Re: [Crash-utility] Crash utility on Android

----- Original Message ----- Not that it's necessarily indicative of the problem at hand, but I am curious as to how your build came up with a target of "arm-linux-androideabi"? When building crash with "make target=ARM", the configure command argument passed to the embedded gdb build is "--target =arm-elf-linux", and on a RHEL or Fedora x86_64 host, "gnu" gets appended to it, such that the resultant target is shown like this in the gdb banner: This GDB was configured as "--host=x86_64-unknown-linux-gnu --target=arm-elf-linux"... Is your x86_64 host machine also running an Android environment? Anyway, it actually made it quite far into the initialization before it fails with the "crash: cannot hash task_struct entries", which I believe is caused by a bug in the ARM code here in the unwind_arm.c function init_module_unwind_tables(): hq_open(); cnt = do_list(&ld); if (cnt == -1) { error(WARNING, "UNWIND: failed to gather unwind_table list"); return FALSE; } If you put a call to hq_close() here, it should at least continue on a bit farther in the initialization sequence: hq_open(); cnt = do_list(&ld); if (cnt == -1) { error(WARNING, "UNWIND: failed to gather unwind_table list"); + hq_close(); return FALSE; } I'm not sure what's happening w/respect to the crash_notes and unwind info gathering. The ARM guys on this list may have some ideas about that? Also, when posting crash initialization-type failures, can you also attach a full debug output by putting "-d8" on the crash command line? Thanks, Dave