[sherenkov]

unoptimizable-recursion recognised the plain-name self-call of a top-level object only as Φ.<name>. In a program with a +package meta, eo-parser 0.63.0 puts the top-level object into its package, so calc (n.minus 1) inside [n] > calc of +package foo becomes base="Φ.foo.calc". No self-call matched, and the lint reported nothing for such a program, although the compiler does not turn Φ.foo.calc into a loop either.

The sheet now reads the package from the +package meta, and a new eo:plain function gives both spellings of a plain name, Φ.<name> and Φ.<package>.<name>. eo:is-self-call, eo:is-self-dispatch and eo:is-self-direct use it where they compared with concat('Φ.', $name), so the three shapes of the plain-name self-call are reported in a packaged program as they are in a program without a package.

New packs in packs/single/unoptimizable-recursion/, all with +package foo:

• catches-plain-name-tail-call-in-a-package.yaml: the example from the issue, which reported no defect before and now gets the "names the formation directly" warning on line 7;
• catches-plain-name-self-call-in-a-package.yaml: a self-call under n.mul, reported as "no self-call sits in a tail position";
• catches-a-dispatch-on-the-plain-name-self-call-in-a-package.yaml: calc.plus 1, reported as "the self-call is the receiver of a dispatch";
• allows-tail-call-in-an-if-branch-in-a-package.yaml: a nested ^.rec tail call, which still gets no defect.

The first three fail on master and pass with this change. mvn test -Dtest='LtByXslTest#checksAllLints' passes.

Closes #1687.

[sherenkov]

@yegor256 CI is green and the branch is up to date with master, could you take a look when you have a moment? It makes unoptimizable-recursion recognise the plain-name self-call of a top-level object in a program with +package, which the parser spells Φ.<package>.<name> and the lint used to miss.

[yegor256]

The change matches issue #1687 and the new packs cover the reported case and keep neighbouring behavior intact.