-----"Shane Duan" <[EMAIL PROTECTED]> wrote: -----
FYI: The change to the build file is now in.
>I just thought from the earlier conversations that we are on 1.5.
>Sorry (On the other hand, the generics were just language level, you
>can always compile it for the target of 1.4. )
If we ever make this change, some warning will be nice; I am currently compiling with a 1.4 compiler so it falls over for me. There may be others out there doing the same.
>Removing the generics usage should be a lot better than simply revert
>all the changes? Just compile and fix.
The generics usage was only one of the problems; I made a big deal of it to remind everyone that we were still on 1.4 (I've accidentally used 1.5 APIs in the past, so you're not the only one).
>For me to reapply it now,
>I'll have to go through each file manually, revert the revert, and
>remove the change.
In my despotic opinion, the code isn't needed. Please feel free to ping me if you really believe that's not the case.
In any case, you can revert all the changes from an entire revision (r796 is my revert, which you could revert in turn). This is especially easy if you're using Windows and Tortoise.
>About the argument for scanning behaviours, I am sorry but are you
>saying that you want to use Ant or Maven classes, something that is
>more complicated and less tested,
The ant tasks are all fairly uncomplicated, TBH. I didn't write the original BehaviourRunners, just extended them for the story running and printing and found it fairly easy.
As for being less tested, I've used filesets in every Enterprise app I've worked on. There's no substitute for testing in the real world! The ant filesets have been tested in JBehave's build since the examples went into the main build targets, anyway.
>rather than use Cotta, something
>much smaller and designed just for directory and file operations, and
>fully tested? Like I said, Cotta works with class files, so it does
>not matter where your source it, and it works if the classes are in a
>jar file instead of directory.
I 'stole' the bit of code which allowed .java, .class and jarred files. It was extremely valuable, so thank you very much for that.
>And I did leave it out of the code, it is on test only.
Please remember that if we need something in order to use JBehave effectively, our users will need it too. It's even worse to provide a solution for ourselves, then keep it from the users, than it is to simply put it in the jar. And I'd rather not, at this stage, put Cotta in the jar.
If you want to talk to me more about this, please feel free to email off-list or ping me on Yahoo.
Cheers,
Liz.
--
Elizabeth Keogh
[EMAIL PROTECTED]
http://sirenian.livejournal.com
http://jbehave.org
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
FYI: The change to the build file is now in.
>I just thought from the earlier conversations that we are on 1.5.
>Sorry (On the other hand, the generics were just language level, you
>can always compile it for the target of 1.4. )
If we ever make this change, some warning will be nice; I am currently compiling with a 1.4 compiler so it falls over for me. There may be others out there doing the same.
>Removing the generics usage should be a lot better than simply revert
>all the changes? Just compile and fix.
The generics usage was only one of the problems; I made a big deal of it to remind everyone that we were still on 1.4 (I've accidentally used 1.5 APIs in the past, so you're not the only one).
>For me to reapply it now,
>I'll have to go through each file manually, revert the revert, and
>remove the change.
In my despotic opinion, the code isn't needed. Please feel free to ping me if you really believe that's not the case.
In any case, you can revert all the changes from an entire revision (r796 is my revert, which you could revert in turn). This is especially easy if you're using Windows and Tortoise.
>About the argument for scanning behaviours, I am sorry but are you
>saying that you want to use Ant or Maven classes, something that is
>more complicated and less tested,
The ant tasks are all fairly uncomplicated, TBH. I didn't write the original BehaviourRunners, just extended them for the story running and printing and found it fairly easy.
As for being less tested, I've used filesets in every Enterprise app I've worked on. There's no substitute for testing in the real world! The ant filesets have been tested in JBehave's build since the examples went into the main build targets, anyway.
>rather than use Cotta, something
>much smaller and designed just for directory and file operations, and
>fully tested? Like I said, Cotta works with class files, so it does
>not matter where your source it, and it works if the classes are in a
>jar file instead of directory.
I 'stole' the bit of code which allowed .java, .class and jarred files. It was extremely valuable, so thank you very much for that.
>And I did leave it out of the code, it is on test only.
Please remember that if we need something in order to use JBehave effectively, our users will need it too. It's even worse to provide a solution for ourselves, then keep it from the users, than it is to simply put it in the jar. And I'd rather not, at this stage, put Cotta in the jar.
If you want to talk to me more about this, please feel free to email off-list or ping me on Yahoo.
Cheers,
Liz.
--
Elizabeth Keogh
[EMAIL PROTECTED]
http://sirenian.livejournal.com
http://jbehave.org
