Tab completion with Leiningen REPL after typing a period in a namespace triggers an NPE #2380
Comments
|
Thanks for reporting this. It's probably a bug in the clojure-complete library, so it will need to be fixed in https://github.com/ninjudd/clojure-complete/issues
|
|
Great, I've opened an issue over there. |
|
I tried reproducing this but it works fine for me. I tried a test on clojure-complete ➜ foo-2380 lein repl
nREPL server started on port 50500 on host 127.0.0.1 - nrepl://127.0.0.1:50500
REPL-y 0.3.7, nREPL 0.2.12
Clojure 1.8.0
Java HotSpot(TM) 64-Bit Server VM 1.8.0_102-b14
Docs: (doc function-name-here)
(find-doc "part-of-name-here")
Source: (source function-name-here)
Javadoc: (javadoc java-object-or-class-here)
Exit: Control+D or (exit) or (quit)
Results: Stored in vars *1, *2, *3, an exception in *e
user=> clojure.core
clojure.core clojure.core.ArrayChunk clojure.core.ArrayManager clojure.core.Eduction clojure.core.IVecImpl
clojure.core.Vec clojure.core.VecNode clojure.core.VecSeq clojure.core.protocols clojure.core.protocols.CollReduce
clojure.core.protocols.IKVReduce clojure.core.protocols.InternalReduce clojure.core.server
user=> clojure.core
➜ foo-2380 lein -v
Leiningen 2.8.0 on Java 1.8.0_102 Java HotSpot(TM) 64-Bit Server VMclojure-complete test (is (= '("clojure.core" "clojure.core.ArrayChunk" "clojure.core.ArrayManager" "clojure.core.IVecImpl" "clojure.core.Vec" "clojure.core.VecNode" "clojure.core.VecSeq" "clojure.core.protocols" "clojure.core.protocols.InternalReduce") (completions "clojure.co"))) |
|
Thanks for the follow-up - I noticed you're running Java 8 and that seems to be the trigger factor here (I'm on JDK 9). I tested it out with Java 8 and the bug did not reproduce so there's got to be some change in Java 9's classloading behavior that's causing this. |
|
@venantius Can confirm on Java 9 on Ubuntu $ lein repl
nREPL server started on port 38202 on host 127.0.0.1 - nrepl://127.0.0.1:38202
REPL-y 0.3.7, nREPL 0.2.12
Clojure 1.8.0
OpenJDK 64-Bit Server VM 9-internal+0-2016-04-14-195246.buildd.src
Docs: (doc function-name-here)
(find-doc "part-of-name-here")
Source: (source function-name-here)
Javadoc: (javadoc java-object-or-class-here)
Exit: Control+D or (exit) or (quit)
Results: Stored in vars *1, *2, *3, an exception in *e
user=> clojure.coNullPointerException complete.core/fn--647/iter--639--648/fn--649 (core.clj:66)
CompilerException java.lang.ClassNotFoundException: clojure.co, compiling:(null:0:0)
user=>As @technomancy mentioned it's a problem with upstream clojure-complete on Java 9. I tried the test which passed on Java 8, it fails on Java 9 as below : Test case : (is (= '("clojure.core" "clojure.core.ArrayChunk" "clojure.core.ArrayManager" "clojure.core.IVecImpl" "clojure.core.Vec" "clojure.core.VecNode" "clojure.core.VecSeq" "clojure.core.protocols" "clojure.core.protocols.InternalReduce") (completions "clojure.co")))$ clj
Clojure 1.9.0
user=> (import [java.io File])
java.io.File
user=> (for [prop ["sun.boot.class.path" "java.ext.dirs" "java.class.path"]
path (.split (System/getProperty prop) File/pathSeparator)] path)
NullPointerException user/eval171/iter--165--172/fn--173 (NO_SOURCE_FILE:5)
user=> (.split (System/getProperty "sun.boot.class.path") File/pathSeparator)
NullPointerException user/eval192 (NO_SOURCE_FILE:7)
user=> (.split (System/getProperty "java.ext.dirs") File/pathSeparator)
NullPointerException user/eval194 (NO_SOURCE_FILE:8)
user=> (.split (System/getProperty "java.class.path") File/pathSeparator)
#object["[Ljava.lang.String;" 0x78de58ea "[Ljava.lang.String;@78de58ea"] |
|
While we've merged a fix to clojure-complete, @ninjudd has warned us that the library is no longer maintained. I don't yet have deploy rights to Clojars, so I don't have a way of actually releasing the fix. It would seem to me that an alternative approach here would be to migrate Leiningen's dependency off of clojure-complete to https://github.com/alexander-yakushev/compliment, which is actively maintained. |
|
@venantius I think leiningen uses reply for repl which in turn uses clojure-complete. I couldn't see clojure-complete being used directly in leiningen. The issue of deprecation is already an issue in clojure-complete repo. I think it will be better if we can bump the release with the help of @ninjudd and to have reply and leiningen to migrate to compliment in the long term since they may take significant changes. I will open an issue in reply to migrate to compliment. |
|
@tirkarthi Agree that's the right thing to do short and long term. I took a quick look at what would need changing in reply and it looks like a fair bit of work, otherwise I would have tried to fix it this afternoon. Opening an issue with them is a good start. |
|
@venantius I have opened up an issue in reply trptcolin/reply#176 . There is already an issue in clojure-complete about deprecation ninjudd/clojure-complete#18 . I will take a stab at migrating reply to compliment over the weekend 👍 |
|
@venantius Seems like there is already a three year old PR to replace clojure-complete with compliment : trptcolin/reply#153 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Bit of an oddly specific bug, I know.
TL;DR - if I type
clojure.coand then try to tab-complete, an NPE is triggered. Tab-completion works fine if I hit tab atclojor atclojure.core/, but after that first period and before the slash I'll get NPE'd. I ran into this on a project and then was able to reproduce outside of a project context.Now the expected behavior would be for tab completion to work the way it did when I hit tab at
cloj, i.e. to show me the list of namespaces that are potential results for what I've typed so far.Full command history and stacktrace:
The text was updated successfully, but these errors were encountered: