I've requested that my account be updated. Theoretically when that happens folks should get notified and we can check and/or restore my permissions.

I have the tests/jira directory working, with replacement tests for tests/2.7.3_release. Currently that works like the tests/bugzilla directory, able to support both simple compare-transformation-against-expected tests and java-based tests for things that require either more setup (eg changing timezone) or more detailed examination of the output (checking the formatting of that output).

One of these bug-report tests appears to be  recastable as an equivalent of tests/conf-fail for exslt -- which would require adding an exslt-fail directory and configuring another build target for it.


Ideally, I'd like to reconsider the test architecture, since java-based tests may be appropriate in other categories; having type of test hardcoded into the build by directory is a bit inelegant. Among other things, as bugzilla/jira reports are resolved, it would be good to move them out of those directories and into the appropriate directory for the section of the system they're testing. An alternative would be to have bugzilla-resolved and jira-resolved directories, which would avoid disrupting the existing test framework, but which would mean looking multiple places to find a test related to a given set of functionality.

Note that a number of the tests in the bugzilla directory are now passing, which suggests that they *should* be moved/marked as resolved (assuming that the issues they were responding to have indeed been addressed, and that they aren't failing on a platform I haven't tested on.)


I have an eclipse .project file for the current arrangement of xalan-java and xalan-test as peers. It includes necessary paths to eliminate the classpath issues a naive load into Eclipse will report for xalan-test. If we decide to refactor xalan-test into xalan-java, I'll want to rework this but could then commit it to git so setup is easier for other folks.


Re test output: The logging system does have a distinction between ERROR, INFO, and STATUS messages. We've often been a bit sloppy about distinguishing these, but we could review and make sure things are categorized correctly, then have the smoketest builds run with only ERROR and above enabled. That should cut down a lot of the chatter.

Similarly, there are a few thousand compiler warnings being issued, and many of them can be trivially fixed -- for example, the type-mismatched comparisons against Strings generally just want a .toString() inserted on the non-String side. Others might take a bit more effort but should be resolvable. Again, a cleaner build makes seeing actual issues much easier.

And I believe we still have some open issues to be addressed.


I'd suggest resolving technical debt  would be a worthwhile investments in the codebase before starting significant new development. It isn't as exciting, but it avoids trying to develop on top of a questionable base.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@xalan.apache.org
For additional commands, e-mail: dev-h...@xalan.apache.org

Reply via email to