| Summary: | build tests can not run due to blocking XML Validation Tests | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | [WebTools] WTP Source Editing | Reporter: | David Williams <david_williams> | ||||||||
| Component: | wst.xml | Assignee: | Lawrence Mandel <lmandel> | ||||||||
| Status: | CLOSED FIXED | QA Contact: | |||||||||
| Severity: | blocker | ||||||||||
| Priority: | P3 | CC: | jeffliu, naci.dai, ryman, thatnitind | ||||||||
| Version: | unspecified | ||||||||||
| Target Milestone: | 1.0 M6 | ||||||||||
| Hardware: | PC | ||||||||||
| OS: | Windows XP | ||||||||||
| Whiteboard: | |||||||||||
| Attachments: |
|
||||||||||
|
Description
David Williams
Created attachment 24667 [details]
Patch for org.eclipse.wst.xml.validation.tests
This patch will set the wtp.quiet system property when AllXMLTest is run.
Just to update everyone on this. Reason why the fix was not done on the master test script level is because passing in the "extraVMArgs" properties from the master script will override other VM args specified in the individual scripts (for example, Tomcat install location). So if we want to pass in this property from the master script, we need to rework some of the individual scripts, and that will be dangerous at this point. So the safest approach (at least for now) is to pass in the property from the XML validation test script. Thanks for the patch and the conceptual explaination, Jeffrey. BTW, I had tried "forcefully" removing the XMLValidator test causing the problem, but the hang just occurred later .... at least in J2EE tests, if not else where, so there is a lot to work through. BTW, is it possible to take advantage of Ant's -noinput property in this regard? Seems that should be set anyway, right? And maybe could be done globally and wouldn't require "wtp.quiet" to be set? We should probably pass in -noinput regardless. However, I don't think this option will help the current hang because our hang has to do with a user click. Since the scope of this problem is more than just XML validation, I'll look into the "pass in property from the master script" approach... Stay tunned. From http://codefeed.com/blog/?p=68: "For unattended builds, you can use the -noinput flag whereupon any attempt to read input in Ant will cause an end-of-file result to be given to the read." Created attachment 24683 [details]
Patch for releng.builder
This patch will pass in the wtp.quiet property from the master test script
level without overriding the custom VM args defined in the individual test
scripts. However, I wasn't able to test it on my local build because something
else failed... not sure what it is. Anyways, are we brave enough to try it?
Thanks Jeffrey, it looks like this patch would work transparently, to me. Naci, can you please review too? Lawrence, 1. I would like the wtp.quiet property renamed, though ... "quiet" usually refers to output ... I think wtp.noninteractive would be better. (And, please, if you find that value set 'true' in your "product" code, please log it to console log ... it might be a source of reports during non-test use of something "not working" if some user were to innocently set it, not knowing fully what it was for ... perhaps a name of wpt.internal.testonly.noninteractive would be better? -- in addition to logging it. 2. I would like Bug 103611 addressed before the cache license re-released. 3. I would also like to be sure there is a "practice build" done before the cache license is re-released. So ... I'm assuming this will all be done post RC1. If you feel we should not declare RC1 without the license code in place, then please have Arthur review issures and work items, for his help in getting the corrective actions in place, and make a decision if further delays warrented. David, 1. Good idea. Do you think wtp.internal.testonly.noninteractive is necessary? Will wtp.noninteractive be fine? How about wtp.autotest.noninteractive? (The name just needs to be clear, the exact name is not very important to me.) I have a fix ready that will log the use of this property to the log now we just need to decide on the var name. 2. I'm working on bug 103611 but so far have been unable to reproduce. 3. Agreed. I want to make it clear that the license code is already in place. The significant change was to enable the cache by default. With the cache off by default a user can still choose to enable the cache and will then be prompted for a license and may run into the same problem as you have. Jeff thanks, the patch seems ok. Lawrence, I think "wtp.autotest.noninteractive" would be good. The name isn't *real* important to me .. but "wtp.quiet" sounds like something I'd just try out for fun just because its sounds like it must be something really good. :) I'll check my browser versions and setup later tonight. I guess technically, we don't need to do a "practice build" ... just to set up an "ant test environment" and run the built, packaged JUnits, with just that one prefernce or jar changed ... if that's any easier for you or others. I've changed the system property to wtp.autotest.noninteractive. Created attachment 24744 [details]
Patch for releng.builder to support wtp.autotest.noninteractive property
The patch above is similar to Jeff's patch. Two differences.
1. The property name is updated to reflect the change.
2. Instead of creating a second library xml file the extraVMargs property is
passed directly to the test script.
I am running unit tests on the 0713 build now. So far so good. I'll report the
complete status once they finish.
The test results look good on my machine. The tests ran unattended and completed successfully. I ran the tests on a WindowsXP box after downloading the 0713 build. 1. As stated, the rename of the system property is complete. 2. In progress. 3. Success in running the ant test environment. I have reviewed and approve this fix and moving forward on getting this change into the builds. Fix committed. Test build ran to completion. How's this for my fastest verification :) Thanks David. Closing bug. |