[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [aspectj-users] Compile-time signature

Iâve made some modifications - basically changed all the thisJoinPoint and thisEnclosingJoinPoint code generation for all join point kinds. These are in the 1.9.0.BUILD-SNAPSHOT build available from the repository:

<repository> <id>maven.springframework.org</id> <url>http://maven.springframework.org/snapshot</url> </repository>

or directly from here:


Because I consider it experimental you need to opt in to trying it out, the option is:

-Xajruntimetarget:1.9

(This option has been used previously to deliberately target an old runtime, basically just added 1.9 as a target here)

But if the build plugin you are using doesnât allow you to pass that through, you can set ASPECTJ_OPTS more globally as a system property/variable and it should be picked up:

export ASPECTJ_OPTS=â-Xajruntimetarget:1.9â

If you try it out, let me know how it goes.  Iâd be interested in the impact on startup time and on size of the built artifacts.

cheers,
Andy

On Feb 26, 2018, at 11:59 AM, Fabian Bergmark <fabian.bergmark@xxxxxxxxx> wrote:

Ok! Yes sure!

On Feb 26, 2018 20:31, "Andy Clement" <andrew.clement@xxxxxxxxx> wrote:
I did do some initial exploration work and it turned out to be bigger impact than Iâd hoped. Doesnât mean it wonât happen - just that itâll take longer. If I did get you a snapshot build with prototype code in it, would you be able to try it out and see if it behaves in your situation?

cheers,
Andy

> On Feb 26, 2018, at 4:41 AM, Fabian Bergmark <fabian.bergmark@xxxxxxxxx> wrote:
>
> Any progress?
>
> Cheers,
> Fabian
>
> 2018-01-10 19:48 GMT+01:00 Andy Clement <andrew.clement@xxxxxxxxx>:
>> The proper optimization here is to use the LDC byte code that takes a
>> constant class, rather than the string variant and then class loading. The
>> original AspectJ used the string form because LDC didn't take a class
>> originally.  Over the years a few places have been updated but not
>> everywhere ( thisEnclosingJoinPoint is one of them that hasn't been done I
>> think). Let me have a look at what I can do. For a while we had to tolerate
>> both modes because LDC class was quite new, but it has been there so long
>> now I think we can assume you are on a JVM that will have it.
>>
>> cheers,
>> Andy
>>
>>
>> On 10 January 2018 at 01:42, Fabian Bergmark <fabian.bergmark@xxxxxxxxx>
>> wrote:
>>>
>>> I'm using both AspectJ and Proguard in a project. On of my aspects
>>> inserts MDC-information in logging messages, including source location
>>> and method:
>>>
>>> void around() : call(void org.slf4j.Logger.trace(..)) || call(void
>>> org.slf4j.Logger.debug(..)) || call(void org.slf4j.Logger.info(..)) ||
>>> call(void org.slf4j.Logger.warn(..)) || call(void
>>> org.slf4j.Logger.error(..)) {
>>>        try (MDC.MDCCloseable c3 =
>>> MDC.putCloseable("logging_aspect::location",
>>> thisJoinPoint.getSourceLocation().toString());
>>>              MDC.MDCCloseable c4 =
>>> MDC.putCloseable("logging_aspect::method",
>>> thisEnclosingJoinPointStaticPart.getSignature().toString())) {
>>>            proceed();
>>>        }
>>>    }
>>>
>>> However, when enabling obfuscation in proguard, this results in log
>>> messages like:
>>>
>>> "mdc" : {
>>>    "logging_aspect::method" : "void
>>>
>>> java.lang.ClassNotFoundException.onCreateConnectionChannelResponse(ClassNotFoundException,
>>> CreateConnectionChannelError, ConnectionStatus)",
>>>    "logging_aspect::location" : "File.java:299"
>>> }
>>>
>>> I think there should be a way to access the Signature at the time of
>>> compilation. This would also be faster (?) than looking up classes
>>> during runtime.
>>> _______________________________________________
>>> aspectj-users mailing list
>>> aspectj-users@xxxxxxxxxxx
>>> To change your delivery options, retrieve your password, or unsubscribe
>>> from this list, visit
>>> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>>
>>
>>
>> _______________________________________________
>> aspectj-users mailing list
>> aspectj-users@xxxxxxxxxxx
>> To change your delivery options, retrieve your password, or unsubscribe from
>> this list, visit
>> https://dev.eclipse.org/mailman/listinfo/aspectj-users
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/aspectj-users

_______________________________________________
aspectj-users mailing list
aspectj-users@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/aspectj-users
_______________________________________________
aspectj-users mailing list
aspectj-users@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/aspectj-users