Move tests/2.7.3 into main test driver framework, document bugzilla test status (#1)
* Complete moving 2.7.3_release to jira directory, add some comments, chmod u+x *.sh, remove jira .class file
* Analyze AMBG bugzilla testcases, create reference output files when not present, document status and recommended action. Most are obsolete and should be either discarded, folded into our test suite, or moved to tests/contrib; awaiting consensus on that decision.
* A few more gold outputs
* Sloppily make 1266 FAIL rather than AMBG, by throwing arbitrary exception in the unexpected cases. We do still seem to be failing to log the warning event.
* Document status of Bugzilla6392: Test doesn't self-evaluate, can't reproduce reported issue, but there may be an edge case accidentally exposed that is worth sanity checking. Dig more deeply.
* Bugzilla 1288 is Excluded, referring as it does to a class that no longer exists. Drop it entirely? Figure out what it was testing and rewrite it?
* Bugzilla3001 analyized, user error, working as designed.
* Replace Bugzilla3001.out with correct output (rather than expected output).
* Bugzilla 3489 analyzed; Working As Designed.
* Forgot to fix filename in code when changing filename convention to Bugzilla. My fault.
* doc
* Improve Bugzilla datalet failure message so we know what fell over.
* Fix exclute, filename. Need to look at why 2925 is find finding its .out file; probably related to the fact that it's writing the .xsr output to .
* More user error cases. We should improve bugzilla example to find gold file both locally and in ./tests/bugzilla, since depending on how invoked it might be in either... could fix in framework, but since bugzilla directory is being phased out I'm not sure it's worth the effort.
* Ah. The mistake in the .error file was deliberate to provoke the hang/no-report. Putting that back, tagging it as such.
* 1266 analyzed, appears to be a genuine initialization sequencing bug.
* Bugzilla1266 analyzed, possible fix proposed (not yet tested).
* Bugzilla1251: Description is backward, but appears to be a valid complaint that TransformerIdentityImpl, unlike TransformerImpl, does not have the code needed to redirect events to a user-provided TransformerHandler/m_serializationHandler. Genuine missing feature.
* doc
* Further evaluation of tests, in particular those reported as INCP.
* Put a rough timestamp on my analys comments, for cleanliness.
* jkesselmn, not -mn.
* Missing goldfile.
* BugzillaNodeInfo case analyzed. Enabling source_location does not record data for temporary trees loaded from stylesheet literal elements. Unclear whether that should be considered a bug or a feature request.
* doc
* Folks nitpicked about -- versus em-dash. The former is pervasive through Xalan, and is Not Wrong, but people wanted this instance changed. We can do that.
* Upon reconsideration, replacement of -- in non-javadoc comment with em-dash is not accepted as a friendly amendment. If you want that change made, submit your own change request, preferably across the entire codebase, and offer a justification that overcomes the additional barrier this presents to community participation.
* Bugzilla2925 analysis: Doesn't seem to test what it apparently wanted to test. We would need to turn it into a workable test before we could determine what to do with it... unless we just want to discard it as incomplete/test-author-error.
* Clarify status comment. The only testing issue is that BugzillaTestletDriver executes with current directory xalan-test, whereas this code assumes cd is xalan-test/tests/bugzilla. See open Jira issue re rationalizing test driver/test conventions.
* Bugzilla1283: Can not reproduce error. See comments.
* BugzillaNNNN also has the issue of local file references -- since it was being used as template for other bugzilla tests, that's where they got the problem from. See open Jira issue, and comment in this file that having BugzillaTestletDriver run the java-based tests from tests/bugzilla would work around the issue. Same probably applies to JiraTestletDriver, since I did a quick clone from how we were handling bugzilla.
* Review bugzilla StylesheetTestlets. All are currently passing in Linux. Comment them to that effect. I believe all Bugzilla testcases checked in here are now reviewed.
---------
Co-authored-by: kubycsolutions <keshlam@kubyc.solutions>
73 files changed
