8226242 : Diagnostic output for posix_spawn failure

[Roger Riggs]

Hi David,

The error that is being reported is the absence of an ack from 
jsspawnhelper (CHILD_IS_ALIVE)
indicating that it has started (and is about to exec the target exe).
The child process should continue to run; reporting the status of the child
will hopefully lead to identifying why the child process terminated 
without exec'ing the target.

Thanks, Roger


On 6/17/19 6:16 PM, David Holmes wrote:
> Hi Roger,
>
> On 18/06/2019 6:00 am, Roger Riggs wrote:
>> Hi,
>>
>> Updated:
>> http://cr.openjdk.java.net/~rriggs/webrev-spawn-diag-8225192-3/
>
> +     if (WIFEXITED(status)) {
> +         snprintf(ebuf, sizeof ebuf,
> +             "Failed to exec spawn helper: pid: %d, exit value: %d",
> +             pid, WEXITSTATUS(status));
>
> WIFEXITED returns non-zero when the process terminates normally, so 
> that need not be an error AFAICS.
>
> David
>
>>
>> On 6/17/19 2:20 PM, Martin Buchholz wrote:
>>>
>>>
>>> On Mon, Jun 17, 2019 at 10:47 AM Thomas Stüfe 
>>> <thomas.stuefe at gmail.com <mailto:thomas.stuefe at gmail.com>> wrote:
>>>
>>>     Hi Roger,
>>>
>>>     I think this is fine.
>>>
>>>     Could you please also print out WIFEXITED and WIFSIGNALLED? Since
>>>     I am not
>>>     sure if the WEXITSTATUS contains valid info if WIFEXITED is 0, 
>>> or just
>>>     random noise. Same for WTERMSIG.
>>>
>> I separated the causes, both exits are abnormal since the exec 
>> didn't' do what was expected.
>> Only an exit status of 0 is 'normal' and in this case, that is not 
>> what expected.
>>
>> Regards, Roger
>>>
>>>
>>>     Alternatively:
>>>     if (WIFEXITED) print("exited normaly with %d", WEXITSTATUS)
>>>     else if (WIFSTOPPED) print("terminated with signal %d", WTERMSIG)
>>>
>>>
>>> thanks for doing this.
>>>
>>> I also think it's best to have different detail messages for normal 
>>> termination and death by signal, as some shells do.
>>>
>>> I might put the exception throwing code into a separate function.
>>