RakuAST is already bringing some breaking changes that fix old mistakes in the rakudo implementation, bringing rakudo more inline with the speculations. Currently there are two proposed regex changes that would also fall into this category:
Apply :ratchet to an atom followed by sigspace whitespace
rakudo/rakudo#6658
This one broke the following modules during a Blin run: ASN::Grammar, CSS::Grammar, CSS::Module, Grammar::Modelica, Math::Symbolic, SQL::Lexer. These can be fixed in a way that would work on older rakudo as well as a rakudo with the proposed PR merged (and I actually already have fix branches for each of these modules locally to confirm this).
edit: I've opened PRs for those 6 modules
End the declarative prefix at a frugal quantifier
Raku/nqp#872
I'm not sure what this change would break, but I suspect it would be similar number of module breakage. jnthn himself seemed to think this was a good thing to do (see the linked issue in the pull request).
We could potentially revision gate these (although I haven't confirmed the feasibility), but I'm not sure if that is the best idea for these in particular since changing it won't be obvious changing the revision changes the behavior (and thus why not just not revision gate it at all which has the same drawback). Given the RakuAST stuff is strategically breaking things to improve things from a language design perspective I'm personally inclined to say we should move forward with this sans revision gate.
RakuAST is already bringing some breaking changes that fix old mistakes in the rakudo implementation, bringing rakudo more inline with the speculations. Currently there are two proposed regex changes that would also fall into this category:
Apply :ratchet to an atom followed by sigspace whitespace
rakudo/rakudo#6658
This one broke the following modules during a Blin run: ASN::Grammar, CSS::Grammar, CSS::Module, Grammar::Modelica, Math::Symbolic, SQL::Lexer. These can be fixed in a way that would work on older rakudo as well as a rakudo with the proposed PR merged (and I actually already have fix branches for each of these modules locally to confirm this).
edit: I've opened PRs for those 6 modules
End the declarative prefix at a frugal quantifier
Raku/nqp#872
I'm not sure what this change would break, but I suspect it would be similar number of module breakage. jnthn himself seemed to think this was a good thing to do (see the linked issue in the pull request).
We could potentially revision gate these (although I haven't confirmed the feasibility), but I'm not sure if that is the best idea for these in particular since changing it won't be obvious changing the revision changes the behavior (and thus why not just not revision gate it at all which has the same drawback). Given the RakuAST stuff is strategically breaking things to improve things from a language design perspective I'm personally inclined to say we should move forward with this sans revision gate.