[rfc][icedtea-web] Don't be silent if JNLPClassLoader gets a ClassFormatError
Jiri Vanek
jvanek at redhat.com
Wed Mar 20 02:51:58 PDT 2013
On 03/19/2013 08:42 PM, Adam Domurad wrote:
> On 03/19/2013 03:19 PM, Jiri Vanek wrote:
>> On 03/19/2013 05:00 PM, Adam Domurad wrote:
>>> I think this is worthy because it would have saved me a lot of time investigating a recent bug.
>>> This will continue as normal, but ClassFormatError's really should be logged. It will make
>>> resulting ClassNotFoundException's less mysterious.
>>>
>>> 2013-19-03 Adam Domurad <adomurad at redhat.com>
>>>
>>> * netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java
>>> (findClass): Print stacktrace for ClassFormatError
>>>
>>>
>> Hmm... why keep other ones silent?
>
> You mean PrivilegedActionException? It is worthy perhaps, shall I include it when I push ?
> (NullJnlpFileException already stores the exception and rethrows, and CNFE represents a failure that
> we fully expect when we fail to load a classs.)
>
>> Why do not rethrow?
>
> Well my logic was simply to be as lenient as possible, eg in case we ever pick something up that
> isn't really a class. However rethrow sounds good to me too. If you prefer this I can do it (I do
> not expect it to break anything, especially anything sanely organized). I looked at the history of
> this catch and it is present in the initial commit, so no telling why it was added.
>
>>
Yup :)) I like your points... But I'm not so sure that it will break anything....
I'm for rethrowing where possible, and be laud/log where we can not rethrow.
But in this particular part of code it will need a lot of empiric tastings, or somebody with insight
into classlaoders... Omair? (silently CCed O:)
In case that we will no able to rethrow, isnt worthy to get rid of printstacktrace and to add
standard logger? Just thing to think about....
J.
More information about the distro-pkg-dev
mailing list