[news.eclipse.technology.linuxtools] Re: OProfile on openSUSE 11.1

[aegges@xxxxxx (Axel Mueller)]

Kent Sebastian wrote:

> Axel Mueller wrote:
> > Axel Mueller wrote:
> >
> > > Hold your breath. It is now nearly working!!
> >
> > > At least I can choose the events in the launch configuration and have 
> > > a look  on the oprofiled log.
> >
> > > Guess what, I have substituted the opxml in the x86_64 directory with 
> > > the 32bit exe!!! I am running a 32bit OS!
> > > I came to this strange idea because there was a log entry from the p2 
> > > installer mocking about something in the x86_64 directory.
> > The patch from here https://bugs.eclipse.org/bugs/show_bug.cgi?id=265495 
> > resolves the problem (about the event view not the session view).
> >
> >  Axel

> Very odd.. The x86 and x86_64 fragments have a platform filter, and only 
> one is loaded. If the log entry was something like this:

> !MESSAGE Bundle 
reference:file:/notnfs/ksebasti/workspace/org.eclipse.linuxtools.oprofile.core.linux.x86/ 
> [264] was not resolved.
> !SUBENTRY 2 org.eclipse.linuxtools.oprofile.core.linux.x86 2 0 
> 2009-02-20 10:20:53.598
> !MESSAGE Platform filter did not match: (& (osgi.os=linux) (osgi.arch=x86))

> then that's normal (I'm on x86_64).
I don't remember the exact error message but it was different. I do not 
see this message now.


> If the plugin couldn't find the opxml binary then it would spit out an 
> error saying so, so I'm even more puzzled as to why putting the 32bit 
> binary in the 64bit fragment helps. That patch was applied to trunk 
> yesterday but I'm not sure if the nightly update site caught it yet, 
> although since I'm running the same version as you (except openjdk vs 
> sun jdk, but that shouldn't be an issue AFAIK), I don't see why it 
> wouldn't work.
I guess the plugin did find the opxml binary (the file was actually there) 
but had problems to run it. The patch was not in the nightly build. I 
applied it manually.

> By the session view you mean the samples not showing any details past 
> the '100% in <binary name>'? 
Yes.

> I'm looking into this since a colleague of 
> mine is having the same problem, (although profiling the same binary is 
> showing results on my system, puzzling yet again).

> I'll be sure to test 0.2 on various distros :p


> Thanks,

> -Kent

 Axel