Thanks for chasing down, Ian. Even if its easy to fix your tests, I'd suggest also opening a bug about the breaking behavior change (on Platform IU).
There is a chance it was an accidental change they would appreciate knowing about. Even if intentional, required, or desired, it does sound like it may be "breaking API behavior" and should be well documented. Perhaps your tests were a special, unrealistic case, but still ... a change in behavior is a change in API. Thanks, From: Ian Trimble <[email protected]> To: "General discussion of project-wide or architectural issues." <[email protected]> Date: 11/27/2009 05:40 PM Subject: RE: [wtp-dev] Informally Promoted Build for wtp-R3.2.0-I: I-3.2.0-20091127073002 Sent by: [email protected] Sure enough, org.eclipse.ui.wizards.datatransfer.ImportOperation has changed in a way that breaks a few of our (JSF Tools Project) tests. ImportOperation used to tolerate being constructed with an empty (but not null) path, and now it does not. We will need to change our test data and test utility slightly to accommodate the change. - Ian From: Ian Trimble Sent: Friday, November 27, 2009 10:45 AM To: General discussion of project-wide or architectural issues. Subject: RE: [wtp-dev] Informally Promoted Build for wtp-R3.2.0-I: I-3.2.0-20091127073002 I’ve taken a look at the failing JSF test cases and have narrowed it down to all tests that use a specific JSF testing utility method (ProjectTestEnvironment.createFromZip(…)). This method creates an org.eclipse.ui.wizards.datatransfer.ImportOperation with an empty “containerPath”. Running the current JSF code under an older platform/WTP build (Galileo, or at least close to it) does not complain about this empty path. When I can find time, I will set up a newer build environment to see if the change is, as I suspect, to ImportOperation. I don’t think ImportOperation is necessarily flawed; I think it’s just no longer somehow dealing with being passed an empty path. The short version is, JSF tests are currently failing due to how the tests are written, and not due to non-test code breakage. It’s highly doubtful that this could affect anyone else (but you’d be well-advised to check any of your code that uses ImportOperation). Happy Thanksgiving, US folks, - Ian From: David M Williams [mailto:[email protected]] Sent: Friday, November 27, 2009 6:40 AM To: [email protected] Subject: [wtp-dev] Informally Promoted Build for wtp-R3.2.0-I: I-3.2.0-20091127073002 At our status meeting last week, we decided we would not have a declared build this week, since so many were out on holiday. But a few days after the meeting, some committers (not in the US :) said they would like a build promoted to 'downloads' to enable some early adopter testing and development, since M4 is not that far off. So, I'll promote this latest build, but call it "informally promoted" instead of "declared" since it has not been tested as we normally do. I will leave it "invisible" if you look at the plain downloads page, http://download.eclipse.org/webtools/downloads/, but if you know this whole URL you can get right to it. Download Page: http://download.eclipse.org/webtools/downloads/drops/R3.2.0/I-3.2.0-20091127073002/ Hopefully this will prevent people from casually downloading it, thinking it was just like every other I-build. Hopefully it is a good build, but since we have not smoke tested it, we can not confidently say much about it. There are a few JUnit failures that appear significant. They occur in the JSF tests, but suspect they are due to changes made elsewhere? _______________________________________________ wtp-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/wtp-dev
_______________________________________________ wtp-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/wtp-dev
