Clojure "clj" scripts mess up command line arguments

[jsg8pitt@…]

clj myscript.clj foo bar baz should have *command-line-args* internally bound to ("foo" "bar" "baz"). However, because of the implementation of the clj wrapper scripts, it ends up being bound to ("myscript.clj" "--" "foo" "bar" "baz").

This happens because $* is being used to execute the actual clojure command. However, that repeats the name of the script which was used. Preceding the invocation of clojure with a call to shift will yank the script name out of the arguments.

I can't tell what the purpose of "--" is, and it sort of strikes me that if I need that argument, then I should pass it myself, but I don't really know.

Patch attached.

[jsg8pitt@…]

Fixes the behavior of the command line args in the clojure shell wrapper.

[mdippery@…]

Looks good to commit.

[jsg8pitt@…]

So, could someone commit it then?

[alex.taggart@…]

The patch is not correct since it leaves the scriptname as the first element in the list. Just get rid of the scriptname altogether:

else
  exec java -classpath $cp $cljclass $*
fi

[mdippery@…]

Okay, I've written up a new patch which should fix this latest issue.

I also removed the jline variants since it's functionally the same as the rlwrap variant, but the rlwrap variant does it better anyway, and I was getting tired of maintaining *three* nearly identical scripts.

[mdippery@…]

New Clojure patch: Fixes issues and removes readline variant (in favor of rlwrap)

[alex.taggart@…]

Is this fix going to be applied anytime soon?

As of today:

$ tail test.clj (println *command-line-args*) $ clj test.clj foo bar (-- test.clj foo bar)

Should be (foo bar)

[mdippery@…]

Replying to alex.taggart@…:

Is this fix going to be applied anytime soon?

As of today:

$ tail test.clj (println *command-line-args*) $ clj test.clj foo bar (-- test.clj foo bar)

Should be (foo bar)

Is that output with or without the latest patches?

[alex.taggart@…]

Is this port abandoned? Can we transfer control to someone who will actually apply the fix?

This bug has been languishing here, unpatched, for months. Every time I try to encourage coworkers to try out clojure on their macs, I have to "fix" this bug, which makes the community seem like a collection of illiterates. This is not a hard bug to test.

$ echo "(println *command-line-args*)" > test.clj
$ clj test.clj foo bar
(-- test.clj foo bar)

Where the correct output should be (foo bar).

If you're going to take the one good name for a clojure port at least maintain the damn thing.

[mdippery@…]

Replying to alex.taggart@…:

Is this port abandoned? Can we transfer control to someone who will actually apply the fix?

I've authorized the patch. I don't have commit access to the MacPorts SVN repo.

[jsg8pitt@…]

Do you need to do something in particular then to get the patch committed?

I see that past changes to the Portfile have been made in the past by ryandesign, macsforever2000, blb and jmr. Can one of them help here?

[mdippery@…]

I'll ping the mailing list.

[raimue (Rainer Müller)]

Committed in r69628.

[raimue (Rainer Müller)]

Fixed conflict with non-existing variant in r69629.