RFR (S) 8167548: [TESTBUG] Logging tests put log files in source tree

[Harold Seigel]

Hi David,

Thanks for looking at this.

The test's logging file name includes the job's start time.  But, the 
test's call to LogFileOutput::set_file_name_parameters(0) sets the job's 
start time to the beginning of time: 1970-01-01_01-00-00.  The logging 
code then converts this from UTC to local time and uses the local time 
in the file name.  And, since Burlington MA is five hours behind UTC, 
the resulting time used in the file name is: 1969-12-31_19-00-00.  So, 
my 'fix' would only work when the test was run in a timezone that's five 
hours behind UTC.

It looks like many of the gtest logging tests create files in the source 
directory.  So, I'm withdrawing this change until this bigger problem 
can be addressed.

Thanks, Harold

On 2/7/2019 7:26 PM, David Holmes wrote:
> Hi Harold,
>
> On 8/02/2019 10:08 am, Harold Seigel wrote:
>> Hi,
>>
>> Please review this fix for JDK-8167548.  The test used the wrong 
>> 'predictable' file names so the files were not getting deleted.
>
> I'm confused. Where does the magic "1969-12-31_19-00-00" date string 
> come from?
>
> Second, this still means the test puts logs into the source tree, 
> which is a bad thing. can we not fix this so that the log file is in a 
> tmp directory, or a jtreg working directory or ...?
>
> Thanks,
> David
>
>> Open Webrev: 
>> http://cr.openjdk.java.net/~hseigel/bug_8167548/webrev/index.html
>>
>> JBS Bug:  https://bugs.openjdk.java.net/browse/JDK-8167548
>>
>> The fix was tested by hand on Linux x64 to make sure the files were 
>> deleted.  It was also regression tested on Windows, Solaris, and Mac.
>>
>> Thanks, Harold
>>