RFR: 8154110: Update class* and safepoint* logging subsystems

Christian Thalinger christian.thalinger at oracle.com
Wed Apr 27 19:19:21 UTC 2016


> On Apr 27, 2016, at 2:19 AM, Coleen Phillimore <coleen.phillimore at oracle.com> wrote:
> 
> 
> 
> On 4/27/16 8:17 AM, David Holmes wrote:
>> 
>> 
>> On 27/04/2016 7:41 PM, Marcus Larsson wrote:
>>> 
>>> 
>>> On 04/27/2016 11:19 AM, David Holmes wrote:
>>>> On 27/04/2016 6:39 PM, Marcus Larsson wrote:
>>>>> Hi,
>>>>> 
>>>>> 
>>>>> On 04/27/2016 10:32 AM, Ioi Lam wrote:
>>>>>> 
>>>>>> 
>>>>>> On 4/26/16 6:13 PM, David Holmes wrote:
>>>>>>> On 27/04/2016 10:21 AM, Ioi Lam wrote:
>>>>>>>> The changes look good to me.
>>>>>>>> 
>>>>>>>> I think each line of the "constraints" prints only one item (?).
>>>>>>>> If so,
>>>>>>>> we should use the singular form. (Max, could you post a sample output
>>>>>>>> from -Xlog?)
>>>>>>> 
>>>>>>> The point is that it shows all the constraints.
>>>>>>> 
>>>>>> 
>>>>>> We currently have a mix-match of singular/plural forms in the log tags:
>>>>>> 
>>>>>> singular:
>>>>>>  LOG_TAG(alloc)
>>>>>>  LOG_TAG(region)
>>>>>> 
>>>>>> plural:
>>>>>>  LOG_TAG(defaultmethods)
>>>>>> 
>>>>>> Sometimes whether to use an (s) can be a personal preference. However,
>>>>>> the poor user would need to try 2^n times to get the right combination
>>>>>> of (s) or (no s), and that's too much.
>>>>> 
>>>>> Yeah, we should definitely be consistent with this. I vote for tags in
>>>>> singular, since it makes them ever so slightly shorter.
>>>> 
>>>> Hmmm, so data becomes datum? Or is that the exception to the rule? ;-)
>>> 
>>> Seems like a good exception to me. :) Perhaps the rule should be to
>>> avoid trailing s if it signifies plural.
>>> 
>>>> 
>>>> No matter what you choose some things will be unintuitively awkward:
>>>> defaultmethods, exceptions, constraints, all suit plural forms and are
>>>> awkward in singular. There are numerous plural forms in use already.
>>> 
>>> I don't see what distinguishes the tags you mention from the others. I
>>> guess I don't see what makes them awkward in singular. If I want
>>> defaultmethod logging I enable that tag. It's obvious I will get
>>> multiple log messages for it so I don't see the purpose of the 's'.
>> 
>> Maybe it is just familiarity with the old TraceXXX forms. I prefer the noun forms to be plural - as many of them are. Many of the ones that are not are a prefix taken from a longer form ie TraceThreadXXX became thread + xxx; whereas TraceExceptions became exceptions.
>> 
>> It is all somewhat arbitrary in my view.
> 
> LOL, we said the same thing.

I’d like to propose another approach:  constraintz 

It’s way cooler and rulezzzz!

> 
> Coleen
> 
>> 
>> David
>> 
>>> Thanks,
>>> Marcus
>>> 
>>>> 
>>>> David
>>>> 
>>>>> Thanks,
>>>>> Marcus
>>>>> 
>>>>>> 
>>>>>> - Ioi
>>>>>> 
>>>>>>> David
>>>>>>> 
>>>>>>>> Thanks
>>>>>>>> - Ioi
>>>>>>>> 
>>>>>>>> On 4/26/16 2:05 PM, Max Ockner wrote:
>>>>>>>>> New webrev: http://cr.openjdk.java.net/~mockner/8154110.hotspot.02/
>>>>>>>>> 
>>>>>>>>> I have responded to these suggestions:
>>>>>>>>> - class+loaderdata -> class+loader+data
>>>>>>>>> - class+loaderconstraints -> class+loader+constraints
>>>>>>>>> - I did not change constraints to constraint. I agree with Coleen's
>>>>>>>>> reasoning, but please let me know if you think there is a good
>>>>>>>>> reason
>>>>>>>>> to change it.
>>>>>>>>> 
>>>>>>>>> Thanks,
>>>>>>>>> Max
>>>>>>>>> 
>>>>>>>>> On 4/26/2016 4:05 PM, Coleen Phillimore wrote:
>>>>>>>>>> 
>>>>>>>>>> Hi,
>>>>>>>>>> 
>>>>>>>>>> Yes, his looks good, but I agree that class, loaderconstraints =>
>>>>>>>>>> class, loader, constraints
>>>>>>>>>> and class, loaderdata => class, loader, data
>>>>>>>>>> 
>>>>>>>>>> But I think constraints should be plural, we're logging the
>>>>>>>>>> constraint checking (more than one constraint) so plural makes more
>>>>>>>>>> sense to me.
>>>>>>>>>> 
>>>>>>>>>> Okay, I think I now prefer the combination of short words to do
>>>>>>>>>> logging after seeing this change.
>>>>>>>>>> 
>>>>>>>>>> Thanks,
>>>>>>>>>> Coleen
>>>>>>>>>> 
>>>>>>>>>> 
>>>>>>>>>> On 4/26/16 3:44 PM, Robbin Ehn wrote:
>>>>>>>>>>> Hi Max,
>>>>>>>>>>> 
>>>>>>>>>>> I was just looking at these tags and thought maybe they should be
>>>>>>>>>>> split, nice !
>>>>>>>>>>> 
>>>>>>>>>>> I have request for the ' constraints' tag, can make it singular
>>>>>>>>>>> (without ending 's') ?
>>>>>>>>>>> 
>>>>>>>>>>> (there are a few other ones in plural which we also should fix)
>>>>>>>>>>> 
>>>>>>>>>>> Otherwise I think this looks good!
>>>>>>>>>>> 
>>>>>>>>>>> Thanks!
>>>>>>>>>>> 
>>>>>>>>>>> /Robbin
>>>>>>>>>>> 
>>>>>>>>>>> On 04/26/2016 07:42 AM, Max Ockner wrote:
>>>>>>>>>>>> Hello,
>>>>>>>>>>>> 
>>>>>>>>>>>> This change factors the tags from the class and safepoint logging
>>>>>>>>>>>> subsystems into smaller tags, including "class" and "safepoint"
>>>>>>>>>>>> tags which are included in tag combination in their respective
>>>>>>>>>>>> subsystems.
>>>>>>>>>>>> 
>>>>>>>>>>>> classresolve -> class+resolve
>>>>>>>>>>>> classload -> class+load
>>>>>>>>>>>> classunload -> class+unload
>>>>>>>>>>>> classpath -> class+path
>>>>>>>>>>>> classloaderdata -> class+loaderdata
>>>>>>>>>>>> classload+constraints -> class+loaderconstraints
>>>>>>>>>>>> classinit -> class+init
>>>>>>>>>>>> classload+preorder -> class+preorder
>>>>>>>>>>>> 
>>>>>>>>>>>> safepointcleanup -> safepoint+cleanup
>>>>>>>>>>>> 
>>>>>>>>>>>> class+loaderdata can be further factored into class+load+data.
>>>>>>>>>>>> Same
>>>>>>>>>>>> with class+loaderconstraints.
>>>>>>>>>>>> 
>>>>>>>>>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8154110
>>>>>>>>>>>> webrev(hotspot):
>>>>>>>>>>>> http://cr.openjdk.java.net/~mockner/8154110.hotspot/
>>>>>>>>>>>> webrev(jdk): http://cr.openjdk.java.net/~mockner/8154110.jdk/
>>>>>>>>>>>> 
>>>>>>>>>>>> Tested with jtreg hotspot tests. jdk tests currently running.
>>>>>>>>>>>> 
>>>>>>>>>>>> Thanks,
>>>>>>>>>>>> Max



More information about the hotspot-dev mailing list