Remove methods deprecated during gsoc-core

[Don-vip]

Deprecated features from gsoc-core to remove, which are still used in core or plugins:

• LatLon.heading(): used in 5 plugins: improveway, iodb, nanolog, photoadjust, rex, ​SimplifyArea
• Preferences.getColor(): used in MapCSSParser + plugins: utilsplugin2, commandLine, routing, graphview
• Layer.isChanged(): used in MapView, GpxLayer, AbstractTileSourceLayer + plugin osminspector
• ImageryLayer.getDx()/getDy()/setOffset()/displace(): used in ApplyOffsetAction, ImageryLayer + 4 plugins: commandLine, imageryadjust, iodb, irsrectify

[michael2402]

I don't have much time right now and won't have until March 31. I would like to re-schedule this one for milestone 17.4 (I just put it to 17.3, since there is no 17.4 yet).

[Don-vip]

ok thanks for the heads up :)

[stoecker]

In 11623/josm:

see #14120 - fix deprecation in core

[Don-vip]

In 11631/josm:

see #14120 - update to JavaCC 7.0.2 - patched to avoid error-prone compilation warnings by removing everything related to ReInit methods in ParseGen.java and SimpleCharStream.template

[Don-vip]

The only warnings in core now concern Layer.isChanged().

[stoecker]

@michael2402:

Can you please give fixing the deprecations highest priority? We need to get a clean state, before we start other deprecations.

[Don-vip]

14.04 is due in two weeks, can you please fix the core deprecation warnings before that?

[Don-vip]

I think I have updated all the plugins now. But it seems we still have a bug in core:

• #13503: IAE: Cannot paint layer, it is not registered (TMSLayer)
• #13604: IAE: Cannot paint layer, it is not registered (OsmDataLayer)
• #13886: IAE: Cannot paint layer, it is not registered (ValidatorLayer)
• #14234: IAE: Cannot paint layer, it is not registered (GeoImageLayer)
• #13589: IAE: Cannot paint layer, it is not registered (ElevationProfileLayer)

[Don-vip]

One week left before 14.04. Because of #14604 impacting lots of users, and with #7427 being fixed, the release date this month will not shift.

[michael2402]

I've been busy the last week, but I have lots of time this one.

[michael2402]

In 12014/josm:

See #14120: Use a listener to get notified of way segment / virtual node highlight changes.

[Don-vip]

Two unit tests are failing.

[michael2402]

I already saw that. Some problem with the test framework removing a layer twice, i'm debugging it but tests always take some time to run ;-)

[Don-vip]

ok thanks :) I tried to fix them but didn't understand the error.

[Don-vip]

In 12020/josm:

see #14120 - checkstyle

[michael2402]

In 12023/josm:

See #14120: Do not overwrite the mapView field for unit tests. Fix map view scale/center.

[michael2402]

In 12024/josm:

See #14120: Fix unit tests by using a separate layer manager when mocking the map view.

[michael2402]

In 12025/josm:

See #14120: Throw a nicer exception if a layer is destroyed twice.

[michael2402]

I found the problem:

The map view handles the layer lifecycle. If tests create multiple map views for the same layer, each map view destroys the layers during test exit - so we have several destroy() calls. I removed one of the places where MapViewMock is used. We could completely remove that Mock, I think.

[Don-vip]

In 12028/josm:

see #14120 - fix java warnings

[Don-vip]

there are still 10 warnings to fix.

[Don-vip]

Also 6 unit tests are failing for 23/26 builds.

[michael2402]

I've seen them. The most important thing for me is getting the warnings safely away until next release. They are caused by a selection change event that was used to trigger a map view repaint in most cases. There is a reason why someone added those calls back then, so I'm going trough the UI code and fixing them one by one (or check if it is already fixed by some other mechanism that triggers a repaint). And I'm fixing some other repaint/layout related issues on the way.

The AlignInLineActionTest(s) are caused by the selection not being sorted for a single addSelection(). I tried around a bit and it seems the LinkedHashSet order gets lost in the streams, even when using a LinkedHashSet as collector. Have to look into it more. I was not aware that the selection order is used that much.
If users select the points, multiple addSelection calls will be issued. So for now, this is only a problem if a plugin uses this functionality.

I'll probably set up travis/git again next week, so that I will be able to run the tests in background more easily.

After the selection stuff is done, I already have a patch for the GPX layer that removes the update counting code. That way, we can get rid of the isChanged() methods and rely on layer invalidation events in the future. This should allow us to remove most of the MapView.repaint() calls in code.

[Don-vip]

OK thanks :)

[Klumbumbus]

Replying to michael2402:

I was not aware that the selection order is used that much.
...
So for now, this is only a problem if a plugin uses this functionality.

(Just in case you didn't know: When creating turn restrictions with the turnrestrictions plugin, the selection order is required as the roles are set automatically depending on the selction order.)

[michael2402]

In 12103/josm:

See #14120: MapView should repaint on temporary layer changes automatically.

[bastiK]

Only one failed unit test remaining...

[michael2402]

In 12156/josm:

See #14120: Don't make gpx tracks depend on the isChanged method, use a listener based approach instead.

[michael2402]

In 12170/josm:

"See #13175, see #14120: Remove isChanged() method call in map view - all layers should invalidate correctly now."

[Don-vip]

Only one test to fix and we won't get spammed by Jenkins at each build anymore: MapCSSRendererTest.testRender[order]. After that I expect the build to be stable a longer time. A test failing for 91 builds is a lot of noise :(

[michael2402]

I did not know you get a mail for every test run...

But now they are all fixed. There a still a a few deprecations to track down, but then everything should be stabilized for the next release.

[michael2402]

I don't want to remove the remaining selection changed calls now since I did not fully test if/where they are required.

[Don-vip]

only one java warning left.

[Don-vip]

Michael can you please fix the last deprecation warning?

[michael2402]

I'll have a look at it.

The line can simply be removed in theory, the hard part is to check if all UI components are actually redrawn correctly if commands are executed.

[Don-vip]

Ticket #15251 has been marked as a duplicate of this ticket.

[bastiK]

In 13010/josm:

see #14120 - fix use of deprecated API (selection handling)

[bastiK]

Replying to michael2402:

The line can simply be removed in theory, the hard part is to check if all UI components are actually redrawn correctly if commands are executed.

Looks like it is working. Why is the private method in DataSet named fireDeprecatedSelectionChange?

[michael2402]

Replying to bastiK:

Why is the private method in DataSet named fireDeprecatedSelectionChange?

Because I wanted to deprecated the old selection listener mechanism first - but it was used in too many places so that I could not do this on the inital patch. And then I forgot about the method name...

See #13467

[Don-vip]

@team: I'm not sure how to update last plugins regarding deprecation of heading()/getDx()/getDy() I'd like to get some help.

[Don-vip]

Help is still welcome.

[Don-vip]

​rex plugin updated.

[Don-vip]

iodb updated in [o34337:34338]

[Don-vip]

nanolog updated in [o34339:34340].

[Don-vip]

​improveway plugin updated.

[Don-vip]

In 13966/josm:

see #14120, see #12524 - remove deprecated method LatLon.heading()

[Don-vip]

irsrectify plugin deleted in [o34341].

[Don-vip]

commandline updated in [o34342:34343].

[Don-vip]

In 13967/josm:

fix #14120 - remove deprecated methods ImageryLayer.getDx()/getDy()/setOffset()