8205054: Could not find "lsof" on test machine

[David Holmes]

Hi Leo,

Inine ...

On 15/06/2018 6:06 PM, Leo Korinth wrote:
> Hi David.
> 
> On 15/06/18 04:08, David Holmes wrote:
>> Sorry after doing more testing and looking at the output I think we 
>> need to do better. For a machine with no lsof we see:
>>
>> using command: <not found>

That looks like an error ...

>> (Second VM) Open file descriptors:
>>
>> using command: <not found>

That looks like an error too - but why did we keep going after the first 
error?
>> (Third VM) Open file descriptors:
>>
>> VM RESULT => RETAINS FD
>> VM RESULT => VM EXIT
>>
>> Log file was not inherited by third VM

But that's not true - we couldn't determine anything! So this looks like 
an accidental pass rather than an deliverate skip.

>> ---
>>
>> That looks like an error is occurring to me and that we pass by 
>> accident. I think we need to terminate the test, cleanly of course, as 
>> soon as we can't find the tool and report that the test can't run 
>> because of that.
>>
>> Thanks,
>> David
> 
> I could change the output to "using command: <not found, terminating 
> test as successful as we can not finish it in a meaningful way>" or 
> maybe something else that you prefer that is better.
> 
> I do believe this is the easy way to terminate the test cleanly without 
> leaving running JVMs. We need to communicate back to the first JVM that 
> the test was successful. Is it the string "<not found>" that looks like 
> an error and that we pass by accident? Or is it something in the test 
> case code that looks fishy?

See above.

I'd suggest locating the command as one of the first things in the test 
in the first VM and then bail out if not found.

Thanks,
David
------

> Or maybe the comment should be improved?
>    // if command can not be found return list without log file (some 
> machines does not have "lsof" in the expected place). (add this->)The 
> empty list will make the test exit successfully.(<-add this)
> 
>> On 15/06/2018 11:21 AM, David Holmes wrote:
>>> Hi Leo,
>>>
>>> This is what I was concerned about - and yes the CI testing did shake 
>>> it out in a few cycles.
>>>
>>> I agree with the fix and will sponsor it for you.
> 
> Thanks David.
> /Leo
> 
>>>
>>> Thanks,
>>> David
>>>
>>> On 15/06/2018 3:31 AM, Leo Korinth wrote:
>>>>
>>>>
>>>> On 14/06/18 19:10, Lindenmaier, Goetz wrote:
>>>>> Hi Leo,
>>>>>
>>>>> the fix looks good.  It makes sense to skip a test if the 
>>>>> infrastructure is
>>>>> insufficient to run it.
>>>>>
>>>>> Could you please put the long comment at the end of line 179
>>>>> into a line of it's own?
>>>>
>>>> Fixed
>>>>
>>>>> Also, please take the chance and move the Copyright message
>>>>> to the beginning of the file.
>>>>
>>>> Fixed
>>>>
>>>>> Best regards,
>>>>>    Goetz.
>>>>>
>>>>
>>>> Thanks for the fast review; I need a second reviewer and someone to 
>>>> help me push (I am not a commiter).
>>>>
>>>> New webrevs:
>>>> http://cr.openjdk.java.net/~lkorinth/8205054/00_01/ (incremental)
>>>> http://cr.openjdk.java.net/~lkorinth/8205054/01/    (full)
>>>>
>>>> Thanks,
>>>> Leo
>>>>
>>>>>> -----Original Message-----
>>>>>> From: hotspot-runtime-dev [mailto:hotspot-runtime-dev-
>>>>>> bounces at openjdk.java.net] On Behalf Of Leo Korinth
>>>>>> Sent: Donnerstag, 14. Juni 2018 18:55
>>>>>> To: Hotspot dev runtime <hotspot-runtime-dev at openjdk.java.net>
>>>>>> Subject: RFR: 8205054: Could not find "lsof" on test machine
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Unfortunately some test machines does not have "lsof" installed, 
>>>>>> or it
>>>>>> can not be found.
>>>>>>
>>>>>> I do not know how to find "lsof" on the test machine where the test
>>>>>> fails; the command might not be installed. Here is a fix to make the
>>>>>> test case succeed if "lsof" can not be found. I believe it is the 
>>>>>> best
>>>>>> option (at least for now). I am sorry for all the noise I am 
>>>>>> generating.
>>>>>>
>>>>>> Bug:
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8205054
>>>>>>
>>>>>> Webrev:
>>>>>> http://cr.openjdk.java.net/~lkorinth/8205054/00/
>>>>>>
>>>>>> Testing:
>>>>>> I have tested this on my machine, and I have a mach5 test job 
>>>>>> running.
>>>>>>
>>>>>> Thanks,
>>>>>> Leo