<< Back to previous view

[CLJ-2048] java.lang.ClassCastException: [Ljava.lang.Object; cannot be cast to [Ljava.lang.StackTraceElement Created: 21/Oct/16  Updated: 21/Oct/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.8, Release 1.9
Fix Version/s: None

Type: Defect Priority: Major
Reporter: Gerrit Jansen van Vuuren Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: None

Attachments: Text File CLJ-2048-b.patch    
Patch: Code
Approval: Prescreened


clojure.core/throw-if creates an array to call Exception.setStracktrace() without specifying the array type. This works fine when passed at least one StackTraceElement, but in the case where passed no stack trace elements (all are filtered or stack traces are being elided by the JVM), this will be an Object[] which results in a ClassCastException:

Exception in thread "main" java.lang.ClassCastException: [Ljava.lang.Object; cannot be cast to [Ljava.lang.StackTraceElement;
        at clojure.core$throw_if.invokeStatic(core.clj:5649)
        at clojure.core$load_one.invokeStatic(core.clj:5698)
        at clojure.core$compile$fn__5682.invoke(core.clj:5903)
        at clojure.core$compile.invokeStatic(core.clj:5903)
        at clojure.core$compile.invoke(core.clj:5895)
        at clojure.lang.Var.invoke(Var.java:379)
        at clojure.lang.Compile.main(Compile.java:67)

This is tricky to reproduce because it involves stack trace filtering so there is no reproducing case here.

Patch: CLJ-2048-b.patch
Prescreened by: Alex Miller

Comment by Gerrit Jansen van Vuuren [ 21/Oct/16 4:18 AM ]

patch calls into-array with StackTraceElement type

Comment by Alex Miller [ 21/Oct/16 8:01 AM ]

How do you cause this to occur?

Comment by Alex Miller [ 21/Oct/16 8:11 AM ]

into-array will create a typed array based on the first element of the seq it is passed, so generally you should see a StackTraceElement[] here. I think the only time this would fail is if it was passed no stack trace elements.

Comment by Alex Miller [ 21/Oct/16 8:19 AM ]

I'd be happy to move this through screening, but the patch needs to be in the proper format (see http://dev.clojure.org/display/community/Developing+Patches).

Comment by Gerrit Jansen van Vuuren [ 21/Oct/16 8:42 AM ]

I'm trying to reproduce this in a way that can be presented here, but I got the compile error just after doing some serious package renaming, and can't reproduce it outside of the project itself.

Comment by Gerrit Jansen van Vuuren [ 21/Oct/16 8:45 AM ]

ok, I'll reformat the patch after reading (http://dev.clojure.org/display/community/Developing+Patches)

Comment by Gerrit Jansen van Vuuren [ 21/Oct/16 9:15 AM ]

I've created a new patch based on the guidelines, attached as file: CLJ-2048-b.patch.

Just to summarise:
The into-array returns the correct type if its provided with a none empty sequence, but if the sequence is empty it cannot know the type and then returns an object array. Because we set the array here to a java method Exception::setStackTrace passing it an object array causes a ClassCastException. One fix is to check for an empty sequence, but a less verbose way is just to pass the type which is known as part of the call to into-array, this is what is done in the patch CLJ-2048-b.patch.

[CLJ-2046] generate random subsets of or'd required keys in map specs Created: 17/Oct/16  Updated: 17/Oct/16

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

Type: Enhancement Priority: Major
Reporter: David Chelimsky Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: generator, spec

Attachments: Text File map-spec-gen-enhancements.patch    
Patch: Code and Test
Approval: Screened


(s/keys :req [(or ::x ::y)]) always generates maps with both ::x and ::y but it should also generate maps with either ::x or ::y.

The attached patch supports arbitrarily deeply nested or and and expressions within the values of :req and :req-un in map specs. It also uses the same 'or' mechanism for :opt and :opt-un keys, thereby replacing the use of clojure.core/shuffle with clojure.test.check.generators/shuffle, ensuring repeatability of the generators.

Patch: map-spec-gen-enhancements.patch

Screened by: Alex Miller

[CLJ-2045] bean function not working Created: 17/Oct/16  Updated: 17/Oct/16  Resolved: 17/Oct/16

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

Type: Defect Priority: Major
Reporter: Roy Varghese Assignee: Unassigned
Resolution: Duplicate Votes: 0
Labels: None

(defproject hello-jcloud "0.1.0-SNAPSHOT"
:description "FIXME: write description"
:url "http://example.com/FIXME"
:license {:name "Eclipse Public License"
:url "http://www.eclipse.org/legal/epl-v10.html"}
:dependencies [#_[org.clojure/clojure "1.8.0"]
[org.clojure/clojure "1.9.0-alpha13"]
[org.clojure/tools.logging "0.1.0"]
[org.apache.jclouds/jclouds-all "2.0.0-SNAPSHOT"]]
:repositories {"jclouds-snapshots" "https://repository.apache.org/content/repositories/snapshots"})


user> (bean (java.util.Date.))
UnsupportedOperationException empty clojure.core.proxy$clojure.lang.APersistentMap$ff19274a.empty (:-1)

It works when clojure/1.8.0 is uncommented.

Comment by Alex Miller [ 17/Oct/16 7:38 AM ]

This is a dupe of http://dev.clojure.org/jira/browse/CLJ-2027 which will probably be in the next alpha.

[CLJ-2044] Four functions in clojure.instant have incomplete documentation Created: 15/Oct/16  Updated: 17/Oct/16

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.4, Release 1.5, Release 1.6, Release 1.7, Release 1.8, Release 1.9
Fix Version/s: None

Type: Enhancement Priority: Major
Reporter: Bruce Adams Assignee: Bruce Adams
Resolution: Unresolved Votes: 0
Labels: docstring, instant

Attachments: Text File defns-for-instant.patch    
Patch: Code
Approval: Triaged


Of the five public functions defined in clojure.instant, these four:

  • parse-timestamp
  • read-instant-calendar
  • read-instant-date
  • read-instant-timestamp

are each declared as a Var without any arglists metadata. This means documentation does not contain function calling information.

In http://clojure.github.io/clojure/clojure.instant-api.html, each of these functions is described as a var and there is no Usage: ... information given.

The output of doc does not include argument list information. For example:

user=> (doc clojure.instant/read-instant-date)
  To read an instant as a java.util.Date, bind *data-readers* to a map with
this var as the value for the 'inst key. The timezone offset will be used
to convert into UTC.

A related problem is that stack traces show anonymous functions instead of the names for any of these functions. For example:

user=> (clojure.instant/read-instant-timestamp "123")
RuntimeException Unrecognized date/time syntax: 123  clojure.instant/fn--6879/fn--6880 (instant.clj:107)

Proposed: Refactor the code into defn functions which makes the code clearer and addresses the documentation issue. An alternate approach would be to apply :arglists metadata to the vars.

Patch: defns-for-instant.patch

Comment by Bruce Adams [ 15/Oct/16 12:40 PM ]

Proposed solution: refactor the definitions of the four problematic functions to be defined using defn.

Comment by Bruce Adams [ 16/Oct/16 5:24 PM ]

Some of my thinking leading to the solution I propose.

Initially, when I realized that I didn't know what arguments parse-timestamp required, I assumed the appropriate fix was to enhance the docstring. Then I noticed that the on-line documentation for functions was formatted quite differently from the text output by doc. Any decent fix was going to have to feed function information into different variants of documentation formatting code.

I can guess, from other examples such as first, that :arglists metadata is what indicates that a var is a function. One possible solution would be to add :arglists to each of the four functions. It felt cleaner to refactor the code into simple defn functions. Refactoring code just for the side effect of documentation seems a bit odd, but the code itself strikes me as more legible after my refactoring.

Comment by Alex Miller [ 17/Oct/16 9:53 AM ]

This seems reasonable to me. I would move the timestamp regex into a separate (private) var instead of creating it in parse-timestamp.

It's possible the way these functions were defined was designed to minimize startup time or something like that, but I don't have any background on that.

Generated at Fri Oct 21 22:46:39 CDT 2016 using JIRA 4.4#649-r158309.