added oprofile-list back to cc.
Jan Brunnert wrote:
> Maynard Johnson wrote:
>>> ok, i tried again with a modified command
>>> opcontrol --stop ; opcontrol --dump ; opreport -l | head -n 20 ;
>>> opcontrol --reset ; opcontrol --start
>>> after 5 times the same stuff happened again, this time the last words in
>> The "same stuff", meaning you're still seeing the corrupted double-lined list? The message below ("sfile_log_sample_count: Invalid argument") results in the oprofile daemon aborting. Never seen that one before either. Do you have *any* problems if you simply run oprofile in the usual way from the command line? That is . . .
>> 1. start oprofile
>> 2. start your application
>> 3. wait a while to collect profile data
>> 4. stop oprofile
>> 5. generate report
>> 6. analyze report
> the failure occurs also entirely without using my script by just issuing
> the above command several times.
Assuming you're *not* getting the corrupted double-lined list error anymore, right? So "the failure" now is the "sfile_log_sample_count: Invalid argument" error? Something unique about your environment or procedures, since I don't believe this error has ever been reported before.
>> I don't understand why your script is generating so many reports? Why don't you just let the profiler run for a longer period of time without interruption? I'm not saying you're not running into a real bug here. But reproducing, debugging and fixing may be very difficult and time consuming. For you to make progress, maybe the best thing is to figure out a way to use oprofile that avoids these errors. Also, please upgrad to 0.9.6 if you haven't already.
> the reason why the script is frequently processing reports is that i
> need to profile the behavior of a function depending on parameters that
> are passed into the function. therefore i need to save a result before
> and after the function is called. i tried opcontrol --save instead of
> evaluating a report on the fly but that also makes the daemon die.
> i will try upgrading to 0.9.6 and see if that helps. i will also try to
> run outside of the xen environment and see if that works better.
I doubt that 0.9.6 will help, but if you can, try changing the version first before running outside of xen. That way, if the problem goes away, you'll know which action helped.
> many thanks,