RFR: 8073357: schema1.xsd has wrong content. Sequence of the enum values has been changed

[Aleksej Efimov]

On 04/21/2015 07:36 PM, huizhe wang wrote:
>
> On 4/21/2015 5:14 AM, Aleksej Efimov wrote:
>> Hi Joe,
>> Thank you for you comments - I have modified the test [1] to avoid 
>> multiple file open operations - now the test reads the file content 
>> one time after each call  to schemagen tool.
>
> Ok.
>
>> Golden/expected output is a good suggestion, but we'll depend here on 
>> a format of generated schema file - it can be changed in future. So, 
>> I better check only for a test specific parts of the generated schema 
>> file.
>
> The report was "schema1.xsd has wrong content", which was why I was 
> looking for schema1.xsd. The lib conforms to the schema specification, 
> shall therefore be quite stable. Our XML tests for example, used 
> golden files created since 2005. They have not been a problem 
> themselves, rather they are helpful in catching issues / regressions. 
> I'm fine with your test. But it's something you may consider or talk 
> to Miran.
True, the title is confusing/not a good one - "schema1.xsd" is a default 
file name generated by schemagen and plus it was a follow-up to 
JDK-8046817 bug - we fixed the generation of Enum's, but the order was 
different than in JDK7. Thanks for explanation about the golden file - 
we definitely will need to discuss it with you/Miran and I think we 
should end-up with unified approach for such tests. I will start 
different thread to discuss it.

Best Regards,
Aleksej
>
> Best,
> Joe
>
>>
>> With Best Regards,
>> Aleksej
>>
>> [1] http://cr.openjdk.java.net/~aefimov/8073357/9/01/
>>
>> On 04/21/2015 02:28 AM, huizhe wang wrote:
>>> Hi Aleksej,
>>>
>>> You may also consider using a golden/expected output file to be 
>>> compared with the output. It makes a test more readable (easily see 
>>> what's expected), easier to maintain with a separated golden file 
>>> than multiple literal strings in the test, and technically simpler 
>>> logic (e.g. runSchemaGen and then compareResult) and without opening 
>>> the output file multiple times.
>>>
>>> Regards,
>>> Joe
>>>
>>> On 4/20/2015 1:22 PM, Lance Andersen wrote:
>>>> Hi Aleksej,
>>>>
>>>> The updates to the test seem reasonable.
>>>>
>>>> Best
>>>> Lance
>>>> On Apr 20, 2015, at 2:00 PM, Aleksej Efimov 
>>>> <aleksej.efimov at oracle.com> wrote:
>>>>
>>>>> Hello,
>>>>>
>>>>> The JDK9 schemagen tool hadn't preserved order of the enum values 
>>>>> [1] and it was fixed in standalone project and was synced to JDK 
>>>>> as part of JAXWS integration [2].
>>>>> Can I have a review for the 
>>>>> 'test/javax/xml/ws/8046817/GenerateEnumSchema.java' test update 
>>>>> [3] to include test case for this functionality. The test and the 
>>>>> partial backport of [2] will be backported to JDK8.
>>>>> The modified test shows no failures with JPRT on all platforms.
>>>>>
>>>>> With Best Regards,
>>>>> Aleksej
>>>>>
>>>>> [1] JBS bug: https://bugs.openjdk.java.net/browse/JDK-8073357
>>>>> [2] JAXWS integration: 
>>>>> https://bugs.openjdk.java.net/browse/JDK-8076549
>>>>> [3] Webrev: http://cr.openjdk.java.net/~aefimov/8073357/9/00/
>>>>
>>>>
>>>> Lance Andersen| Principal Member of Technical Staff | +1.781.442.2037
>>>> Oracle Java Engineering
>>>> 1 Network Drive
>>>> Burlington, MA 01803
>>>> Lance.Andersen at oracle.com
>>>>
>>>>
>>>>
>>>
>>
>