Retaining whitespace in comments

Here's a testcase that shows the potentially desired behavior:

    public void testWhitespaceCanBeRetainedInJavadoc() {
        String sourceCode = ""
                + "package com.acme.thing;\n"
                + "\n"
                + "/**\n"
                + " * This class does something.\n"
                + " *     chalala\n"
                + " *         cha  **  lala\n"
                + " **/\n"
                + "public class AClassName {\n"
                + "}";
        JavaDocBuilder builder = new JavaDocBuilder();
        builder.addSource(new StringReader(sourceCode));
        JavaClass aClass = builder.getClassByName("com.acme.thing.AClassName");
        assertEquals("This class does something.\n"
                + "    chalala\n"
                + "    cha  **  lala", aClass.getComment());
    }

— as you can see, I think the concern is mainly about leading whitespace on each line.

Gregory, look at line 211- 226 in lexer.flex.

I think the stripping of spaces is baked into QDox at the parser stage. It would most likely need to be a change for QDox rather than a switch.

In AbstractJavaEntity.commentHeader(..) there is an attempt to format again in /** Javadoc Style */ but it does not add back the spaces.

I've had a look at the lexer and I think it's a bit too complecated right now.
Why not read per line instead of per word? Everything after a javaDocComment-linestart can be copied as it is, remaining spaces, asterisks, etc.
The only reason I can think of is detecting doclet-tags.
For the doclet case you could use javaDocComment-linestart followed by an '@', push state and read until the space and pop back.

Are there cases which won't fit this strategy?

If this seems solid enough, then most lexer-tests specific for javadocs have to be rewritten. Any ideas?

Take a look at the test below. I think I found a way to parse javadoccomments, but it looks like both tests collide with each other.
Unless we say: JavadocTags should remove leading spaces, comments not.
But then there´s another thing: in the chalala-example I think the last line should contain the extra 4 spaces in front of 'cha'.

Any ideas?

    public void testTagValuesCanSpanMultipleLines() {
        String source = "" +
                "/**\n" +
                " * @bar.baz foo=\"this is\\\n" + 
                " *       multilined\"\n" +
                " */\n" +
                "class x{}";
        builder.addSource(new StringReader(source));
        JavaClass x = builder.getClassByName("x");
        DocletTag tag = x.getTagByName("bar.baz");
        assertEquals("foo=\"this is\\\nmultilined\"", tag.getValue());
        assertEquals("this is\nmultilined", tag.getNamedParameter("foo"));
    }

Issue resolved like mentioned: add those 4 extra spaces so the comment stays as it is.