dtrace discuss - Aug 2006 - Re: the ability to specify probes that don''t yet exist that will be enabled when they show up

Adam Leventhal wrote:
>I think it''s probably been backported. If you''re trying to
enable a probe
>that doesn''t exist and you''re getting that warning, add
the -Z option to
>tell DTrace not to complain if your program matches zero probes.
>
Yes, I do exactly this.

Some more info just to be sure everything works as it is supposed to work :)

In jdk6.0 hotspot probes definitions are defined in libjvm.so.
As result when you try to do something like this:
# dscript.d -c "java ...."

hotspot probes could be not loaded yet (because libjvm.so could be not 
loaded yet)
and as result warning like following ones are thrown:
 dtrace: failed to compile script ... : probe description 
hotspot4530:::class-loaded does not match any probes

If I add flag ''-Z'' warnings will disappear.
Also hotspot:::class-loaded probes begins to fire after some time

Can you confirm that doing this way I will not miss any 
hotspot:::class-loaded probes?

Thanks a lot,
-katya
>Adam
>
>On Thu, Aug 10, 2006 at 12:17:04PM +0400, Ekaterina Pavlova wrote:
>
>>Hi Adam,
>>
>>I working on Dtrace samples for mustang (jdk6) and come up with question
>>regarding ''-c'' option.
>>I remember some time ago the problem which is mentioned on your blog:
>>"
>>And while dtrace(1M) has a -c option that lets you specify a command to 
>>examine with DTrace the dvm probes don''t show up until a little
later
>>when the JVM has
>>initialized. It''s worth mentioning that on the next version of
Solaris
>>(available via Solaris Express we''ve added a feature that lets
you
>>specify probes that don''t yet
>>exist that will be enabled when they show up; that feature will be in 
>>an early Solaris 10 update.
>>"
>>
>>Can you confirm whether it was already fixed in Solaris updates?
>>
>>thanks,
>>-katya
>>

That''s how it''s supposed to work. If it fires at all you
should be getting
the right results.

Adam

On Thu, Aug 10, 2006 at 07:46:48PM +0400, Ekaterina Pavlova
wrote:
> 
> 
> Adam Leventhal wrote:
> 
> >I think it''s probably been backported. If you''re
trying to enable a probe
> >that doesn''t exist and you''re getting that warning,
add the -Z option to
> >tell DTrace not to complain if your program matches zero probes.
> >
> Yes, I do exactly this.
> 
> Some more info just to be sure everything works as it is supposed to work
:)
> 
> In jdk6.0 hotspot probes definitions are defined in libjvm.so.
> As result when you try to do something like this:
> # dscript.d -c "java ...."
> 
> hotspot probes could be not loaded yet (because libjvm.so could be not 
> loaded yet)
> and as result warning like following ones are thrown:
> dtrace: failed to compile script ... : probe description 
> hotspot4530:::class-loaded does not match any probes
> 
> If I add flag ''-Z'' warnings will disappear.
> Also hotspot:::class-loaded probes begins to fire after some time
> 
> Can you confirm that doing this way I will not miss any 
> hotspot:::class-loaded probes?
> 
> Thanks a lot,
> -katya
> 
> >Adam
> >
> >On Thu, Aug 10, 2006 at 12:17:04PM +0400, Ekaterina Pavlova wrote:
> >
> >>Hi Adam,
> >>
> >>I working on Dtrace samples for mustang (jdk6) and come up with
question
> >>regarding ''-c'' option.
> >>I remember some time ago the problem which is mentioned on your
blog:
> >>"
> >>And while dtrace(1M) has a -c option that lets you specify a
command to
> >>examine with DTrace the dvm probes don''t show up until a
little later
> >>when the JVM has
> >>initialized. It''s worth mentioning that on the next
version of Solaris
> >>(available via Solaris Express we''ve added a feature that
lets you
> >>specify probes that don''t yet
> >>exist that will be enabled when they show up; that feature will be
in
> >>an early Solaris 10 update.
> >>"
> >>
> >>Can you confirm whether it was already fixed in Solaris updates?
> >>
> >>thanks,
> >>-katya
> >>
-- 
Adam Leventhal, Solaris Kernel Development       http://blogs.sun.com/ahl