Not really an issue ... in the readme the tag for namespaces is not closed/ended properly as namespace is missing the s at the end. It only causes a problem for lazy people. like me, that cut and paste.

Since I am here, is there a way in the configuration to skip the tests ... i tried true in the configuration section of the plugin without any success

Thanks

Thanks

I have a multi-module build where some modules are using the clojure-maven-plugin. When I run maven from the module's directory, everything works fine. E.g.

[INFO] [clojure:test {execution: default}]

Testing tests

Ran 1 tests containing 2 assertions.
0 failures, 0 errors.
[INFO] ------------------------------------------------------------------------ [INFO] BUILD SUCCESSFUL [INFO] ------------------------------------------------------------------------

When I run the build from the project's root directory, I get this error message:

[INFO] [clojure:test {execution: default}] [INFO] ------------------------------------------------------------------------ [ERROR] BUILD ERROR [INFO] ------------------------------------------------------------------------ [INFO] testScript is empty or does not exist! [INFO] ------------------------------------------------------------------------

It seems that clojure-maven-plugin uses the CWD instead of the module's directory.

Situation:
1. Using the task clojure:swank i always get an NumberFormatException while clojure is parsing the protocolVersion "2009-09-14".
2. It seems the that the <"> around the string are vanishing
3. Looking deeper you find: ClojureSwankMojo generates a clojure expression that is later on consumed by clojure.main as command line argument. Part of this is:

    StringBuilder sb = new StringBuilder();
    sb.append("(do ");
    sb.append("(swank.swank/ignore-protocol-version \"");
    sb.append(protocolVersion);
    sb.append("\") ");
    sb.append("(swank.swank/start-server \"");
    sb.append(swankTempFile.getAbsolutePath());
    sb.append("\" :port ");
    sb.append(Integer.toString(port));
    sb.append(" :dont-close true");
    sb.append("))");

But i should look like this:

    StringBuilder sb = new StringBuilder();
    sb.append("(do ");
    sb.append("(swank.swank/ignore-protocol-version \\\"");
    sb.append(protocolVersion);
    sb.append("\\\") ");
    sb.append("(swank.swank/start-server \\\"");
    sb.append(swankTempFile.getAbsolutePath().replace("\\", "/"));
    sb.append("\\\" :port ");
    sb.append(Integer.toString(port));
    sb.append(" :dont-close true");
    sb.append("))");

1. The new line

   sb.append(swankTempFile.getAbsolutePath().replace("\\", "/"));
   

is necessary on windows system, where the temp-File looks like "c:\DOKUME~1\J10839~1.V10\LOKALE~1\Temp\slime.364" with for instance \D gets wrongly interpreted.

Since Clojure supports nested namespaces like Java, does it make sense to recurse into source directories and compile the .clj files in nested subdirectories instead of just the root level? It seems like the maven convention is to point the compiler at a root source folder and assume everything "under there" is code that can be compiled.

As it stands, one must override the configuration to specify each subdirectory that contains Clojure source. This isn't a huge deal given that there usually wouldn't be as much nesting as in a typical Java project due to the more coarse-grained modularity of Clojure source files. However, if the goal is to behave more like other maven plugins, I believe this proposal would move the plugin closer to the principle of least surprise.