Canonicalization tests - N-triples and N-Quads --> fails - most likely it only fails on windows

[arne-bdt]

Version

6.1.0-SNAPSHOT

What happened?

Revision: 7810feb
org.apache.jena.riot.Scripts_C14N does not successfully run unter windows.

**** Test: C14N comment_following_triple
---- Input
<http://example/s> <http://example/p> <http://example/o> .
---- Actual
<http://example/s> <http://example/p> <http://example/o> .
---- Expected
<http://example/s> <http://example/p> <http://example/o> .
--------

org.opentest4j.AssertionFailedError: Does not match expected canonical text ==> 
Expected :true
Actual   :false

The debugger reveals that actual ends with \n, where expected end with \r\n:

In org.apache.jena.riot.writer.WriterStreamRDFPlain#triple and #quad the last line is:
out.print(" .\n");

Changing these two occurences to
out.print(" ." + System.lineSeparator());
fixes the tests.
But the code is that way since 09.03.2013 (DD.MM.YYYY), so I am pretty sure this would not be the desired solution.

Which would be the desired solution?

Relevant output and stacktrace

Are you interested in making a pull request?

Maybe

[afs]

Thanks for the report.

At least for the C14N case it should not be \r\n. It is going to have line endings itself.

[arne-bdt]

The expected value with \r\n is created by Files.readAllBytes, so this seems to be the way windows reads it.
Or the way it is written during git checkout on windows.

[afs]

The problem is looking like it's git.

The file needs to be marked "text eol=lf" so that conversion to local OS newlines does not happen. Files.readAllBytes does not process newlines. Jena actual looks right. In canonical form, the choice of newline is fixed as \n (and then file signing works).

The files look to be wrong in the offical rdf-tests --
https://github.com/w3c/rdf-tests/blob/main/rdf/.gitattributes for other files.
Issue raised: w3c/rdf-tests#317

Thank you! It's important to fix at source.