Head towards 4.11 release

[parrt]

Heh, we have to get 4.11 out so that people can use the amazing Go target what day is done by @jimidle.

Stuff to do:

• Make sure to generate a Windows dotnet package from Windows to get the signing right
• Update antlr.org quick start guide to use https://github.com/antlr/antlr4-tools
• @jimidle any more Go Target stuff that needs to go in?
• Make sure 4.11.0 tag is also they're not just 4.11 (Go requirement)
• C++ issue regarding EOF symbol: [Cpp] Can't use antlr and cstdio together because of undef EOF #2914
• @KvanTTT's Constant folding for Python and other runtimes #3701
• Dependency declaration error in ANTLR 4.10.1 #3733 ST uses 3.5.2 antlr not most recent.
• Fix old error Cpp, Go, JavaScript, Python2/3: Template rendering error. #2016, merging Fix 2016 [WIP] #2022
• Fix 2016 round 2 #3841
• Cpp: Link to threads library #3794

What else needs to go in 4.11?

@jimidle @ericvergnaud @mike-lischke @marcospassos @pboyer, @janyou, @ewanmellor, @lingyv-li have comments?

[KvanTTT]

I wish the following issues were resolved:

• ATN optimization: removing of redundant epsilon transitions and related basic states #3676 It decreases the size of generated files and improves performance
• New action-related warnings: Warnings ACTION_SHOULD_BE_PLACED_AFTER_PREDICATES, PREDICATE_AT_THE_BEGINNING_OF_LEXER_RULE_DEGRADES_PERFORMANCE and misc fixes #3626 (it's better to reimplement them on ATN instead of grammar)
• Fixes related to "mutually left recursive rules" errors: Report all "mutually left recursive rules" errors and fix positions #3779
• Last but not least refactoring: Refactoring of grammar literals parsing #3780

[parrt]

Hmm..not sure I have the time to go through those for now. It will require a massive effort to remember all of the ATN stuff before I can evaluate those. The last one is something I probably won't accept because it's just refactoring and incurs a risk.

[lingyv-li]

🙌 Nothing to add for Dart.

[mike-lischke]

The EOF macro problem is tricky to solve. Either you annoy users who use C libs with ANTLR4 or you annoy existing project owners if you rename that symbol. Given that there are only very few complains I tend to keep the current solution. For the threads patch @jcking can probably help.

[parrt]

Yeah, breaking all existing users' code that ref EOF is problematic; i.e., we shouldn't rename. The undef trick @jcking suggested might not even be enough since someone can stick an include of C stdio.h after every include file. Wouldn't it be reasonable to ask the few users that need C libs to add #undef EOF right after ever #include that pulls in a EOF macro? Let me see if I can do a test.

[marcospassos]

I don't think those issues impact the PHP target 🚀

[jimidle]

I will try to get in the remaining fixes (such as test failures) in the next two days, then I think the go runtime is good for 4.11 and I can move to more performance analysis