Mailing-List: contact issues-help@karaf.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@karaf.apache.org
Date: Mon, 15 Dec 2014 05:26:13 +0000 (UTC)
From: =?utf-8?Q?Jean-Baptiste_Onofr=C3=A9_=28JIRA=29?= <jira@apache.org>
To: issues@karaf.apache.org
Message-ID: <JIRA.12672983.1381308648000.21417.1418621173266@Atlassian.JIRA>
In-Reply-To: <JIRA.12672983.1381308648000@Atlassian.JIRA>
References: <JIRA.12672983.1381308648000@Atlassian.JIRA>
 <JIRA.12672983.1381308648753@arcas>
Subject: [jira] [Updated] (KARAF-2508) make osgi:find-class a more useful
 command
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable


     [ https://issues.apache.org/jira/browse/KARAF-2508?page=3Dcom.atlassia=
n.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jean-Baptiste Onofr=C3=A9 updated KARAF-2508:
----------------------------------------
    Fix Version/s:     (was: 2.4.1)
                   2.4.2

> make osgi:find-class a more useful command
> ------------------------------------------
>
>                 Key: KARAF-2508
>                 URL: https://issues.apache.org/jira/browse/KARAF-2508
>             Project: Karaf
>          Issue Type: Improvement
>          Components: karaf-shell
>    Affects Versions: 2.3.3
>            Reporter: Amichai Rothman
>            Assignee: Jean-Baptiste Onofr=C3=A9
>            Priority: Minor
>             Fix For: 4.0.0, 3.0.3, 2.4.2
>
>
> I recently had to track down which bundle contains a certain class, and f=
ound the find-class command wanting. A few simple improvements would make i=
t much more useful:
> 1. Add an --exact/-e option which doesn't automatically add asterisks at =
both ends of the string. In my case I was looking for the javax.jms.Connect=
ion class, and since the search is only on the last part of the name, this =
returned all classes in all packages which have the word Connection in thei=
r name, which added up to hundreds of results. with --exact I could look fo=
r what I want exactly, or choose to add asterisks at one or both ends expli=
citly in the search string if I wanted to.
> 2. Add a --local/-l option which adds BundleWiring.LISTRESOURCES_LOCAL to=
 the listResources method call bitmask. With the current implementation, it=
 returns both bundles containing the given class name and classes that impo=
rt it from other bundles, which is useful in some cases. But with the --loc=
al option it could show only the bundles that actually contain the class (w=
hether exported or not, unlike the exports command), which is far more usef=
ul in tracking down some classloading issues.
> 3. Add the bundle ID (at least) on the beginning of every printed line. T=
his makes it amenable to grepping (without it a grep loses all bundle infor=
mation, which is the whole point of using this command).
> 4. As a somewhat more complex alternative to #3, it would be nice if you =
could specify a fully qualified class name and not only the last part of th=
e name. The command can split the search string around periods, pass the la=
st part to listResources, then internally grep/match the full name. That wo=
uld make it truly flexible in finding classes. Asterisks would have to be t=
reated accordingly.
> I think adding these options (#1-#3 are trivial implementations), or at l=
east some of them, would make the command extremely useful when performing =
classloading diagnosis.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)