Well, as AbstractLogger is indeed Serializable, it looks like there's no way to remove that at this point. Making the MessageFactory Serializable sounds feasible. Are there any other components that may need to be serialized? If not, I can go ahead with the implementation.
On 15 October 2015 at 09:15, Mikael Ståldal <[email protected]> wrote: > I dislike to have to make a class Serializable just because some stupid > framework (or stupid use of some non-stupid framework) requires it. > > I guess it was a mistake to make org.apache.logging.log4j.spi. > AbstractLogger Serializable in the first place. > > On Thu, Oct 15, 2015 at 6:56 AM, Ralph Goers <[email protected]> > wrote: > >> Well, MessageFactory is not Serializable but AbstractMessageFactory is. >> If the MessageFactory used by the Logger is serializable we could include >> it. If it is not we would have to treat it as transient. Upon >> deserialization we may find that the MessageFactory implementation doesn’t >> exist on the target platform, in which case we would have to just use the >> default. >> >> Ralph >> >> On Oct 14, 2015, at 8:27 PM, Gary Gregory <[email protected]> wrote: >> >> What if a logger does not use the default message factory? >> >> Gary >> >> >> -------- Original message -------- >> From: Ralph Goers <[email protected]> >> Date: 10/14/2015 19:32 (GMT-08:00) >> To: Log4J Developers List <[email protected]> >> Subject: Re: Is there anything besides the Logger name that uniquely >> identifies a Logger? >> >> I am not sure why you would need or want to serialize any plugins. >> Logger basically references the LoggerContext and the PrivateConfig. Both >> of these should be transient as it makes very little sense for those to be >> deserialized on a target implementation. But even serializing the actual >> Logger makes very little sense. On the target system you would want to call >> LogManager.getLogger(name) to recreate it. >> >> What I would suggest is to use the Proxy pattern that is used by >> Log4jLogEvent to serialize and deserialize the Logger. What would be >> different is that the serialization would basically only serialize the >> logger name and deserialization would call LogManager.getLogger(name). >> >> Ralph >> >> On Oct 14, 2015, at 7:02 PM, Matt Sicker <[email protected]> wrote: >> >> Basically, to naively serialize a Logger, you need to serialize all the >> plugins associated with it. As most things in log4j-core can be classified >> as either plugins or "framework" code, that's really most of the codebase. >> >> On 14 October 2015 at 18:00, Gary Gregory <[email protected]> wrote: >> >>> If it's really 50%, then yeah, that's suspicious. I'd like to hear if >>> Ralph or Remko have any insights here. >>> >>> Gary >>> >>> On Wed, Oct 14, 2015 at 3:04 PM, Matt Sicker <[email protected]> wrote: >>> >>>> Most people use a static field to store the Logger, so most use cases >>>> don't require serialization. For instance fields, it might work better to >>>> declare it transient, and in that case, our implementation of Logger should >>>> not be Serializable at all. Otherwise, there are ways to serialize >>>> everything, but the way it looks, that will require making over 50% of the >>>> code base Serializable which doesn't smell right to me. >>>> >>>> On 14 October 2015 at 16:54, Gary Gregory <[email protected]> >>>> wrote: >>>> >>>>> It would be a neat trick to only use the logger name for ser/deser. >>>>> But a logger only exists in a LC, so how would you re-create the Logger >>>>> object. LogManager.getLogger(String) can't account for the message factory >>>>> for example. Would knowing the class within which the static Logger >>>>> resides >>>>> be enough to know which LC to use? I do not see how :-( I think we need >>>>> Ralph's insight here. >>>>> >>>>> The alternative would be... to recommend that all Logger declarations >>>>> be transient? That does not seen realistic, especially accounting for code >>>>> you cannot change. >>>>> >>>>> Gary >>>>> >>>>> On Wed, Oct 14, 2015 at 2:48 PM, Matt Sicker <[email protected]> wrote: >>>>> >>>>>> Perhaps besides a particular LoggerContext. I have an idea on how to >>>>>> significantly simplify the serialization of Logger, and if we can simply >>>>>> unserialize it based purely on its name, then that would save a lot of >>>>>> trouble. I don't remember if we've discussed this idea in the past, but I >>>>>> think this would be the best way to implement serialization in Logger. I >>>>>> wouldn't want to pass a Logger over the wire and clobber a possibly >>>>>> different configuration already in memory at the time, for instance. >>>>>> >>>>>> -- >>>>>> Matt Sicker <[email protected]> >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> E-Mail: [email protected] | [email protected] >>>>> Java Persistence with Hibernate, Second Edition >>>>> <http://www.manning.com/bauer3/> >>>>> JUnit in Action, Second Edition <http://www.manning.com/tahchiev/> >>>>> Spring Batch in Action <http://www.manning.com/templier/> >>>>> Blog: http://garygregory.wordpress.com >>>>> Home: http://garygregory.com/ >>>>> Tweet! http://twitter.com/GaryGregory >>>>> >>>> >>>> >>>> >>>> -- >>>> Matt Sicker <[email protected]> >>>> >>> >>> >>> >>> -- >>> E-Mail: [email protected] | [email protected] >>> Java Persistence with Hibernate, Second Edition >>> <http://www.manning.com/bauer3/> >>> JUnit in Action, Second Edition <http://www.manning.com/tahchiev/> >>> Spring Batch in Action <http://www.manning.com/templier/> >>> Blog: http://garygregory.wordpress.com >>> Home: http://garygregory.com/ >>> Tweet! http://twitter.com/GaryGregory >>> >> >> >> >> -- >> Matt Sicker <[email protected]> >> >> >> >> > > > -- > [image: MagineTV] > > *Mikael Ståldal* > Senior software developer > > *Magine TV* > [email protected] > Regeringsgatan 25 | 111 53 Stockholm, Sweden | www.magine.com > > Privileged and/or Confidential Information may be contained in this > message. If you are not the addressee indicated in this message > (or responsible for delivery of the message to such a person), you may not > copy or deliver this message to anyone. In such case, > you should destroy this message and kindly notify the sender by reply > email. > -- Matt Sicker <[email protected]>
