Fix folding in user-defined languages for non-windows line endings

[ssk97]

fixes #2873 and fixes #350 (duplicate issues)

I also replaced two instances of "true" with EOL_SKIP_CHECK since that's what it evaluates to (and to reduce confusion since "true" would imply EOL_FORCE_CHECK)

[ssk97]

Update: I realized that the previous fix was only for operators (like semicolon in C), I've now fixed it so that delimiters (like quotes) don't break the folding either.

[rmcurtis]

I'm so glad to see a fix for this! It's been bugging me for years. We even keep a bunch of files in Windows EOL format just so code folding works with our UDL. Otherwise they should be Unix EOL format.

I managed to compile Notepad++ with your changes and it works great, just like it should. Hope to see this fix in a released version soon.

Thanks again!

[donho]

@ivan-radic Could you validate this PR please?

[ivan-radic]

Thanks for notification @donho and thanks to @ssk97 for the fix.
I agree that "true" should be replaced with EOL_SKIP_CHECK. Variable checkEOL used to be bool, but got converted to int during development. I wonder why compiler never warned about implicit conversion??

Anyway, looking at the code variable checkEOL is used to force EOL processing when Delimiter close statement "consumes" EOL. Typical example are line comments, EOL is consumed by line comment (or delimiter that is closed by ((EOL)) special character).

Of top of my head, I don't remember why was it necessary to skip EOL processing if Delimiter open and Folding keywords that are immediately followed by EOL char, but it obviously is missing for Opetaror1 stye. Thanks @ssk97 for finding that bug. First commit should be merged into main branch.

As for second commit, it is not good:

`
if (newState != SCE_USER_STYLE_COMMENTLINE || (sc.ch != '\r' && sc.ch != '\n'))
sc.Forward(iter->length());

- if (sc.atLineStart)
- checkEOL = EOL_FORCE_CHECK;
+ if (sc.atLineEnd)
+ checkEOL = EOL_SKIP_CHECK;
`

EOL at the end of line comments is not "skipped forward" because of potential nested line comment sequence. Therefore, we can't check atLineEnd but must instead rely on atLineStart which is safe even in case of multiple nested line comments.

[ivan-radic]

Long story short: use first commit, discard second one. Some basic testing in pre-release phase by beta testers would be wise.

Best regards,
Ivan

[IvanKostyuk]

The changes resolve the problem. Folding works well with all types of EOL (CRLF, LF, CR).

[Medal-x]

Still not work when using "SCE_USER_STYLE_COMMENTLINE" & "SCE_USER_STYLE_DELIMITER" (1-8)
I added :
if (sc.atLineEnd) checkEOL = EOL_SKIP_CHECK;
About everywhere after:
dontMove = true;
in LexUser.cxx

Seem to work now!

[gwttk]

Can we have this at 7.6.6 please?

[donho]

Cherry pick Fix folding in user-defined languages for non-windows line endings but ignore Fix delimiters as well as operators as @ivan-radic suggested.