Visual Prolog: multiple cases of different styles between \r and \n

[rdipardo]

Similar to #57, testing LexVisualProlog found mixed styles at CRLF line ends wherever StyleContext::atLineEnd is used to find newlines:

~/../AllStyles.pro: 1: different styles between \r and \n at   33: 5, 0
~/../AllStyles.pro: 4: different styles between \r and \n at   75: 5, 0
~/../AllStyles.pro: 7: different styles between \r and \n at  126: 5, 0
~/../AllStyles.pro:10: different styles between \r and \n at  176: 5, 0
~/../AllStyles.pro:25: different styles between \r and \n at  561: 5, 0
~/../AllStyles.pro:26: different styles between \r and \n at  572: 5, 0
~/../AllStyles.pro:27: different styles between \r and \n at  584: 5, 0
~/../AllStyles.pro:29: different styles between \r and \n at  618: 5, 0
~/../AllStyles.pro:33: different styles between \r and \n at  679: 5, 0
~/../AllStyles.pro:36: different styles between \r and \n at  732: 5, 0
~/../AllStyles.pro:39: different styles between \r and \n at  781: 5, 0
~/../AllStyles.pro:40: different styles between \r and \n at  795: 16, 19
~/../AllStyles.pro:42: different styles between \r and \n at  838: 5, 0
~/../AllStyles.pro:45: different styles between \r and \n at  909: 5, 0
~/../AllStyles.pro:48: different styles between \r and \n at  988: 5, 0
~/../AllStyles.pro:49: different styles between \r and \n at 1002: 20, 22
~/../AllStyles.pro:50: different styles between \r and \n at 1014: 20, 22

Namely:

• line comment style (id 5) ends mid-EOL:

• an error indicator style for unterminated string literals (SCE_VISUALPROLOG_STRING_EOL_OPEN [id 19]), is applied too late:

• a line continuation indicator style for multi-line verbatim strings (SCE_VISUALPROLOG_STRING_VERBATIM_EOL [id 22]), is also applied too late:

Here are the test cases that currently raise the diagnostics: visualprolog.zip

Note that style ids 13, 14 and 15 are not implemented:

git grep -E "SCE_VISUALPROLOG_CHARACTER(|_TOO_MANY|_ESCAPE_ERROR)" -- *.cxx

echo $?
1

This patch fixes the EOL splitting: 0001-LexVisualProlog-Prefer-MatchLineEnd-to-atLineEnd.diff.txt

[rdipardo]

[Correction]: all the faulty state changes including line comments are caused by late detection of DOS EOLs

[nyamatongwe]

Note that style ids 13, 14 and 15 are not implemented

Character and string literals only differ in there being a single character between the quotes for a SCE_VISUALPROLOG_CHARACTER character literal. Thus SCE_VISUALPROLOG_CHARACTER_TOO_MANY doesn't seem to make sense.

https://wiki.visual-prolog.com/index.php?title=Language_Reference/Lexical_Elements#Character_Literals

[rdipardo]

Character and string literals only differ in there being a single character between the quotes for a SCE_VISUALPROLOG_CHARACTER character literal.

Perhaps the unrealized intention was to enforce a pedantic limit on the length of isolated escape characters or UTF-8 codepoints.

You may also notice the proposed test cases don't iterate every possible combination of verbatim string delimiter.
I figure the behaviour under test is what happens after a style's case label is reached, not the domain-specific logic that leads control to there.

[nyamatongwe]

0001-LexVisualProlog-Prefer-MatchLineEnd-to-atLineEnd.diff.txt seems fine but Lexilla is in release freeze so will be looked at after release.

I meant AllStyles.* to have examples of all styles as it is sometimes difficult to work out the purposes of styles. Exhaustive coverage of cases (or paths in the code) is a lot more work so hasn't been a priority.

[nyamatongwe]

Committed with addition of Unicode escape example '\uAB12'.