Good suggestion. Having both the "by component" and "higher-level
aggregate" packaging for tests and samples makes sense to me.
Similarly, we should place any DAS-specific tests and samples under
the /java/das/ directory.
Simon
Kenneth Tam wrote:
+1 for this, but I don't think the two layouts are actually in conflict.
I completely agree SDO and SCA should build/test/package/potentially
ship separately, and the directory structure should support that. I
think:
cpp/sca/
cpp/sdo/
java/sca/
java/sdo/
is the right place to start. As far as samples and test, I think it
makes sense for each of the above directories to have "samples" and
"test" subdirs for their specific tests, but it also makes sense for:
cpp/
java/
to have their own "samples" and "test" subdirs to hold integration
tests and samples across that language's implementations. Ultimately
it might even make sense to add infrastructure into there to support
cross language testing/samples.
On 12/19/05, Pete Robbins <[EMAIL PROTECTED]> wrote:
For our C++ tree we placed samples and testing under sca and sdo:
cpp/sca/samples
cpp/sca/test
...
cpp/sdo/samples
cpp/sdo/test
This is because we saw SDO as a separate entity (built, tested, packaged
separately from SCA). The SCA has a depenedency on SDO and the SCA
samples/tests also exercise SDO.
We're happy to go along with your split if you think that's best so we are
consistent with the java tree.
We'll even use cxx instead of cpp ;-))
Pete
On 17/12/05, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote:
Looks good. +1
Jeremy Boynes wrote:
Before we import any code I think we should give a little thought to how
we would like to lay out the repository. With the number of bits and at
least two different languages this will be a little trickier than usual.
I think that it is important that someone can checkout a single URL and
expect to get to work. That means that all the things they are likely to
be working on should be grouped together. However, this should not
result in multiple versions of a project being checked out due to SVN
tags being located together.
It should also be easy to tag/branch a module so that people can work on
revolutionary changes in a private area without having to check out the
entire tree.
With that in mind, I'd like to suggest the following initial tree under
https://svn.apache.org/repos/asf/incubator/tuscany/
site # tree for the website
java # main development tree for Java implementation
java/spec/... # tree for spec artifacts
java/sdo/... # tree for the SDO runtime
java/das/... # tree for the DAS runtime
java/sca/... # tree for the SCA runtime
java/tools/... # tree for tools
java/samples/... # tree for samples of all technologies
java/testing/... # tree for acceptance/compatibility testsuite
cxx # main development tree for C++ (structure like java tree)
branches/${branchName}/... # a copy of any subtree from above
tags/${tagName}/... # a tag of any subtree above
sandbox/${name}/... # a experimental copy of any subtree
This is close to the current layout with a few minor changes.
Thoughts?
--
Jeremy
--
Jean-Sebastien Delfino
--
Simon C Nash IBM Distinguished Engineer
Hursley Park, Winchester, UK [EMAIL PROTECTED]
Tel. +44-1962-815156 Fax +44-1962-818999
