Fix TextBlockLiteralExpr in LexicalDifferenceCalculator

[blacelle]

Suggested fix for #3936.

[jlerbsc]

I'm not sure if this is the right solution. I need to take the time to look into this problem.

[blacelle]

#3936 (comment)

We also see a comment:

// FIXME: csm should be CsmTextBlock -- See also #2677

And in CsmTextBlock, we see a \n after """:

@Override
public void prettyPrint(Node node, SourcePrinter printer) {
    // Note that values within TextBlocks ALWAYS have the \n line ending, per https://openjdk.java.net/jeps/378#1--Line-terminators
    printer.print("\"\"\"\n");
    // TODO: Confirm if we need to force this to use {@code \n} separators
    printer.print(property.getValueAsStringAttribute(node));
    printer.print("\"\"\"");
}

This made me quite confident some '\n' has been lost in the process.

[jlerbsc]

The pretty printer and LPP do not work the same way at all. I have no doubt that it is necessary to add a line break but it is more the solution raises some questions. I need to dig a little deeper into the subject.

[jlerbsc]

Since the parser is not very restrictive on text blocks and the definition of element syntax in the LPP does not categorize text block delimiters, your proposal can be accepted.

However, you should simplify your unit test and avoid referring to your project. Thanks.

[jlerbsc]

To set the most appropriate eof character, you can probably use Node.getLineEndingStyle().toString().

[blacelle]

This is not a matter of style, it is a matter of Java specification Just like prettyPrinter adds a \n.

On a system where EOL is '\r', Node.getLineEndingStyle().toString() would break TextBlockLiteralExpr

[blacelle]

This is similar to CsmTextBlock.prettyPrint(Node, SourcePrinter) not relying on printer.println()

[blacelle]

https://docs.oracle.com/en/java/javase/14/docs/specs/text-blocks-jls.html

Reading https://www.vojtechruzicka.com/java-text-blocks/ makes me feel I overlooked part of the spec

Line terminators are normalized to the ASCII LF character, as follows:

1. An ASCII CR character followed by an ASCII LF character is translated to an ASCII LF character.
2. An ASCII CR character is translated to an ASCII LF character.

I'm less comfortable with dynamic EOL in this case, but you're the boss.

Should I do the same in CsmTextBlock.prettyPrint(Node, SourcePrinter) ?

[blacelle]

    // Note that values within TextBlocks ALWAYS have the \n line ending, per https://openjdk.java.net/jeps/378#1--Line-terminators

(CsmTextBlock author also felt '\n' were preferable)