Tiny differences in projection regression unit test with Java 9

[Don-vip]

I have added Java 9 basic support to JOSM in r8777:8778.

JOSM now compiles and works fine, as only one unit test is failing.

I think this is due to a change of implementation of Math.toDegrees in Java 9, see ​javabug:8051808 and ​javabug:4477961.

I'm not sure what's the best solution for this issue. I'd like to have the test working fine for the three configurations (JDK7, 8, 9).

[wiktorn]

Wouldn't change of regression test asserts from *.equals(...) to *.equalsEpsilon(...) suffice?

Though, this could impact other places, where we are comparing using equals(...) of LatLon's and EastNorth's

But isn't the thing, that we should do anyway?

[Don-vip]

It would work, yes. Paul, Simon, what do you think?

[simon04]

To my understanding, performing a projection (and back) often involves a loss of precision. So I doubt that bitwise double equivalence is necessary. equalsEpsilon seems an adequate replacement to me …

[Don-vip]

I'm afraid that's against the spirit of this test. I'm going to take a safer approach right now, we'll see later if we can switch back to an epsilon-based version.

[Don-vip]

In 8793/josm:

fix #11889 - adapt projection regression test to Java 9. Safe approach: keep exact comparison but duplicate data files.

[Don-vip]

In 8796/josm:

fix #11889 - adapt projection regression test to Java 9 for Jenkins build

[Don-vip]

In 8797/josm:

fix #11889 - adapt projection regression test to Java 9 for Jenkins build

[bastiK]

In 12013/josm:

see #11889 - backport improved version of Math.toDegrees and Math.toRadians from Java 9

[Don-vip]

In 12015/josm:

see #11889 - fix javadoc warnings

[Don-vip]

In 12131/josm:

see #11889, see #11924, see #13387 - use backported versions of Math.toDegrees/toRadians (more accurate and faster) - to revert when migrating to Java 9