Mailing-List: contact dev-help@jmeter.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@jmeter.apache.org
MIME-Version: 1.0
In-Reply-To: <CAH9fUpZ-3s_tgZeaQwe1=ENEA1dYdBX9ZiVnpdMAeEFRrDE2LA@mail.gmail.com>
References: <CAH9fUpZd_-1oyLjBKJ=q_Bb_ZPfpmN0_NXiySCYv3p2wdCEsMw@mail.gmail.com>
 <CAH9fUpZNPXC1URCf2vQ6g39qh227E4wDWazx1hQt0StZNKVQ2w@mail.gmail.com>
 <CAOGo0VY5MkW1V2R92qQ7RSMtABCS4w=KWOTnHtkGNa262RzZdg@mail.gmail.com> <CAH9fUpZ-3s_tgZeaQwe1=ENEA1dYdBX9ZiVnpdMAeEFRrDE2LA@mail.gmail.com>
From: sebb <sebbaz@gmail.com>
Date: Wed, 22 Nov 2017 11:29:11 +0000
Message-ID: <CAOGo0VYD3GE0CTm+V6fWkk=gfzBu1+tTZ6-ksX6ZHeUi0FbYdA@mail.gmail.com>
Subject: Re: BUG 61567
To: dev@jmeter.apache.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
archived-at: Wed, 22 Nov 2017 11:29:23 -0000

On 21 November 2017 at 22:25, Philippe Mouawad
<philippe.mouawad@gmail.com> wrote:
> Hello,
> There is clearly a bug. You can try the attached project.

I think the bug is in the POM.
See Bugzilla

> The fix would be just to rename our jars by suffixing them with version

That's much better than changing the ClassFinder code.
However I'm not sure it's necessary, if the POM is set up correctly.
If one is using Maven, one should get all the JMeter jars from Maven or non=
e.

> Regards
>
> On Sun, Nov 19, 2017 at 12:22 AM, sebb <sebbaz@gmail.com> wrote:
>
>> The code is quite complicated, so I would be very wary of changing it.
>> And it's important not to include jars in the classpath that don't belon=
g.
>>
>> JMeter runs quite happily if the jars are named anything you like so
>> long as they are in the lib/ext directory or they are added to the
>> appropriate path.
>>
>> So I'm not sure that the Bug is valid.
>>
>> On 18 November 2017 at 22:42, Philippe Mouawad
>> <p.mouawad@ubik-ingenierie.com> wrote:
>> > Hello,
>> > Any thoughts on this ?
>> > Thanks
>> >
>> > On Sat, Nov 11, 2017 at 10:41 PM, Philippe Mouawad <
>> > p.mouawad@ubik-ingenierie.com> wrote:
>> >
>> >> Hello,
>> >> Analyzing bug 61567 , I end up thinking that what we're doing in the
>> below
>> >> method is not correct:
>> >>
>> >>    - ClassFinder#getClasspathMatches
>> >>
>> >> This method does the following:
>> >>
>> >>    - It takes a list of jars in search_path
>> >>    - It takes the classpath
>> >>    - For each entry in classpath , it tests if it ends with one of th=
e
>> >>    jar of the first list
>> >>    - If one match is found, the entry will be returned , otherwise it
>> >>    will be ignored
>> >>
>> >> So what is triggering the bug is the following, as search_path is emp=
ty,
>> >> only content of lib/ext is used.
>> >>
>> >> So first list consist of :
>> >>
>> >>    - ApacheJMeter_jms.jar
>> >>    ApacheJMeter_junit.jar
>> >>    ApacheJMeter_native.jar
>> >>    ApacheJMeter_ftp.jar
>> >>    ApacheJMeter_components.jar
>> >>    /data/jmeter/jmeters/apache-jmeter-3.3/lib/ext
>> >>    ApacheJMeter_ldap.jar
>> >>    ApacheJMeter_mail.jar
>> >>    ApacheJMeter_tcp.jar
>> >>    ApacheJMeter_http.jar
>> >>    ApacheJMeter_java.jar
>> >>    ApacheJMeter_core.jar
>> >>    ApacheJMeter_functions.jar
>> >>    ApacheJMeter_jdbc.jar
>> >>    ApacheJMeter_mongodb.jar
>> >>
>> >> Classpath consists of many entries and specifically , JMeter artifact=
s
>> end
>> >> with -3.3.jar
>> >>
>> >> None of the entries are added due to endsWith test failing, example:
>> >>
>> >> Testing if /MAVEN .m2/repository/org/apache/
>> jmeter/ApacheJMeter_functions/
>> >> 3.3/ApacheJMeter_functions-3.3.jar ends with ApacheJMeter_jms.jar *wi=
ll
>> >> return false*
>> >>
>> >> As a consequence, no function is found
>> >>
>> >>
>> >> *One fix would be in JMeter to suffix artifacts in bin and lib/ext wi=
th
>> >> the version number.*
>> >>
>> >> *But is this behaviour correct ?*
>> >>
>> >> --
>> >> Regards.
>> >> Philippe M.
>> >>
>> >
>> >
>> >
>> > --
>> > Cordialement.
>> > Philippe Mouawad.
>> > Ubik-Ing=C3=A9nierie
>> >
>> > UBIK LOAD PACK Web Site <http://www.ubikloadpack.com/>
>> >
>> > UBIK LOAD PACK on TWITTER <https://twitter.com/ubikloadpack>
>>
>
>
>
> --
> Cordialement.
> Philippe Mouawad.