The way the E4 UISynchronizer runs UI messages seems to be changed.
The usage tests run in the following way:
- the usage_reporting_enabled property is set to false in the parent pom
- Eclipse starts the usage extension point that triggers opening the
dialog asynchronously
- Eclipse runs tests that set the usage_reporting_enabled property to
true
Eclipse < Kepler M6 works as follows:
- the usage_reporting_enabled property is set to false in the parent pom
- Eclipse starts the usage extension point that triggers opening the
dialog asynchronously
- Eclipse tries to open the dialog; since
usage_reporting_enabled=false, the dialog isn't open
- Eclipse runs tests that set the usage_reporting_enabled property to
true
Kepler M6 on Linux works in the following way:
- the usage_reporting_enabled property is set to false in the parent pom
- Eclipse starts the usage extension point that triggers opening the
dialog asynchronously
- Eclipse runs tests that set the usage_reporting_enabled property to
true
- Eclipse tries to open the dialog; since
usage_reporting_enabled=true, the dialog is open
so to sum up >= Kepler M6, the tests are run before the Eclipse
extension tries to open the dialog, right (< Kepler M6 it's the other
way round)?
I have no idea whether this is expected or not. Who does actually decide?
There are no more Eclipse instances when running a test plugin.
I dont get this one, can you please explain more?
Also, it is not important if a test is an UI test or not. Eclipse
will
always call the usage's startup ext. point if the usage plugin is
included which will trigger opening the dialog.
Sure, no doubt.
You can check this as follows:
- clone the jbosstools-build component
- remove -Dusage_reporting_enabled=false from the parent/pom.xml
- mvn clean install -f parent/pom.xml
- cd <usage_test_directory>
- mvn clean verify
The test will always freeze - on Linux, Windows, Mac, Kepler M6,
Kepler M5, Indigo...
Snjeza