After reflection (:D), I think the constraint is weak, the creation of new syntax is rare compared to the rest.

I am confused, the generation of index.katesyntax fails because the list "variables" does not exist (there is no variable predefined in this context). We may be in a situation where a change of branch does not recompile the index file :/.

oops, missing version increase

• fix module
• remove hard coded color

I left it because it goes well with dark or light themes, but I removed it. Is the policy the same for bold, underlined and italic as for color? Because a basic applied distinction between Comment and CommentTitle seems to me preferable. I imagine not, but I would like to be sure.

OK I understand. Initially I had not changed the version number of txsh.xml and it is based on the reference in /usr/share/org.kde.syntax-highlighting/syntax/index.katesyntax rather than in the repository.
And the fact that the versions in the index file were newer than my xml files... I had errors with highlight.fish.

I forgot to ask about running the tests. Is it normal that system syntax files take precedence if we have an index.katesyntax? Currently I have to run the tests by setting the environment variable XDG_DATA_DIRS.

I agree that some rules are excessive or even false, but the last time I watched (it was several months ago now), some regexes suggested to be DetectString seemed to have writing errors (mainly \\ count as 2 characters in xml) and some more or less useful propositions. If there has been no change at this level, it should still be checked.

• fix the recursion detection and handling of nested included keyword lists

(I inadvertently edit the operators list in the previous commit)

• restore operators list
• lazy load keyword lists

resolveIncludeKeywords() would then use loadKeywords() instead of load() ? This should be enough.

Sorry to answer so late, I had trouble logging in.

include rule in keyword list to import those from another file

Yes, my compilation folder was pointing to my fork on github from which I make the patches (as I do not have the right version for ECM, I have a script that does among other things a new CMakeLists.txt in another folder and adds symbolic links for project files). If there was no more error it was because I saw them in the conflicts.

Ok, i'm doing this in the week.

There have been changes in the xml files that will cause conflicts. I would look in the next few days.

I just thought of Definition::keywordLists() which does not list the keywords used by this new syntax.

Yes, I can start tomorrow.

I copied all the keywords in the css.xml file and it's a shame. I think that Keyword could refer to those of other files with ## as for IncludeRules.

This follows https://mail.kde.org/pipermail/kde-frameworks-devel/2018-July/066564.html

It is done. So much emotion :')

Thank you, but I already made the request and it was accepted last night. I will do the next push :)

I do not have a developer account, but I can do it. This is not a problem because I will contribute again.

• C++: test with clang:: in an attribute

@dhaumann The difference is in the dynamic attribute of a context. Kate propagates the captures only if the value is true whereas SyntaxHighlighting never looks at that value and only uses the dynamic attribute of the rules.

• Awk: add collating symbols and equivalence classes

There is indeed a problem with Kate, but it works with KSyntaxHighlighting. I will look in what part of doxygenlua.xml the problem is.

The page on Phabricator is much better than before. My problem was about creating a branch without the -t option which requires me to specify a commit/branch to do the diff.

I am completely lost with arc diff. The command adds files that I did not normally modify in this branch and specifiying a reference commit does not seem to work.

• SQL: various improvements and fix if/case/loop/end detection with SQL (Oracle)
• add sql tests
• update kateversion to 5.0 and allows multiple spaces for "end if", "end case" and "end loop"
• autotests/input/highlight.sh.syntax and autotests/input/highlight.sh are restored

• diff with the proper branch

• autotests/input/highlight.sh.syntax and autotests/input/highlight.sh are restored

• SQL: various improvements and fix if/case/loop/end detection with SQL (Oracle)
• add sql tests
• update kateversion to 5.0 and allows multiple spaces for "end if", "end case" and "end loop"

New suggestions:

• Regex "^xyz\b" -> WordDetect with column=0
• Regex "^xyz" with xyz a string of more than 2 characters -> StringDetext with column=0

More suggestions and fixes some false positives.

There is a bug on which I fell while completing the tests: the output is not stable.

I just tried with a format containing 5000 <RegExpr attribute="Keyword" context="\#stay" String="\baa\b"/> rules. This gives 2000 to 6000 more allocation than with WordDetect. Same for StringDetect vs Detect2Chars.

I've already planned to fix all the suggestions, but I lost a lot of time updating the sql*.xml files and came across a very strange bug related to QRegularExpression by adding some tests. At the moment, I do not know if it's my tool to test, my version of Qt that is old or a real bug. Like last week my hard drive is dead, I took the opportunity to have a more recent configuration. I will make a few patches this weekend when I finished setting my new environment.

It seems that KTextEditor does not have the problem because KateHlWordDetect is based on KateHlStringDetect which uses the inSensitive parameter.

Pony: Support for attributes and best behavior for function definitions. Add license and removed autotests/foldingtest.pony

This commit is already merged.