Bug 141800 - [1.5][compiler] Necessary cast is marked as "unnecessary" for cast check in 3.2RC4
Summary: [1.5][compiler] Necessary cast is marked as "unnecessary" for cast check in 3...
Status: VERIFIED FIXED
Alias: None
Product: JDT
Classification: Eclipse Project
Component: Core (show other bugs)
Version: 3.2   Edit
Hardware: PC Windows XP
: P3 normal (vote)
Target Milestone: 3.2 RC5   Edit
Assignee: Philipe Mulet CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-05-15 12:40 EDT by John Liu CLA
Modified: 2006-05-19 11:16 EDT (History)
6 users (show)

See Also:

Attachments
Proposed patch (3.48 KB, patch)
2006-05-17 12:08 EDT, Philipe Mulet CLA 		no flags Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description John Liu CLA 2006-05-15 12:40:05 EDT 
For the below application the ((Object[])test) is line is marked as unnecessay.  This seemed to happen only in 3.2RC4 and not in previous builds.

import java.util.Arrays;
class Test {
   public static void main( String args[] ) {
      Object test = new Object[] { "Hello", "World" };
      System.out.println(Arrays.asList(test));
      System.out.println(Arrays.asList((Object[])test)); // Warning here
   }
}

If the application is run.. the results are quite different

[[Ljava.lang.Object;@1372a1a]
[Hello, World]
Comment 1 Olivier Thomann CLA 2006-05-15 12:44:48 EDT 
Reproduced with RC4.
Comment 2 Philipe Mulet CLA 2006-05-15 12:55:40 EDT 
Regression. 3.2RC2 didn't report the problem (haven't checked RC3 yet).
Comment 3 Philipe Mulet CLA 2006-05-17 09:28:57 EDT 
Problem got introduced in RC4. Consequence of fix for bug 139931
Comment 4 Philipe Mulet CLA 2006-05-17 12:08:44 EDT 
Created attachment 41743 [details]
Proposed patch

The patch is nuking the cast warning in case there is a dimension mismatch (ie. less warnings are going to be emitted no matter what). Thus it is safe. An extra warning is causing grief, since quickfix is going to clean code which it shouldn't.

The patch has no influence on codegen, which always was correct here. Just an inacurrate warning notification.
Comment 5 Philipe Mulet CLA 2006-05-17 12:09:54 EDT 
Given the regression was quickly spotted, and it may affect quickfix; we may want to consider it for 3.2RC5.

Martin - what is your vote ?
Comment 6 Philipe Mulet CLA 2006-05-17 12:10:35 EDT 
Olivier - pls review the patch
Comment 7 Philipe Mulet CLA 2006-05-17 12:11:00 EDT 
Dani - what is your vote ?
Comment 8 Philipe Mulet CLA 2006-05-17 12:11:49 EDT 
Mike - what is your vote ?
Comment 9 Mike Wilson CLA 2006-05-17 12:47:14 EDT 
+1
Comment 10 Philipe Mulet CLA 2006-05-17 12:50:40 EDT 
+1 for 3.2RC5. As said earlier, only possible downside of this fix is that we could nuke too many warnings (only for varargs method) if we got the fix wrong.
Comment 11 Martin Aeschlimann CLA 2006-05-17 12:57:25 EDT 
+1
Comment 12 Philipe Mulet CLA 2006-05-18 06:41:46 EDT 
Actually, since dani went away.
Darin - would you please vote instead ?
Comment 13 Philipe Mulet CLA 2006-05-18 06:43:09 EDT 
Tim - what do you think ?
Comment 14 Darin Wright CLA 2006-05-18 10:19:51 EDT 
+1
Comment 15 Philipe Mulet CLA 2006-05-18 10:25:59 EDT 
Released for 3.2RC5
Comment 16 Olivier Thomann CLA 2006-05-19 11:16:37 EDT 
Verified with I20060519-0010 for 3.2RC5