RFR 9: 8034903: Add Logging of Process.start arguments and resulting pid

Martin Buchholz martinrb at google.com
Sat Feb 15 18:27:36 UTC 2014


On Fri, Feb 14, 2014 at 12:14 PM, Alan Bateman <Alan.Bateman at oracle.com>wrote:

> On 14/02/2014 18:10, Martin Buchholz wrote:
>
>> I'm not excited about having logging statements added throughout the jdk
>> implementation.  PlatformLogger is an implementation detail, so real users
>> cannot benefit.  Your OS should already provide facilities to let you
>> trace
>> process creation.  E.g. on Linux you can do this with strace (and I have
>> done this myself in the past to debug subprocess creation).  So I'd say
>> adding logging just to provide a small rarely used convenience for jdk
>> developers is Not Worthwhile.
>>
>>  Just on PlatformLogger then it just forwards to j.u.logging when it is
> present.
>
> OK, so is this code for the benefit of "real" users or openjdk developers?
 If the latter, it doesn't pull its weight - "Dude, you forgot to remove
your printf debugging statements".  If the former, this feature needs to be
explained somewhere - your JDK will randomly start doing more logging that
users didn't ask for and is undocumented.

It's very nuanced, because I am fully in favor of the similar
feature "_JAVA_LAUNCHER_DEBUG".  Maybe because strace/truss solved my need
for subprocess invocation debugging on Linux/Solaris.



More information about the core-libs-dev mailing list