-12-05 at 12:41 +0100, Per Liden wrote:
>> Hi,
>>
> [...]
>>>> Coleen and I talked about this, and neither of us wants yet
>>>> another naming convention for these files.
>>>> See her comment in the bug report for her
>>>> preference. (vmG1Operations). I don’t care that much
>>>> about the specifics, so long as there is consistency (which there
>>>> admittedly is not, at present, but
>>>> this proposed change just makes that worse).
>>>>
>>> I could rename the other gc files too to make them consistent, e.g.
>>> cmsVMOperations :)
>>>
>>> However I can be convinced to go with vmG1Operations although it
>>> does read backwards imho. At least it gets rid of the underscore.
>>>
>>> New webrev with that latter change:
>>> http://cr.openjdk.java.net/~tschatzl/8214791/webrev.1
>> I don't know... I would suggest we have either a prefix or a suffix.
>> The "midfix" (is that a word?) style is just strange. For example:
>>
>> runtime/vmOperations.hpp
>> gc/shared/gcVMOperations.hpp
>> gc/g1/g1VMOperations.hpp
>> gc/parallel/psVMOperations.hpp
>> gc/cms/cmsVMOperations.hpp
> I would prefer this (this is actually the original suggestion) and will
> do the work if everyone agrees on that. Widening to hotspot-runtime-dev
> in the hope that we can catch all that discussion there too then.
I'm okay with this convention too (despite what I wrote in the bug), if
they're all changed. I can guess that name when looking for it.
>> or maybe (to follow the vmStructs style):
>>
>> runtime/vmOperations.hpp
>> gc/shared/vmOperations_gc.hpp
>> gc/g1/vmOperations_g1.hpp
>> gc/parallel/vmOperations_ps.hpp
>> gc/cms/vmOperations_cms.hpp
I don't like underscores.
>>
>> But then I also don't think there is any real value in always having
>> VM_Operation (or other things like OopClosure) classes declared in
>> separate files, like we currently do. In some cases such classes
>> have very little in common and are better place in the context they
>> are used.
>> For example, in ZGC the VM_ZOperation class is an internal part of
>> the ZDriver, and not declared in a separate file.
I like the separate files. Then I can find them all and try to figure
out what the vm operation flags mean.
> I agree, I saw that and think it is nice(r) :) However e.g. the
> g1CollectedHeap.cpp file is big enough as it is, and the files are
> there already.
>
> It's just that every time I look at these names in the file browser I
> see these very strangely named files and think to myself that this
> should be fixed. Now here we are...
Me too! Thank you for fixing these!
Coleen
>
> Thanks,
> Thomas
>
>
More information about the hotspot-gc-dev
mailing list