Add a suggestion for non-English locale of Linux in the test doc

Jing Tian jingtian at loongson.cn
Tue Apr 16 08:10:43 UTC 2019


Thank you for your suggestions. I have reworked the documentation and I 
think this is a prudent approach so far. Regarding the other seven test 
cases, I will continue to find out why they can't pass the test.

JBS: https://bugs.openjdk.java.net/browse/JDK-8222444
Webrev: http://cr.openjdk.java.net/~lzhai/8222444/webrev.01/

Jing Tian

在 2019/4/15 23:24, naoto.sato at oracle.com 写道:
> As for the wording, I'd suggest "Non-US Locale" instead of 
> "Non-English." Some tests may depend on US customary behavior, such as 
> date format, decimal separator, etc.
>
> Naoto
>
> On 4/15/19 6:59 AM, Erik Joelsson wrote:
>> Hello,
>>
>> Documenting this is certainly the least we can do. If our tests 
>> depend on the locale being set to en_US, then I think the best action 
>> would be to provide such a configuration directly in RunTests.gmk. 
>> Exporting LANG should work for all Unix OSes, but most likely not on 
>> Windows. The extra VM_OPTIONS would fix most of them it seems. Would 
>> it be worth investigating the remaining 7 and get them fixed?
>>
>> In the meantime, documenting seems prudent. I would suggest something 
>> like this:
>>
>> ### Non-English Locale
>>
>> If your locale is non-English, some tests are likely to fail. To work 
>> around this you can set the locale to English. On Unix platforms 
>> simply setting `LANG=en_US` in the environment before running tests 
>> should work. On Windows, setting 
>> `JTREG="VM_OPTIONS=-Duser.language=en -Duser.country=US"` helps for 
>> most, but not all test cases.
>>
>> /Erik
>>
>> On 2019-04-14 20:28, Jing Tian wrote:
>>> Hi,
>>>
>>> We have discussed the issue of the test cases fail because of locale 
>>> before[1].
>>>
>>> Thanks for the suggestions given by Naoto and David. I think we can 
>>> put this advice in the test doc, which may be better for people to 
>>> test. This advice can avoid the problem that caused by locale and we 
>>> can pay more attention to the functional points that the test itself 
>>> focuses on.
>>>
>>> Set JTREG="VM_OPTIONS=-Duser.language=en -Duser.country=US" , it 
>>> does pass most test cases, but there are still very few test cases(7 
>>> in total) can't pass the test when they are in a non-English locale.
>>>
>>> I think if 'make test' in a non-English locale, we can set the 
>>> locale to English first. Use 'export LANG="en_US"'. But this method 
>>> is just for Linux. I test "tier1 tier2 tier3" after setting 
>>> LANG="en_US". The problems caused by the local settings have not 
>>> appeared anymore.
>>>
>>>
>>> JBS: https://bugs.openjdk.java.net/browse/JDK-8222444
>>> Webrev: http://cr.openjdk.java.net/~lzhai/8222444/webrev.00/
>>>
>>> The testing.html is updated automatically using "make 
>>> update-build-docs" with pandoc version 2.7.2.
>>>
>>> [1] 
>>> https://mail.openjdk.java.net/pipermail/compiler-dev/2019-March/013144.html 
>>> <https://mail.openjdk.java.net/pipermail/compiler-dev/2019-March/013144.html> 
>>>
>>>
>>> Cheers,
>>> Jing Tian
>>>




More information about the build-dev mailing list