Bug 210432

Summary:	BodyDeclaration.getJavadoc() returns null if javadoc follows an annotation
Product:	[Eclipse Project] JDT	Reporter:	Mike Yawn <mike>
Component:	Core	Assignee:	Frederic Fusier <frederic_fusier>
Status:	VERIFIED INVALID	QA Contact:
Severity:	normal
Priority:	P3	CC:	eric_jodet, jerome_lanneluc, Olivier_Thomann
Version:	3.2.2
Target Milestone:	3.4 M4
Hardware:	PC
OS:	Windows 2000
Whiteboard:

Description Mike Yawn

2007-11-20 13:21:17 EST

Build ID: 3.2.2

Steps To Reproduce:
1. Create source file with method containing both an annotation and a Javadoc comment
2. Parse the AST of the method
3. If the Javadoc precedes any annotations, BodyDeclaration.getJavadoc() returns it.  If the Javadoc comes between annotations and method declaration, .getJavadoc returns null


More information:
This parses correctly:
	/** Javadoc on ID field  */
	@Id
	@GeneratedValue
	public int getId();

This does not:
	@Id
	@GeneratedValue
	/** Javadoc on ID field  */
	public int getId();

Comment 1 Olivier Thomann

2007-11-20 14:27:14 EST

I would expect the javadoc to be placed before the start of the method. The modifiers (annotation) are part of the method declaration.
Therefore I would expect the javadoc to be located before the first modifier. I don't find the actual behavior as being boggus.

We should actually see if the apt type model reports a javadoc for such a declaration.

Frederic, any comment?

Comment 2 Eric Jodet

2007-11-20 16:54:42 EST

(In reply to comment #1)
> I would expect the javadoc to be placed before the start of the method. The
> modifiers (annotation) are part of the method declaration.
> Therefore I would expect the javadoc to be located before the first modifier. I
> don't find the actual behavior as being boggus.

Tested the javadoc tool behavior:
- with the javadoc comment before the first modifier (1st syntax), the comment is correctly generated.
- with the javadoc comment located between the annotations and the public modifier (2nd syntax) , no javadoc gets generated.

Comment 3 Mike Yawn

2007-11-20 17:03:12 EST

This is looking like it might be a user misunderstanding on my part of what legal syntax is.   I'm curious (if anyone knows), what happens to the 'misplaced' Javadoc in this case?  Is it demoted to a regular-style comment, or is it just thrown away?

Comment 4 Frederic Fusier

2007-11-21 01:57:19 EST

Your test case is like writing:
public /** comment */ int /** xxx */ getId();

In this case you'd never expect a javadoc to be generated by javadoc tool or parsed by the Eclipse compiler (Eric's comment 2 confirmed that). So, I agree with Olivier that we behave correctly and then consider this bug as INVALID.

(In reply to comment #3)
> This is looking like it might be a user misunderstanding on my part of what
> legal syntax is.   I'm curious (if anyone knows), what happens to the
> 'misplaced' Javadoc in this case?  Is it demoted to a regular-style comment, or
> is it just thrown away?
> 
Nothing is done with this comment. It's just available in the comments list of the  compilation unit (you can see it in the ASTView), but it's not attached to any node, so not really usable...

Comment 5 Jerome Lanneluc

2007-12-11 09:33:01 EST

Verified for 3.4M4 using I20071210-1800