you are wrong on both accounts.

> It doesn't matter when the Try::Tiny is compiled.

I've explicitly tested this with feature::feature_enabled() and order definitely matters.

It's basically the difference between:

use feature "try";
use Try::Tiny;
[download]

vs

use Try::Tiny;          # no chance (well see footnote°)
use feature "try";
[download]

And that's because lexical scope starts right at the point of declaration.

> it's out of reach of Try::Tiny since I believe it's intentionally a pure-Perl module.

It's very much possible with a pure-Perl module like I said. The _enabled routines in feature are basically checking the hinthash %^H returned by caller for the needed level.

And the hint-key is called "feature_try"

A slight complication is that feature introduced "try" before it provided the check-enabled routines. IOW one needs to copy the hint-hash lookup logic for older Perl versions.

Footnotes

°) the wrong order could be fixed by using an INIT {} block doing the check after all compilations are completed, but I question the wisdom of such a fragile construct. Too many edge cases.

update

little demo v5.38

 perl -E'BEGIN{say join "\n",feature::features_enabled()}'
bareword_filehandles
bitwise
current_sub
evalbytes
fc
indirect
isa
module_true
multidimensional
postderef_qq
say
signatures
state
unicode_eval
unicode_strings
[download]