I still have a major bug to fix in cal10n. Next, is a review of Carl's contributions in ch.qos.logback.*.net package. Once that is done, I'll look at new bug reports and pull requests. Thus, unfortunately I have not yet had a chance to look at your changes to AppenderTrackerImpl and co. In my defense, my todo list is ever expanding. Even when I spend the whole day working on logback, at the end of the day, there are more logback-related items on my todo list than the day started with. Not to rant but bug reports and pull requests are generated at a rate faster than I can cope with them. As jira.qos.ch states, popular issues, those with most votes and highest activity, get treated first. On 13.04.2013 10:24, David Roussel wrote:
Ceki,
Would you be interested in accepting this change? Is there anything else you feel needs to be done?
I'd be interested in seeing it land in the next version.
David
On 13 Apr 2013, at 01:01, "Becker, Thomas" <Thomas.Becker@netapp.com> wrote:
There was already a Map there. To your point though, that bit of rework to AppenderTrackerImpl was not strictly necessary but I felt the existing code, which essentially re-implemented LinkedHashMap in a clumsy way was pretty low hanging fruit. I'd be willing to make additional changes if the maintainer showed interest in accepting the patch, but my immediate need has passed. My observation has been that the majority of the pull requests to Logback seem to get no comments at all. But I went ahead and reopened this one here: https://github.com/qos-ch/logback/pull/107
-Tommy
________________________________________ From: Logback-user [logback-user-bounces@qos.ch] on behalf of diroussel [nabble@diroussel.xsmail.com] Sent: Friday, April 12, 2013 12:12 PM To: logback-user@qos.ch Subject: Re: [logback-user] SiftingAppender and RollingFileAppender, roll on close.
Thomas,
That functionality looks good.
In order to get accepted, maybe you can make your changes to AppenderTrackerImpl less invasive?
In AppenderTrackerImpl, you are subclassing LinkedHashMap instead of using ch.qos.logback.classic.pattern.LRUCache. But really there is no need for a map. Just keep the existing head and tail references, keep a count variablle. If the count gets to high, then remove entries from the tail.
I like that you've added unit tests and the code formatting is consistent with the rest of the code base.
David
-- Ceki 65% of statistics are made up on the spot