Closed rschnekenbu closed 5 years ago
Actually, there's also a test failure that is currently hidden by one of the assumption violations, that indicates a real regression. In test LifelineSwitchingUITest::ExecutionSpanningOccurrences::switchLifeline
, it used to be that when a message becomes a self-message by virtue of the execution specification moving from one lifeline to another when re-targeting a synchronous message, that self-message was shaped properly with the minimal vertical span. However, with the changes in this pull request, that message now has no vertical size. It's this scenario:
But worse than this is what happens when attempting the same edit but by moving the execution specification directly (not re-targeting the request message):
This really breaks the diagram, including an error dialog with cycle in the dependency graph. Looks like an inconsistency in the creation/ordering of SetOwnerCommand
s and SetCoveredCommand
s?
Hi, @rschnekenbu. I've made pull request #452 targeting your branch that should clear up the assumption violations on Mac platform. It also fixes some that I think were happening on all platforms (I see them in the build log also), one of which exposes now a test failure that identifies an actual regression in the editor's behaviour.
Change-Id: Ieb479d5b1109c4f4b4a76c0df5e45b48c73192d2 Signed-off-by: Remi Schnekenburger rschnekenburger@eclipsesource.com