Mailing-List: contact harmony-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: harmony-dev@incubator.apache.org
Received-SPF: neutral (asf.osuosl.org: 202.81.18.152 is neither permitted nor
 denied by domain of paulex.yang@gmail.com)
Message-ID: <44446322.1020507@gmail.com>
Date: Tue, 18 Apr 2006 11:55:14 +0800
From: Paulex Yang <paulex.yang@gmail.com>
User-Agent: Thunderbird 1.5 (Windows/20051201)
MIME-Version: 1.0
To: harmony-dev@incubator.apache.org
Subject: Re: should strings in exceptions match the reference implementation?
References: <011b01c660d5$2e802b50$1101a8c0@OFFICEDESKTOP>
 <44442618.5040403@pobox.com> <444449B3.3090509@gmail.com>
 <44444CF7.8010404@pobox.com>
In-Reply-To: <44444CF7.8010404@pobox.com>
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Geir Magnusson Jr wrote:
>
>
> Paulex Yang wrote:
>> Geir Magnusson Jr wrote:
>>>
>>>
>>> Nathan Beyer wrote:
>>>> This seems overly excessive and I'm fairly sure that other JREs 
>>>> don't match
>>>> every message of every exception in the RI.
>>>
>>> Really?  Every other JRE uses the classlibrary from sun.  They would 
>>> have the same messages, wouldn't they?
>> I'm sure not *every*, at least GNU classpath has no chance to use 
>> classlib from Sun ;-) .
>
> It's not a JRE :)
>
> But BEA, IBM and Sun use the library from Sun. I'd bet Apple does too. 
> As far as I know, that's about it for maintream JRE/JDKs...
OK, AFAIK as a user, IBM JDK includes much class lib implementation from 
Sun, but not all, because you can see com.ibm.*** randomly from the 
stack trace:).

Anyway, it's not the key point, the key point is the what is our 
baseline considering "compatibility", I assume the baseline is to pass 
TCK and not to break existing applications. While I have no any idea 
about TCK, I believe few application depends on message, so I see no 
reason so far why we must pay great effort (even greater than exception 
type compatibility I think) to have same message with RI.

And legal issue stand out here yet another time....
>
> Good question though - what does GNU Classpath do?
>
> geir
>
>>>
>>> geir
>>>
>>>>
>>>> I'm not opposed to matching messages to help consistency and 
>>>> debugging, but
>>>> I just don't want it to be a dictate for development and testing. 
>>>> The only
>>>> exception, no pun intended, to this would be if the specification 
>>>> defines a
>>>> format for the message and thus making it an explicit part of the API.
>>>>
>>>>> -----Original Message-----
>>>>> From: Mark Hindess [mailto:mark.hindess@googlemail.com]
>>>>> Sent: Saturday, April 15, 2006 2:44 PM
>>>>> To: Harmony Dev
>>>>> Subject: should strings in exceptions match the reference 
>>>>> implementation?
>>>>>
>>>>> Another thing that came up when looking at PatternSyntaxExceptionTest
>>>>> (HARMONY-352) was that the test was testing for the strings in
>>>>> exceptions.  Since these were testing for strings not in the 
>>>>> exceptions
>>>>> thrown by the new implementation (nor by the reference
>>>>> implementation), I thought about removing the tests and just ignoring
>>>>> the strings.  But then...
>>>>>
>>>>> I remembered that two days ago I had solved a problem with tests
>>>>> failing on our windows build machine [0], by doing a google search 
>>>>> for
>>>>> the exact string in the exception that was being thrown.  It would
>>>>> have been much harder to solve if the string didn't match the string
>>>>> thrown by the reference implementation.  So, I think we should try to
>>>>> match strings in exceptions because it will help our users when 
>>>>> trying
>>>>> to debug problems.  Currently we don't in very many cases.
>>>>>
>>>>> What do other people think?
>>>>>
>>>>> Obviously we will have to match them if we are going to test for them
>>>>> in our API tests or they wont pass when run against a reference
>>>>> implementation.
>>>>>
>>>>> My patch for the PatternSyntaxExceptionTest has the string tests with
>>>>> the expected values set to the values from the exceptions trhown by
>>>>> the RI.  But I've commented them out since harmony strings don't
>>>>> currently match and since the strings probably should have different
>>>>> line endings on different platforms.
>>>>>
>>>>> Regards,
>>>>>  Mark.
>>>>>
>>>>> [0] SystemRoot not being set in the environment when called from my
>>>>>     ant task.
>>>>>
>>>>> -- 
>>>>> Mark Hindess <mark.hindess@googlemail.com>
>>>>> IBM Java Technology Centre, UK.
>>>>>
>>>>> ---------------------------------------------------------------------
>>>>> Terms of use : http://incubator.apache.org/harmony/mailing.html
>>>>> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
>>>>> For additional commands, e-mail: 
>>>>> harmony-dev-help@incubator.apache.org
>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> Terms of use : http://incubator.apache.org/harmony/mailing.html
>>>> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
>>>> For additional commands, e-mail: harmony-dev-help@incubator.apache.org
>>>>
>>>>
>>>
>>> ---------------------------------------------------------------------
>>> Terms of use : http://incubator.apache.org/harmony/mailing.html
>>> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
>>> For additional commands, e-mail: harmony-dev-help@incubator.apache.org
>>>
>>>
>>
>>
>
> ---------------------------------------------------------------------
> Terms of use : http://incubator.apache.org/harmony/mailing.html
> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
> For additional commands, e-mail: harmony-dev-help@incubator.apache.org
>
>


-- 
Paulex Yang
China Software Development Lab
IBM


---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
For additional commands, e-mail: harmony-dev-help@incubator.apache.org