Gary,

As far as I know the only outstanding issue is 
https://issues.apache.org/jira/browse/LOG4J2-2038 
<https://issues.apache.org/jira/browse/LOG4J2-2038>. I would point them to 
https://issuetracker.google.com/issues/70118537 
<https://issuetracker.google.com/issues/70118537> as well. Also, in 
https://issues.apache.org/jira/browse/LOG4J2-2133 
<https://issues.apache.org/jira/browse/LOG4J2-2133> Oleg said he was going to 
follow-up and see if Google had fixed the issue with the module-info.class 
files. He never did that.

FWIW, I pinged again on the thread at 
http://mail.openjdk.java.net/pipermail/core-libs-dev/2017-December/050382.html 
<http://mail.openjdk.java.net/pipermail/core-libs-dev/2017-December/050382.html>
 to see if I can get an answer on whether Google has implemented a fix to 
ignore files in META-INF/versions.

When you converted HttpClient to use Log4j 2 did you happen to use any API 
methods that SLF4J doesn’t support (for example, custom Messages)? If so I 
would mention that reverting will cause a loss of functionality.

My response would be that the tooling for Android should be catching up shortly 
which will allow the existing Log4j 2 API to work properly.

Ralph

> On Jan 8, 2018, at 10:56 AM, Gary Gregory <[email protected]> wrote:
> 
> Thank you both for replying. It sounds like for now we do not have clean
> Android story. I will reply later today to the HC thread with something
> like 'We plan on having an Android solution in the future" which means HC
> will likely switch to Slf4j for the next alpha/beta. Feel free to comment
> over at HC as well.
> 
> Gary
> 
> On Mon, Jan 8, 2018 at 9:26 AM, Ralph Goers <[email protected]>
> wrote:
> 
>> I have a couple of comments here.
>> 
>> One of the reasons I created Log4j using its own API was due to the
>> introduction of Messages. I had discussions on the SLF4J list that made it
>> clear that Ceki wasn’t interested in modifying the API to support anything
>> more than Strings. In fact, as a whole SLF4J is not very responsive to
>> people proposing changes. Just last week there were emails to the SLF4J
>> from someone begging that their PRs be committed, or at the very least be
>> reviewed. Although it would be great to only have a single Logging API I
>> hold out little hope that SLF4J will ever incorporate what we would need.
>> I have noticed that when people migrate from Log4j 1 to Log4j 2 they are
>> trying to figure out how to simply migrate their code from Log4j 1 syntax
>> to Log4j 2. They are not investigating how Log4j 2 works and trying how to
>> best implement the functionality they require.
>> Android is a problem. I think that is primarily because none of us develop
>> for it. The main reason we have support for OSGi, low GC overhead,
>> asynchronous loggers, etc. is because we have or had committers with a need
>> for those features. Although SLF4J seems to support Android my
>> understanding is that there is an SLF4J Android project that provides more
>> support. I think we just need to determine what the best way to support
>> Android is and implement it.
>> 
>> Ralph
>> 
>>> On Jan 8, 2018, at 9:01 AM, Matt Sicker <[email protected]> wrote:
>>> 
>>> I've noticed that one of the common complaints when migrating from log4j
>> 1
>>> to 2 (or similar) is the API to modify loggers/appenders/etc. at runtime.
>>> Despite efforts to fulfill the user needs without digging into internals,
>>> it seems like some frameworks or tools prefer convoluted access instead
>> of
>>> managing config files.
>>> 
>>> As for an API trim or similar to make Android and future Java use easier,
>>> that would really require a v3 API at this point, and I'd imagine any v3
>>> API we made would use Java 8 as a baseline which doesn't really solve the
>>> Android problem at all.
>>> 
>>> In practical usage, all my applications have to mangle logging
>> dependencies
>>> transitively regardless, so I've given up hope that there will ever be a
>>> simple solution. Sometimes I wonder if it would be more valuable to try
>> and
>>> port more of log4j-api to slf4j-api, or potentially unify the two facades
>>> one day.
>>> 
>>> On 8 January 2018 at 09:37, Gary Gregory <[email protected]> wrote:
>>> 
>>>> Hi All:
>>>> 
>>>> Over on the [email protected] list, we have a thread called "Using
>> SLF4J
>>>> as
>>>> a logging facade; Re: Disagreement on Log4J2" where Oleg (our lead and
>> main
>>>> author) has proposed moving from Log4j2 to SLF4J for the next HttpClient
>>>> Alpha.
>>>> 
>>>> (I had initially switched HC from Commons Logging to Log4j 2)
>>>> 
>>>> Android is the main contention. The size of the API is another but we
>>>> cannot do anything about that.
>>>> 
>>>> Your comments to at least provide direction and expectations would be
>> most
>>>> welcome.
>>>> 
>>>> Thank you,
>>>> Gary
>>>> 
>>> 
>>> 
>>> 
>>> --
>>> Matt Sicker <[email protected]>
>> 
>> 

Reply via email to