search

joepadde / mockito

Automatically exported from code.google.com/p/mockito
0 stars 0 forks source link

Investigate the use of another bytecode manipulation library #217

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago 
CGLIB doesn't seem to have been updated in a long time, and I think Mockito 
should anticipate a full deprecation of this library if this happens.

The technical reason are :
* Today there is some known bugs that affects Mockito, such as issue 212.
* With the upcoming JDK7 (well one day at least), if CGLIB isn't updated, 
Mockito won't be able to create mocks from JDK7 bytecode.

Hibernate staff won't use CGLIB anymore, and will progressively replace CGLIB 
use by the javassist library. (see 
http://in.relation.to/Bloggers/DeprecatedCGLIBSupport)

javassist maybe a good start sinci it is actively maintained by the JBoss 
Community, however certain care might be required regarding the liscence, as i 
don't know the real constraints about MPL or LGPL if there is any regarding the 
Mockito liscence.

Original issue reported on code.google.com by brice.du...@gmail.com on 3 Sep 2010 at 4:02

GoogleCodeExporter commented 8 years ago

Original comment by brice.du...@gmail.com on 31 Oct 2010 at 10:19

GoogleCodeExporter commented 8 years ago 
This issue was closed by revision 966a8bc0e0.

Original comment by szcze...@gmail.com on 31 Oct 2010 at 8:25

GoogleCodeExporter commented 8 years ago

Original comment by brice.du...@gmail.com on 1 Nov 2010 at 10:48

GoogleCodeExporter commented 8 years ago

Original comment by brice.du...@gmail.com on 3 Nov 2010 at 1:48

GoogleCodeExporter commented 8 years ago 
More on Javassist there :
http://www.csg.is.titech.ac.jp/~chiba/javassist/tutorial/tutorial.html

Original comment by brice.du...@gmail.com on 16 Dec 2010 at 4:30

GoogleCodeExporter commented 8 years ago 
So.... this is now a serious bug that means no code under JDK7 will run -when- 
compiled a JDK7 compile target.

Has any movement been done on switching to Javassist at all?

Original comment by m...@talios.com on 27 Nov 2011 at 11:38

GoogleCodeExporter commented 8 years ago 
A movement has begun, but you don't switch that easily to another lib just like 
that, especially since this project is mostly done on our free time which is 
sparse.

Original comment by brice.du...@gmail.com on 28 Nov 2011 at 8:56

GoogleCodeExporter commented 8 years ago 
Hmmm, are you saying Mockito does not work under java7?

Original comment by szcze...@gmail.com on 6 Dec 2011 at 12:32

GoogleCodeExporter commented 8 years ago 
ASM 4.0 supports JDK7, and I've tried to update the sources under cglib-and-asm.

I rebuilt mockito, however this didn't solve the problem.
I'm still getting this:
Caused by: java.lang.VerifyError: Bad type on operand stack in aaload in method 
...
    at java.lang.Class.getDeclaredConstructors0(Native Method)
    at java.lang.Class.privateGetDeclaredConstructors(Class.java:2404)
    at java.lang.Class.getDeclaredConstructors(Class.java:1853)
    at org.mockito.internal.creation.jmock.ClassImposterizer.setConstructorsAccessible(ClassImposterizer.java:70)
    at org.mockito.internal.creation.jmock.ClassImposterizer.imposterise(ClassImposterizer.java:65)

It seems the guys at jetbrains have solved a similar issue: 
http://youtrack.jetbrains.net/issue/IDEA-71870

Just in case I'm attaching the updated cgi-lib/asm sources.

Original comment by idls...@gmail.com on 9 Feb 2012 at 10:22

Attachments:

GoogleCodeExporter commented 8 years ago 
Hi,

Yeah we know about ASM 4, unfortunately, we are using CGLIB that is using ASM. 
If CGLIB doesn't use ASM with new class format instruction then it won't work. 
And CGLIB does seem to be a dead project. I definitely would like to spend more 
time here, but right now I don't have this free time luxury.

Actually our best shot would be to try with Javassist. There might be actually 
some progress on that matter.

Original comment by brice.du...@gmail.com on 10 Feb 2012 at 8:45

GoogleCodeExporter commented 8 years ago 
Anyway thank you so much for your time investigating the problem and possible 
solutions :)

Original comment by brice.du...@gmail.com on 10 Feb 2012 at 8:47

GoogleCodeExporter commented 8 years ago 
But you're using the repackaged version of both cglib and asm. So I assumed 
that replacing a repackaged asm 3.x with a repackaged asm 4.0 and subsequent 
refactoring of the repackaged cglib would to the trick.
Unfortunately I was wrong, there is obviously some little detail missing.

Original comment by idls...@gmail.com on 10 Feb 2012 at 2:20

GoogleCodeExporter commented 8 years ago 
At this point I'm guessing. ASM is pretty low level, if some *new* instructions 
are not used to generate a correct bytecode for the JVM 7 then it won't solve 
the issue. As ASM guys actually made a huge effort to keep the API of version 4 
backward compatible. And as said for now that is CGLIB that issue these 
instructions to ASM.

Original comment by brice.du...@gmail.com on 10 Feb 2012 at 2:28

GoogleCodeExporter commented 8 years ago 
@idlsoft, we will gladly accept patch for the cglib-asm we ship mockito with if 
it solves some problem and does not break existing functionality.

First things first, though. What problem we are trying to solve? Are you saying 
that Mockito does not work with jvm7? If that's the case we need a separate 
ticket to track it and let's start a discussion there.

Cheers!

Original comment by szcze...@gmail.com on 11 Feb 2012 at 3:17

GoogleCodeExporter commented 8 years ago 
Hi. Seems that cglib resurrected. 3.0 was released yesterday:

http://sourceforge.net/news/?group_id=56933

Not yet in maven central.

Original comment by eduardo.macarron on 26 May 2012 at 4:12

GoogleCodeExporter commented 8 years ago 
BTW regarding the license Javassist adopted the ASL 2.0 since 3.15.

Original comment by eduardo.macarron on 27 May 2012 at 6:54

GoogleCodeExporter commented 8 years ago 
Thanks for the info. At the moment we're not actively pursuing replacing cglib. 
There does not seem to be a huge benefit of doing this.

Users are welcome to try out javaassist (or other tools) via Mockito's 
MockMaker interface. It is an extension point that enables using custom proxy 
creation mechanism and avoid using current default (cglib).

Original comment by szcze...@gmail.com on 27 May 2012 at 2:16

GoogleCodeExporter commented 8 years ago 
Thx for the update.
I took a look at what they changed, unfortunately it doesn't really help in our 
case right now.

 - They moved to ASM4
 - They changed a few interfaces of their own

Most of the previous issues are still there, however this might be good news as 
the project might see future works.

Original comment by brice.du...@gmail.com on 29 May 2012 at 6:58

GoogleCodeExporter commented 8 years ago 
Hi Brice. Not sure if that is good news. 

We at MyBatis were thinking about going javassist due to cglib inactivity. Now 
there is not that inactivity! :) But I am not sure if this is a signal for 
future improvements. The new version was not released by one of the regular 
cblib devs and it is not yet in maven central and I do not know if it will be.

Original comment by eduardo.macarron on 4 Jun 2012 at 2:48

GoogleCodeExporter commented 8 years ago 
[deleted comment]
GoogleCodeExporter commented 8 years ago 
I hope that javassist support will be available as optionally attachable 
artifact (in maven terms) out-of-the-box. It seems that CGLIB mock maker 
couldn't handler non-public super classes but Javassist version can.

I pulled out the initial source code but it wasn't compatible with latest 
release. It also suffered from the fact that already proxied classes having 
Proxy(Object) interface(es) already in them are not supported. It's quite easy 
to get the first non javassist super class and gather interfaces that are left 
behind on the way. Of course this will not work if the javassist proxies are in 
the middle of the class hierarchy.

Original comment by TuomasKi...@gmail.com on 8 Aug 2012 at 9:12

GoogleCodeExporter commented 8 years ago 
Hi Eduardo,

A few months / weeks ago I took the opportunity to try Javassist as the 
bytecode engine, but I was a bit disappointed : some NPEs in generated bytecode 
in some cases, some bridge method error on call site (some bridge methods were 
not found when the JVM performed some late binding), plus unpractical stuff in 
the API. However I'm talking about the proxy factory.

Still, I heard that the javassist transformers do seem to work well. Johan uses 
them with Powermock.

In my opinion, a new bytecode engine, maybe on top of the good ASM4. By the way 
this guy is crafting a fluent API over ASM4 : https://github.com/qmx/jitescript

Original comment by brice.du...@gmail.com on 8 Aug 2012 at 9:22

GoogleCodeExporter commented 8 years ago 
@tuomas I'v stopped working on javassist, as I had to patch Javassist myself. 
I've proposed a patch, but there's no movement on that front either. And there 
is still NPEs in generated bytecode, that I didn't investigated.
And I would rather avoid important modifications on Javassist anyway, so we 
updating the javassist code would be a no-brainer task.

Also I've seen movement on the CGLIB front, like CGLIB 3 (which appears to be 
in a very young development state). Still the project site is a mess. I don't 
know how the team is commited. (I understand this an opensource project, but a 
better transparency would be very welcome). I don't know what their plan is. 
They still use CVS. etc.

Original comment by brice.du...@gmail.com on 8 Aug 2012 at 9:38

GoogleCodeExporter commented 8 years ago 
@brice Thanks for the info. CgLibMockMaker is a no-go for me due to '2. you are 
trying to stub a final method, you naughty developer!' but luckily I got the 
general idea from your code and I wrote a version that I could use. I hope that 
it doesn't blow up in any of the situations where I'm using it. BTW: 
ClassPathLoader doesn't work correctly under OSGi environment and hence I got 
thrown back to default CgLibMockMaker implementation.

Original comment by TuomasKi...@gmail.com on 16 Aug 2012 at 10:04

GoogleCodeExporter commented 8 years ago 
@tuomas Could you fill a bug, and even best if you have a fix or an 
environement were we can reproduce, that would be awesome !

Original comment by brice.du...@gmail.com on 16 Aug 2012 at 10:08

GoogleCodeExporter commented 8 years ago 
About spring : https://jira.springsource.org/browse/SPR-9669

They finally decided to keep CGLIB, and jarjar it as well. It seems they are 
using CGLIB 3, which we must investigate the changes (Spring had problems with 
this version it seems but they hotfixed it, see the issue link)

Original comment by brice.du...@gmail.com on 4 Dec 2012 at 9:55