The interp command recognizes "foo" and "::foo" as different interpreter names. However, the two interpreters have the same interpreter command. When "::foo" is created below, the command for interpreter "foo" is overwritten and so interpreter "foo" is deleted.
The problem is that, because "interp" creates a command for each interpreter, its rules for name resolution should not differ from the rules for name resolution of commands.
Suggested remedy: "interp" should recognize interpreter names as equal only if their fully-qualified namespace names are equal.
When there are multiple levels of interpreters, e.g. a slave ::foo with its own slave {::foo ::bar}, the name in the master interpreter is a list, and each element in the list should be compared with its peers by their fully-qualified namespace names. This ensures the same outcome if interpreter {::foo ::bar} is created in the master, as if interpreter ::bar is created in the slave ::foo.
The usual, expected behavior:
% interp create foo
foo
% interp create foo
interpreter named "foo" already exists, cannot create
The unexpected behavior:
% interp create foo
foo
% interp create ::foo
::foo
% interp slaves
::foo
Check that the interpreter really has been replaced:
% interp create foo
foo
% foo eval set a abc
abc
% foo eval set a
abc
% interp create ::foo
::foo
% interp slaves
::foo
% foo eval set a
can't read "a": no such variable
For maximum surprise, move the first interpreter's command so that interpreters foo and ::foo can coexist:
% interp create foo
foo
% interp eval foo set a abc
abc
% interp eval foo set a
abc
% rename foo moved
% interp create ::foo
::foo
% interp slaves
::foo foo
% interp eval foo set a
abc
% interp eval ::foo set a
can't read "a": no such variable
% moved eval set a
abc
|