Add :EXTENSIBLE-SEQUENCES to *FEATURES*

[Michał "phoe" Herda]

While working on the test suite of split-sequence at https://github.com/sharplispers/split-sequence/pull/24#discussion_r549261736 I discovered that there's no way that would allow code to figure out whether an implementation supports extensible sequences. This is no longer a trivial question, since both SBCL and ABCL support that extension, and new implementations can soon follow suit.

Does it sound feasible to add `:extensible-sequences` in `*features*` to solve this?

[Douglas Katzman (dougk)]

SBCL has had the SEQUENCE package for >= 15 years, and I'm disinclined to add "equivalent" features.
Personally I'd suggest that your code do #+sbcl (pushnew :extensible-sequence *features*) if you think it will help either your system or dependents on your system.
Granted, the spec says that ":ieee-floating-point" is present if IEE754 is supported, despite it being standard since 1985. But I think the distinction is that there isn't a simple way to detect use of non-IEE754 floats, and SBCL could theoretically support whatever that other thing is. Whereas, an extra feature for sequences would mean literally no more and no less than the :sbcl feature already conveys.

If others developers disagree with me, I won't object to adding it; but I don't see it as a good solution to anything, just an annoyance, as I'm already striving to _remove_ extraneous features.

[Michał "phoe" Herda (phoe-krk)]

I'm fine with that; all I want is a reliable way to check if the extensible sequences API is available on a given Lisp image. The hypothetical TRIVIAL-EXTENSIBLE-SEQUENCES can do the aforementioned #+sbcl check and PUSHNEW the feature e.g. when it finds the SEQUENCE package. Does this sound OK, or is there a better way?

[Stas Boukarev (stassats)]

Adding :extensible-sequences now would mean all older SBCL will stop supporting it.

[Michał "phoe" Herda (phoe-krk)]

OK. I'll check for #+sbcl (find-package "SEQUENCE") and push the feature myself in my own TRIVIAL- system instead.

Ticket closeable as "invalid", I guess.