biswarup dasgupta wrote:What version of oprofile are you using? Have you tried the latest that's available -- oprofile-0.9.7-RC2?
> I am trying to run opreport command in Android 2.3.3 inside VMware and the
Just out of curiosity . . . why do you need to use --session-dir? I take it you specified the same session-dir when you started oprofile, right?
> command always crashes when I give the symbol path as seen below:
> *release$* ./prebuilt/linux-x86_64/oprofile/bin/opreport --session-dir=.
Why "--merge=all"? What separation parameters are you running with? Generally, I find it most helpful to run with "--separate=lib,kernel". But then using "--merge=all" would negate the benefit.
> --merge=all -d -p $OUT
> *Segmentation fault*
> Overflow stats not available
> line is # ARM V7 PMNC possible unit maskswarning:
> /data/data/com.lifevibes.mediashow/files/libvideoeditor_jni.so could not be
> warning: /dev/ashmem/dalvik-jit-code-cache (deleted) could not be found.
> warning: /no-vmlinux could not be found.
> warning: /system/bin/akmd could not be found.
> warning: /system/lib/egl/libEGL_adreno200.so could not be found.
> warning: /system/lib/egl/libGLESv1_CM_adreno200.so could not be found.
> warning: /system/lib/egl/libGLESv2_adreno200.so could not be found.
> warning: /system/lib/egl/libq3dtools_adreno200.so could not be found.
> warning: /system/lib/libgsl.so could not be found.
> warning: /system/lib/libhtc_ril.so could not be found.
> CPU: ARM V7 PMNC, speed 0 MHz (estimated)
> Counted CPU_CYCLES events (Number of CPU cycles) with a unit mask of 0x00
> (No unit mask) count 1500
Have you tried running opreport under gdb to get a backtrace when the seg fault occurs? If you're using a distro-provided oprofile, it's likely that opreport is stripped. If so, then build/install oprofile-0.9.7-rc2. If you get the same error, then get a backtrace.
Just to be clear about how the "-p" is used . . . opreport will only look in the location specified by the "-p" option if it does not find the binary it's looking for in the expected place. If it finds a stripped binary in, say, /usr/bin, it will use that binary and *not* look in your specified image-path location.
> I have the oprofile data and if I don't specify the -p option then I can see
> the profile distribution library-wise.
> But I want to see where in the library the actual time is spent so I specify
> the symbol path.
> However, it always crashes when the path $OUT is given. In my setup, the
> value of $OUT is release/out/target/product/<product_name>/symbols
> I have created a soft-link of the arm-eabi version of objdump inside my
> release folder and have also added the release folder at the start of my
> $PATH variable. Still the crash is there.
> Can anyone provide any help ?
> Thanks in advance
> All of the data generated in your IT infrastructure is seriously valuable.
> Why? It contains a definitive record of application performance, security
> threats, fraudulent activity, and more. Splunk takes this data and makes
> sense of it. IT sense. And common sense.
> oprofile-list mailing list