[roadmngr] Fixed Position::Delta sometimes flipping dLandId

[brifsttar]

Hi Emil,

This should fix #261. I've taken this opportunity to also make the if condition more readable.

[eknabevcc]

I understand this loop. It seems like a flaw that visited[0] is not the actual first road in the "path".
Still it would be nice to skip the loop. I'll check later if there is another solution (to store FIRST road).

[brifsttar]

The loop is taken from RoadPath::Calculate, which faces the same problematic of finding the first connection. The method could therefore store that value once found, so that it can be used elsewhere.

esmini/EnvironmentSimulator/Modules/RoadManager/RoadManager.cpp

Lines 4505 to 4528 in 73390cc

	while (node)
	{
	if (node->previous == 0)
	{
	// This is the first node - inspect whether it is in front or behind start position
	bool isPred = node->link == startRoad->GetLink(LinkType::PREDECESSOR);
	bool isGTPi2 = abs(startPos_->GetHRelative()) > M_PI_2;
	bool isLT3Pi2 = abs(startPos_->GetHRelative()) < 3 * M_PI / 2;
	bool isSucc = node->link == startRoad->GetLink(LinkType::SUCCESSOR);
	bool isLTPi2 = !isGTPi2;
	bool isGT3Pi2 = !isLT3Pi2;
	bool isPredAndBack = isPred && isGTPi2 && isLT3Pi2;
	bool isSuccAndFront = isSucc && (isLTPi2 || isGT3Pi2);
	if (isPredAndBack || isSuccAndFront)
	{
	direction_ = 1;
	}
	else
	{
	direction_ = -1;
	}
	}
	node = node->previous;
	}

[eknabevcc]

Yes, I like the idea to store the already calculated start direction. And then re-use it the Delta method which could be used frequently, hence we want to minimize unnecessary work.

I took the liberty of committing an attempt to achieve this on your branch. If it builds, please test and report how it works.

[eknabevcc]

Good that a test case caught a bad attempt from my side.

I made another commit that seems to work better. But since it included also changes in RoadManager.hpp (which is not changed in your commit) I can't edit it in this PR. I tried to figure out a way to commit to the PR branch, but I had to give up.

Instead I created a branch fix_flip. Could you perhaps merge that branch on your (PR branch)?

As a last resort we could close this PR and I could merge fix_flip on master instead.

[brifsttar]

I merged your branch on this PR, and I ran the test that initially failed (see #261), and it worked as expected. So all is ok on my end.

[eknabevcc]

Excellent, thanks for the fix!