<< Back to previous view

[CLJ-1970] instrumented macros never conform valid forms Created: 25/Jun/16  Updated: 25/Jun/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.9
Fix Version/s: None

Type: Defect Priority: Major
Reporter: OHTA Shogo Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: spec


 Description   

Although macros can be speced without &form and &env, once they are instrumented they will try to conform the args including &form/&env and fail:

user=> (require '[clojure.spec :as s])
nil
user=> (defmacro m [x] x)
#'user/m
user=> (s/fdef m :args (s/cat :arg integer?) :ret integer?)
user/m
user=> (m 1)
1
user=> (m a)
CompilerException java.lang.IllegalArgumentException: Call to user/m did not conform to spec:
In: [0] val: a fails at: [:args :arg] predicate: integer?
:clojure.spec/args  (a)
, compiling:(NO_SOURCE_PATH:5:1) 
user=> (s/instrument)
[user/m]
user=> (m 1)
ExceptionInfo Call to #'user/m did not conform to spec:
In: [0] val: (m 1) fails at: [:args :arg] predicate: integer?
:clojure.spec/args  ((m 1) nil 1)
  clojure.core/ex-info (core.clj:4718)
user=>

To resolve the situation, I think instrument/instrument-all should avoid speced macros.



 Comments   
Comment by Alex Miller [ 25/Jun/16 9:42 AM ]

This is an issue we're discussing - for the moment, you should not instrument macros. There is no point in instrumenting them as they are automatically checked during macroexpansion.

Comment by OHTA Shogo [ 25/Jun/16 10:00 AM ]

Yes, I know the compiler checks macro specs automatically, but just thought it would be nice if explicit calls to instrument (with no args) and instrument-all would check whether or not each speced var is a macro and filter it out if so.

Comment by Alex Miller [ 25/Jun/16 2:30 PM ]

Totally agreed.





[CLJ-1969] :as form is unbound when no optional keyword arguments is passed even though :or form is provided Created: 24/Jun/16  Updated: 24/Jun/16  Resolved: 24/Jun/16

Status: Closed
Project: Clojure
Component/s: None
Affects Version/s: Release 1.7
Fix Version/s: None

Type: Defect Priority: Trivial
Reporter: Lars Andersen Assignee: Unassigned
Resolution: Declined Votes: 0
Labels: None


 Description   
(defn fn-with-kw-opts [& {:keys [opt] :or {opt 1} :as options}]
  (println "opt " opt "options " options))

user> (fn-with-kw-opts)
;;=> opt  1 options  nil

user> (fn-with-kw-opts :opt 2)
;;=> opt  2 options  {:opt 2}

I would expect options to be bound to the default value when no keyword argument is passed.



 Comments   
Comment by Alex Miller [ 24/Jun/16 7:36 AM ]

:as binds to the original value passed in and will never include values from :or. :or is used to provide defaults for each local being bound when that local is missing in the input.

In the case of (fn-with-kw-opts), the incoming value is nil so options is bound to nil.

Comment by Alex Miller [ 24/Jun/16 7:38 AM ]

Working as designed.





[CLJ-1968] clojure.test/report :error does not flush *out* when the test fails with an exception Created: 23/Jun/16  Updated: 23/Jun/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.8
Fix Version/s: None

Type: Defect Priority: Major
Reporter: Sam Roberton Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: clojure.test

Approval: Triaged

 Description   

Minimal reproduction:

(require 'clojure.test)

(clojure.test/deftest foo-test
  (throw (ex-info "I fail" {})))

(clojure.test/deftest bar-test
  (.println System/out "bar"))

(clojure.test/test-vars [#'foo-test #'bar-test])

Result:

ERROR in (foo-test) (core.clj:4617)
Uncaught exception, not in assertion.
expected: nil
bar
  actual: clojure.lang.ExceptionInfo: I fail
 at clojure.core$ex_info.invokeStatic (core.clj:4617)
...

Note "bar" appearing in the output in the middle of the error report for foo-test.

Analysis:

(clojure.test/report {:type :error, :actual some-exception}) calls stack/print-cause-trace. Unlike other clojure.test/report callpaths, this does not flush on newline. Thus, when tests fail with exceptions and there is anything writing directly to Java's System.out, there can be a large gap between the first part of the error report and the exception trace.

(To explain why this is annoying: we're running Selenium tests via clj-webdriver, and our system under test is logging with log4j via clojure.tools.logging. We invariably see dozens or even hundreds of lines between "expected: ..." and the subsequent "actual: ..." exception trace. This makes it very easy to come to completely the wrong conclusion about when failures occurred with respect to the other events that appear interleaved in the log.)

It would be preferable (in my opinion) if clojure.test/report always constructed the output from each individual invocation into a single string which got written to *out* all at once – that way there could be no way for output to be interleaved from other threads. Absent that, it would at least help a lot if the :error implementation called (flush).






[CLJ-1967] Enhanced namespaced map print support Created: 23/Jun/16  Updated: 23/Jun/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.9
Fix Version/s: None

Type: Enhancement Priority: Major
Reporter: Alex Miller Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: print

Approval: Triaged

 Description   

CLJ-1910 added namespaced map syntax for reader and printer but did not (yet) add these things that may be useful:

  • pprint support for namespaced maps
  • printer flag to suppress printing namespaced maps





[CLJ-1966] :clojure.spec/invalid is not a valid :clojure.spec/any value Created: 21/Jun/16  Updated: 24/Jun/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.9
Fix Version/s: None

Type: Defect Priority: Major
Reporter: Alexander Kiel Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: None


 Description   

(clojure.spec/valid? :clojure.spec/any :clojure.spec/invalid) returns false

This issue gets serious, if one likes to write specs for core functions like = which are used by spec itself. I observed this bug as I wrote a spec for assoc.

A possible solution could be to use an (Object.) sentinel internally and :clojure.spec/invalid only at the API boundary. But I have not thought deeply about this.



 Comments   
Comment by Alexander Kiel [ 24/Jun/16 9:48 AM ]

I have another example were the described issue arises. It's not possible to test the return value of a predicate suitable for conformer, because it should return :clojure.spec/invalid itself.

(ns coerce
  (:require [clojure.spec :as s]))

(s/fdef parse-long
  :args (s/cat :s (s/nilable string?))
  :ret (s/or :val int? :err #{::s/invalid}))

(defn parse-long [s]
  (try
    (Long/parseLong s)
    (catch Exception _
      ::s/invalid)))




Generated at Mon Jun 27 15:26:45 CDT 2016 using JIRA 4.4#649-r158309.