Filter specification bug (was: Re: [VOTE] Felix framework 2.0.0 and related subprojects releases)

[Angelo van der Sijpt <an...@luminis.nl>]

On Sep 9, 2009, at 8:02 PM, Stuart McCulloch wrote:

> 2009/9/10 Angelo van der Sijpt <an...@luminis.nl>
>
>>
>> On Sep 9, 2009, at 7:00 PM, Richard S. Hall wrote:
>>
>> On 9/9/09 12:54, Angelo van der Sijpt wrote:
>>>
>>>> +1 (very non-authoritive)
>>>>
>>>> We checked the new release against our OSGi testing framework (
>>>> http://opensource.luminis.net/wiki/display/OSGITEST/OSGi+testing+framework
>>>> ), and found nothing bad.
>>>>
>>>> It's a shame about the new OSGi-provided filter implementation,  
>>>> by the
>>>> way...
>>>>
>>>>
>>> Why is that?
>>>
>>
>> In some corner cases, the OSGi implementation might conclude that a  
>> filter
>> is invalid, while it is actually permitted by the production rules.  
>> For
>> instance, (!ab=a) states an attribute name of '!ab', which is not  
>> disallowed
>> by spec, and solved correctly by Felix' filter implementation. The  
>> provided
>> one assumes that any ! after a ( will signal a negation, and  
>> complain that
>> it requires a ( .
>
>
> have you raised a bug for this over at
> https://www.osgi.org/bugzilla/enter_bug.cgi ?

Bug filed as https://www.osgi.org/bugzilla/show_bug.cgi?id=57 .

Angelo

Re: Filter specification bug

["Richard S. Hall" <he...@ungoverned.org>]

On 9/10/09 1:18, Angelo van der Sijpt wrote:
>
> On Sep 9, 2009, at 8:02 PM, Stuart McCulloch wrote:
>
>> 2009/9/10 Angelo van der Sijpt <an...@luminis.nl>
>>
>>>
>>> On Sep 9, 2009, at 7:00 PM, Richard S. Hall wrote:
>>>
>>> On 9/9/09 12:54, Angelo van der Sijpt wrote:
>>>>
>>>>> +1 (very non-authoritive)
>>>>>
>>>>> We checked the new release against our OSGi testing framework (
>>>>> http://opensource.luminis.net/wiki/display/OSGITEST/OSGi+testing+framework 
>>>>>
>>>>> ), and found nothing bad.
>>>>>
>>>>> It's a shame about the new OSGi-provided filter implementation, by 
>>>>> the
>>>>> way...
>>>>>
>>>>>
>>>> Why is that?
>>>>
>>>
>>> In some corner cases, the OSGi implementation might conclude that a 
>>> filter
>>> is invalid, while it is actually permitted by the production rules. For
>>> instance, (!ab=a) states an attribute name of '!ab', which is not 
>>> disallowed
>>> by spec, and solved correctly by Felix' filter implementation. The 
>>> provided
>>> one assumes that any ! after a ( will signal a negation, and 
>>> complain that
>>> it requires a ( .
>>
>>
>> have you raised a bug for this over at
>> https://www.osgi.org/bugzilla/enter_bug.cgi ?
>
> Bug filed as https://www.osgi.org/bugzilla/show_bug.cgi?id=57 .

Great. This bug will probably not be addressed for a while, but at least 
it is on our radar...if you want to submit a patch to the filter impl, 
we could potentially include it in a future build of Felix.

-> richard
>
> Angelo