#846 hare test: add a way to disable color output 3 hours ago

Comment by ~sebsite on ~sircmpwn/hare

Dmitry Matveyev referenced this ticket in commit 0429880.

REPORTED RESOLVED IMPLEMENTED

#807 allow using static assert/abort in global context 3 hours ago

Comment by ~sebsite on ~sircmpwn/hare

REPORTED RESOLVED IMPLEMENTED

#897 Attribute lexing differs in harec and spec 3 hours ago

easy added by ~sebsite on ~sircmpwn/hare

#897 Attribute lexing differs in harec and spec 3 hours ago

spec added by ~sebsite on ~sircmpwn/hare

#897 Attribute lexing differs in harec and spec 6 hours ago

Ticket created by ~sebsite on ~sircmpwn/hare

938ac7b adds attributes as tokens, however, it still doesn't match the behavior of harec. Since @ is only allowed within a token when it exactly matches one of the listed attributes, it follows that any trailing characters aren't part of the token, and should be separated. Thus, according to the spec, the following code is valid:

@testfn foo() void = void;
let @threadlocalbar = 0;

#896 Harec CI improvements 21 hours ago

Comment by ~sebsite on ~sircmpwn/hare

We may also want to use -fsanitize=undefined

#893 formatting issues in the syntax summary appendix 3 days ago

easy added by ~sebsite on ~sircmpwn/hare

#893 formatting issues in the syntax summary appendix 3 days ago

Comment by ~sebsite on ~sircmpwn/hare

This isn't an issue with the syntax summary appendix, this is an issue with the nonterminaldefs themselves. Most lines for nonterminals end with \\, but some don't (like in lexical analysis, and ones that are immediately followed by \end{grammar}). So we should go through the spec and make sure we're consistently using \\ everywhere we should be.

#894 Branch result type reduction does not handle if expressions correctly 7 days ago

Comment by ~sebsite on ~sircmpwn/hare

Allowing intermediate results to be (null | str) without just allowing null types on their own in general would be, very weird, and probably not a good idea. I think we should either just make null a regular type, or say that this is intended behavior.

#894 Branch result type reduction does not handle if expressions correctly 7 days ago

Comment by ~sebsite on ~sircmpwn/hare

If we make null an actual representable type then we can just say that at the end of type reduction, if there's one pointer and a null type, it turns into a nullable pointer, but otherwise the null type is still allowed.