Hi Thomas,

On 11/30/2018 02:06 PM, Thomas Stüfe wrote:
Hi Roger,

I updated the CSR according to your feedback. I'm a bit at a loss
about the specification though. How should I specify the behavior of
the API without describing the implementation?
What you wrote is fine. It does need to mention posix_spawn by name,
as that is the OS function being used.

The notes about the behavior of libc, fit better as an explanation
in the description of the Solution than in the Specification section.


Also, since this patch only extends an existing implementation to
Linux, I would have thought there are there technical notes in place
describing POSIX_SPAWN on other platforms, which I would have just
refered to. I searched but could not really find anything.
Nope, that's why it was a debugging tool for Martin,
not a documented implementation feature.

I'm not sure what the proper plural of Unix's is but Unices looks odd.
Perhaps avoid the issue and just say Unix platforms.

Thanks, Roger


Thanks, Thomas


On Fri, Nov 30, 2018 at 3:50 PM Roger Riggs <roger.ri...@oracle.com> wrote:
Hi Thomas,

Looks pretty good.

Usually 'we' avoid the first person writing in the jira.
It makes them more readable in the long term, when there is no context for 'we'.

- Describing it as 'experimental' gives the wrong impression
and has some magnified implications as that term is being used for
other major changes.

- The compatibility risk should be corrected:

Supplying an unknown value on the command line produces a java.lang.Error.

% java -Djdk.lang.Process.launchMechanism=POSIX_SPAWN ...

java.lang.Error: POSIX_SPAWN is not a supported process launch mechanism on 
this platform.

- Since CSRs should be self contained, the specification section should 
explicitly
specify the behavior from the API point of view. CSRs should avoid describing
the implementation (though in this case, its not entirely possible).
The webrev of the impl is not relevant.

Thanks, Roger


On 11/30/2018 03:32 AM, Thomas Stüfe wrote:

CSR for jdk12: https://bugs.openjdk.java.net/browse/JDK-8214511

..Thomas



Reply via email to