Bug#604755: perf timechart: segfault in perf_session__process_events

[Jonathan Nieder <jrnieder@gmail.com>]

Package: linux-tools-2.6.36
Version: 2.6.36-1~experimental.1
Tags: upstream

Can't seem to get "perf timechart" working.

  $ uname -r
  2.6.36-trunk-686
  # perf timechart record echo hi
  hi
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.079 MB perf.data (~3471 samples) ]
  # chmod a+r perf.data
  $ gdb --args perf_2.6.36 timechart
[...]
  (gdb) run
  Starting program: /usr/bin/perf_2.6.36 timechart
  [Thread debugging using libthread_db enabled]

  Program received signal SIGSEGV, Segmentation fault.
  0x08061969 in ?? ()
  (gdb) bt
  #0  0x08061969 in ?? ()
  #1  0x0808998e in ?? ()
  #2  0x080888a6 in ?? ()
  #3  0x080894b8 in __perf_session__process_events ()
  #4  0x08089890 in perf_session__process_events ()
  #5  0x08060041 in cmd_timechart ()
  #6  0x0805210e in ?? ()
  #7  0x080527cd in main ()

Also was reproducible with upstream linux and perf 2.6.37-rc3.
Valgrind trace (source line numbers refer to v2.6.37-rc3):

	Invalid read of size 4
	   at 0x805B5C1: process_sample_event (builtin-timechart.c:505)
	   by 0x808654D: process_finished_round (session.c:410)
	   by 0x8085CF5: perf_session__process_event (session.c:633)
	   by 0x808732F: __perf_session__process_events (session.c:827)
	   by 0x80875CF: perf_session__process_events (session.c:867)
	   by 0x805BDE0: cmd_timechart (builtin-timechart.c:949)
	   by 0x804CCED: run_builtin (perf.c:286)
	   by 0x804D47E: main (perf.c:357)
	 Address 0x5416558 is 8 bytes after a block of size 72 alloc'd
	   at 0x4023F50: malloc (vg_replace_malloc.c:236)
	   by 0x8085EFA: perf_session__process_event (session.c:553)
	   by 0x808732F: __perf_session__process_events (session.c:827)
	   by 0x80875CF: perf_session__process_events (session.c:867)
	   by 0x805BDE0: cmd_timechart (builtin-timechart.c:949)
	   by 0x804CCED: run_builtin (perf.c:286)
	   by 0x804D47E: main (perf.c:357)

Is this a known problem?  Where should it be reported?

(Ooh, this time it wrote a timechart before segfaulting!  Apparently
v2.6.37-rc3 userspace + debian 2.6.36 kernel is the recipe for
success...)

Ciao,
Jonathan