Okay, so the change in the code is ludicrously simple:

yaml.go

	// Line and Column hold the node position in the decoded YAML text.
	// LineEnd and ColumnEnd hold the position of the end of the node.
	// These fields are not respected when encoding the node.
	Line      int
	Column    int
	LineEnd   int
	ColumnEnd int

decode.go

	n.Line = p.event.start_mark.line + 1
	n.Column = p.event.start_mark.column + 1
	n.LineEnd = p.event.end_mark.line + 1
	n.ColumnEnd = p.event.end_mark.column + 1

but the change in the tests looks to be a bit of a pain:

$ grep -R "\b(Line|Column): " node_test.go | wc -l
628

Here's the helper regex I wrote for me:

^(\s*)Line:( *)(\d+),\n(\s*)Column:( *)(\d+),\n
$1Line:$2$3,\n$4Column:$5$6,\n$1LineEnd:$2$3,\n$4ColumnEnd:$5$6,\n

-- now I just have to fix the line and column numbers everywhere ;-;

It appears the p.event.end_mark from https://github.com/yaml/libyaml is unreliable. I checked their repository, and they don't have tests for start_mark and end_mark :/.

The end_mark is not completely useless, but it's just not always accurate. I'm tempted to hack the test function to make it ignore the value of end_mark, since it's inherited from an external library. -- The test should not be done in https://github.com/go-yaml/yaml, but rather in https://github.com/yaml/libyaml, since this is where the code of the parser actually lives.