Hi,
something that I have in mind for some time
now and where I got reminded by Ryan Ovrevik in
XDT-1002 and others is regresion testing.
I basically see three cases:
a) make sure that generated code compiles (and
would even run)
b) validate the generated xml - against a DTD and
some reference xml
c) let the developer validate the written .xdt
files without running against a test case
The first part is easy for a) just add a compile
step into every build of a module.
The second part could mean to run findbugs(*1)
against the generated code and make
sure, that there are no prio 1 errors.
Same for b) the first part is done within the
samples directory. But this doesn't tell
me if a change in some .xdt will leave out some
elements/attrbutes, but shouldn't as
long as the DD ist still valid.
c) is only caught when running the samples _and_
the sample needs the specific .xdt
file, which takes quite some round trip
time.
Obviously the first two are more things for an ant
task, while the latter would be an interactive
thing.
The other thing is layout of tests. There are
already some unit tests in core. Should
unit tests follow this and have a test directory
under each module? Or should it follow
the samples case with a new 'toplevel' directory,
where all test go into (I am in favour of that)?
Is junit the right framework for all this? Are
there better ones?
Suggestions? Opinions? Volounteers?
Heiko
