[
https://issues.apache.org/jira/browse/BCEL-207?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14700060#comment-14700060
]
Mark Roberts edited comment on BCEL-207 at 8/17/15 10:04 PM:
-------------------------------------------------------------
The dispose() change causes a problem with my existing code. But I believe the
real issue is the whole shallow vs deep copy thing.
MethodGen.getLocalVariables() returns a shallow copy. My client code wants to
rebuild the LocalVarableTable. So it calls getLocalVariables to get a copy of
the old table and then calls removeLocalVariables to prepare for creating the
new table. But, of course, the new table is a variation of the old so you need
to copy some of the items from the old to the new table. There is/was a subtle
assumption that getLocalVariables returned a deep copy in order to do this
properly. It turned out even though the copy was shollow, it worked okay,
because removeLocalVariables() didn't really remove anything, just reset the
Targeters. Now dispose really removes things so we have trouble. I'm not
suggesting we remove the dispose stuff. I would like to discuss just what we
mean by a method called getFoo(). Does it return a shallow or deep copy of the
Foo object? I think I am arguing that it should be deep. This issue gets even
messier because some BCEL objects override clone() and some don't. And some
have a copy() method (which I assume is always deep?) and some don't. (oh boy,
now I'm wondering if I should have sent this to the commons list instead.)
Well. as usual, it's more complicated than that. Turns out sometimes we want
getFoo() to be shallow so we can modify the Foo object and sometimes we want
getFoo() to be deep so we can create a new Foo from the copy. So now I'm
thinking getFoo() should be shallow, and the user needs to say getFoo().copy()
if he wants deep? What do you think? (Of course, very few of the BCEL objects
have a copy() method....)
was (Author: markro):
The dispose() change causes a problem with my existing code. But I believe the
real issue is the whole shallow vs deep copy thing.
MethodGen.getLocalVariables() returns a shallow copy. My client code wants to
rebuild the LocalVarableTable. So it calls getLocalVariables to get a copy of
the old table and then calls removeLocalVariables to prepare for creating the
new table. But, of course, the new table is a variation of the old so you need
to copy some of the items from the old to the new table. There is/was a subtle
assumption that getLocalVariables returned a deep copy in order to do this
properly. It turned out even though the copy was shollow, it worked okay,
because removeLocalVariables() didn't really remove anything, just reset the
Targeters. Now dispose really removes things so we have trouble. I'm not
suggesting we remove the dispose stuff. I would like to discuss just what we
mean by a method called getFoo(). Does it return a shallow or deep copy of the
Foo object? I think I am arguing that it should be deep. This issue gets even
messier because some BCEL objects override clone() and some don't. And some
have a copy() method (which I assume is always deep?) and some don't. (oh boy,
now I'm wondering if I should have sent this to the commons list instead.)
> RemoveLocalVariable(s) doesn't remove the associated Targetters.
> ----------------------------------------------------------------
>
> Key: BCEL-207
> URL: https://issues.apache.org/jira/browse/BCEL-207
> Project: Commons BCEL
> Issue Type: Bug
> Reporter: Mark Roberts
> Fix For: 6.0
>
> Attachments: methodgen3.diff
>
>
> RemoveLocalVariable(s) doesn't remove the associated Targetters.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
