future plans for mauve

future plans for mauve

[Noa Resare]

Hello

While using mauve for a while now I have felt an urge to put some time
into change/fix some things that I think could be done better. This will
not happen right now, as I would like to see kaffe-1.1.5 out the door
first and I'm going offline for a week tonight but I thought it would be
a good idea to share my ideas with the list anyway, especially since
David Gilbert seems to be thinking about using a mauve for testing other
code than the class library.

Note that these are just random thoughts, and I'm really open to
suggestions.


* The './configure; make tests ARGS' invocation scheme is a little odd.
I'd like to have a small script perhaps called 'mauve' that took
arguments --java=/path/to/java --javac=/path/to/javac. The rest of the
args should be like the KEYS variable.

* Changes in mauve-KEYWORD files should trigger a recalculation of which
classes should be tested.

* The mechanism now implemented in the choose script should be faster
and written in a language more suited for the task. java perhaps?

* Improved documentation. There should be a 'write a new test guide' of
some sort. 

* Package structure. The part that could be called the 'testlet api'
should live in gnu.testlet. The test runner implementation should be in
gnu.mauve.runner (?) and test hierarchy perhaps in gnu.mauve.tests.

* Improved progress reporting when doing the tests.

/noa

Re: future plans for mauve

[Michael Koch]

Am Freitag, 29. Oktober 2004 08:41 schrieb Noa Resare:
> Hello
>
> While using mauve for a while now I have felt an urge to put some
> time into change/fix some things that I think could be done better.
> This will not happen right now, as I would like to see kaffe-1.1.5
> out the door first and I'm going offline for a week tonight but I
> thought it would be a good idea to share my ideas with the list
> anyway, especially since David Gilbert seems to be thinking about
> using a mauve for testing other code than the class library.
>
> Note that these are just random thoughts, and I'm really open to
> suggestions.
>
>
> * The './configure; make tests ARGS' invocation scheme is a little
> odd. I'd like to have a small script perhaps called 'mauve' that
> took arguments --java=/path/to/java --javac=/path/to/javac. The
> rest of the args should be like the KEYS variable.

I like to use the ./batch_run.  I just ./configure mauve without 
arguments once. Then I can set COMPILER and RUNTIME accordingly to my 
test I wanna do and run ./batch_run. Thats much more easy then 
re-reconfiguring each time. 

> * Package structure. The part that could be called the 'testlet
> api' should live in gnu.testlet. The test runner implementation
> should be in gnu.mauve.runner (?) and test hierarchy perhaps in
> gnu.mauve.tests.

Mauving things around in CVS is a bad idea as the history of changes 
for the moved files get lost. I often look into such informations.


Michael
-- 
Homepage: http://www.worldforge.org/

Re: future plans for mauve

[Michael Koch]

Am Freitag, 29. Oktober 2004 12:53 schrieb Robert Schuster:
> Hi
>
> >>* Package structure. The part that could be called the 'testlet
> >>api' should live in gnu.testlet. The test runner implementation
> >>should be in gnu.mauve.runner (?) and test hierarchy perhaps in
> >>gnu.mauve.tests.
> >
> >Mauving things around in CVS is a bad idea as the history of
> > changes for the moved files get lost. I often look into such
> > informations.
>
> Maybe this was an implicit wish to switch to SVN?
> They have a nice CVS2SVN conversion tool at tigris.org, too:
> http://cvs2svn.tigris.org/

I would like monotone more [1]. It can import CVS history too and is 
much more capable of distributed development and has better merging 
support. Perhaps someone does a drive while I'm in holiday ...


Michael


[1] http://www.venge.net/monotone/
-- 
Homepage: http://www.worldforge.org/