fix off by one error in multiline highlighting

[digama0]

Implements the fix described in #41.

Fixes #41

[zbraniecki]

Thank you for the contribution!

[zbraniecki]

It seems like it makes fixture tests fail. @digama0 can you verify that the test positions should be affected by your patch and that the new end indexes are correct?

[zbraniecki]

I see the following follow up PR fixing the tests:

diff --git a/tests/fixtures/no-color/multiline_annotation.toml b/tests/fixtures/no-color/multiline_annotation.toml
index bdb577f..c3dc1e9 100644
--- a/tests/fixtures/no-color/multiline_annotation.toml
+++ b/tests/fixtures/no-color/multiline_annotation.toml
@@ -33,7 +33,7 @@ range = [5, 19]
 [[slices.annotations]]
 label = "expected enum `std::option::Option`, found ()"
 annotation_type = "Error"
-range = [22, 765]
+range = [22, 766]
 [title]
 label = "mismatched types"
 id = "E0308"
diff --git a/tests/fixtures/no-color/multiline_annotation2.toml b/tests/fixtures/no-color/multiline_annotation2.toml
index 6ec0b1f..845bf9f 100644
--- a/tests/fixtures/no-color/multiline_annotation2.toml
+++ b/tests/fixtures/no-color/multiline_annotation2.toml
@@ -10,7 +10,7 @@ fold = false
 [[slices.annotations]]
 label = "missing fields `lineno`, `content`"
 annotation_type = "Error"
-range = [31, 127]
+range = [31, 128]
 
 [title]
 label = "pattern does not mention fields `lineno`, `content`"

[digama0]

Yes, the new spans are one character too short now but that's because the input span was already compensating for this bug with an end index which ends one character before the actual span. (Was this a real span generated by rustc?) Your suggested fixes put the input spans where they should be, so the output should now be the same as before.

[zbraniecki]

Thank you for adding the tests!