TEST_BUG: com/sun/jdi/BreakpointWithFullGC.sh fails to cleanup in Cygwin

[Chris Hegarty]

On 08/ 9/11 05:18 PM, Jim Holmlund wrote:
> Will this work on older versions of cygwin?  Or do we have a spec
> someplace that says cygwin versions older than xxxx aren't supported?

I have not seen any docs/spec anywhere for support of /dev/null in 
specific Cygwin releases, but I did test on 1.5 and 1.7.9 and MKS, and 
all went fine.

http://www.cygwin.com/cygwin-ug-net/using-specialnames.html

I didn't have time to submit a jprt job on this, but tested on Windows 
Cygwin/MKS, Solaris, and Linux. I believe the TL integrator will be 
doing a full jprt run later today.  If we don't see a few hundred 
failures then we should be good ;-)

-Chris.

> - jjh
>
>
> On 8/9/2011 8:56 AM, Chris Hegarty wrote:
>> Sorry Dan, I'll put a more complete description in the bug report.
>>
>> This issue, from what I can tell only happens with very recent Cygwin
>> versions. I'm running 1.7.9. When running the com/sun/jdi tests in a
>> batch, in samevm mode, this test leave behind a file, NUL, in the
>> scratch directory that cannot be deleted. This causes the remaining
>> tests to also fail.
>>
>> Thanks for the review,
>> -Chris.
>>
>> On 08/ 9/11 04:52 PM, Daniel D. Daugherty wrote:
>>> On 8/9/11 8:52 AM, Chris Hegarty wrote:
>>>> Hi Dan, Alan,
>>>>
>>>> This is a trivial change to use /dev/null instead of NUL for Windows
>>>> test with Cygwin.
>>>>
>>>> http://cr.openjdk.java.net/~chegar/7076756/webrev.00/webrev/
>>>>
>>>> Thanks,
>>>> -Chris.
>>>
>>> I'm good with the change, but the bug report itself is less than clear.
>>>
>>> I run these tests all the time on WinXP/Cygwin so I'm guessing
>>> that there is more going on here than meets the eye. I'm going to
>>> guess that the '-samevm' option is the enabler for this failure
>>> mode. The failure mode that you fixed should also affect tests
>>> other than BreakpointWithFullGC.sh so I'm not quite sure why
>>> this issue appears with just that test. Or maybe it does appear
>>> with other tests, but again the bug report isn't clear.
>>>
>>> In any case, thumbs up on the change.
>>>
>>> Dan
>>>