You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geronimo.apache.org by Jeremy Boynes <je...@coredevelopers.net> on 2003/08/13 21:49:45 UTC
[test] maven unit tests
I just added the unit tests for Maas van den Berg's JSR-88 impl and they do
not seem to be run. Could this be because they are in a javax.* package?
Anyone have any other ideas?
Also, when I ran maven --debug it seemed that every unit test was run in a
separate shell. Is this true, deliberate, and if so desirable given we
should end up with a *lot* of these?
--
Jeremy
RE: [test] maven unit tests
Posted by Jeremy Boynes <je...@coredevelopers.net>.
Duh - I couldn't see it for looking either - will fix now.
> -----Original Message-----
> From: Maas van den Berg [mailto:email@dds.nl]
> Sent: Wednesday, August 13, 2003 1:11 PM
> To: geronimo-dev@incubator.apache.org
> Subject: Re: [test] maven unit tests
>
>
> Jeremy,
>
> Oops, I used the naming convention from work. All test classes are
> prefixed with Test instead of postfixed with Test. Sorry about that.
>
> Maas
Re: [test] maven unit tests
Posted by Maas van den Berg <em...@dds.nl>.
Jeremy,
Oops, I used the naming convention from work. All test classes are
prefixed with Test instead of postfixed with Test. Sorry about that.
Maas
On Wed, 2003-08-13 at 21:49, Jeremy Boynes wrote:
> I just added the unit tests for Maas van den Berg's JSR-88 impl and they do
> not seem to be run. Could this be because they are in a javax.* package?
> Anyone have any other ideas?
>
> Also, when I ran maven --debug it seemed that every unit test was run in a
> separate shell. Is this true, deliberate, and if so desirable given we
> should end up with a *lot* of these?
>
> --
> Jeremy
>
Re: [test] maven unit tests
Posted by Maas van den Berg <em...@dds.nl>.
Jeremy,
Every unit test running in a freshly forked JVM makes sense if you think
about 'static' resources.
Maas
On Wed, 2003-08-13 at 21:49, Jeremy Boynes wrote:
> I just added the unit tests for Maas van den Berg's JSR-88 impl and they do
> not seem to be run. Could this be because they are in a javax.* package?
> Anyone have any other ideas?
>
> Also, when I ran maven --debug it seemed that every unit test was run in a
> separate shell. Is this true, deliberate, and if so desirable given we
> should end up with a *lot* of these?
>
> --
> Jeremy
>
RE: [test] maven unit tests
Posted by Jeremy Boynes <je...@coredevelopers.net>.
> > Would it be better to set it to false so we deal with these from the
> > outset?
>
> Not IMHO. Unless you fancy debugging class loader issues in maven /
> junit / ant / jaxp which has nothing whatsoever to do with Geronimo.
> Note we're talking about class loader issues of actually running the
> Ant JUnit tasks inside Maven here - not Geronimo based class loading
> problems.
>
> After suffering various class loader issues in the past (especially
> about using XML inside a JUnit test case) I always set this flag to
> true. Though be my guest if you wanna turn it off - just be warned &
> get ready to flip it back on later.
>
It a compromise to run one junit task in a separate VM so there is one fork
not a bizillion? This would isolate us from maven/ant issues leaving us with
just junit/geronimo ones (which we ought to fix if this thing is going to be
embeddable).
--
Jeremy
Re: [test] maven unit tests
Posted by James Strachan <ja...@yahoo.co.uk>.
On Wednesday, August 13, 2003, at 09:53 pm, Jeremy Boynes wrote:
>> FWIW right now we're using maven.test.fork=true (in
>> etc/project.properties) to spawn new JVMs when running the unit tests
>> -
>> this fixes all kinds of classpath issues you often get with unit
>> tests,
>> JUnit, Ant & XML parsers etc.
>>
>
> Would it be better to set it to false so we deal with these from the
> outset?
Not IMHO. Unless you fancy debugging class loader issues in maven /
junit / ant / jaxp which has nothing whatsoever to do with Geronimo.
Note we're talking about class loader issues of actually running the
Ant JUnit tasks inside Maven here - not Geronimo based class loading
problems.
After suffering various class loader issues in the past (especially
about using XML inside a JUnit test case) I always set this flag to
true. Though be my guest if you wanna turn it off - just be warned &
get ready to flip it back on later.
James
-------
http://radio.weblogs.com/0112098/
RE: [test] maven unit tests
Posted by Jeremy Boynes <je...@coredevelopers.net>.
> FWIW right now we're using maven.test.fork=true (in
> etc/project.properties) to spawn new JVMs when running the unit tests -
> this fixes all kinds of classpath issues you often get with unit tests,
> JUnit, Ant & XML parsers etc.
>
Would it be better to set it to false so we deal with these from the outset?
Re: [test] maven unit tests
Posted by James Strachan <ja...@yahoo.co.uk>.
On Wednesday, August 13, 2003, at 08:49 pm, Jeremy Boynes wrote:
> Also, when I ran maven --debug it seemed that every unit test was run
> in a
> separate shell. Is this true, deliberate, and if so desirable given we
> should end up with a *lot* of these?
FWIW right now we're using maven.test.fork=true (in
etc/project.properties) to spawn new JVMs when running the unit tests -
this fixes all kinds of classpath issues you often get with unit tests,
JUnit, Ant & XML parsers etc.
James
-------
http://radio.weblogs.com/0112098/