[Bug translator/20599] New: collected bit shift & variance rewritten by new histogram during elaboration

["mcermak at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=20599

            Bug ID: 20599
           Summary: collected bit shift & variance rewritten by new
                    histogram during elaboration
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: translator
          Assignee: systemtap at sourceware dot org
          Reporter: mcermak at redhat dot com
  Target Milestone: ---

Created attachment 9508
  --> https://sourceware.org/bugzilla/attachment.cgi?id=9508&action=edit
proposed patch

Following script fails to compute variance correctly:

=======
$ stap --poison-cache -e 'global x probe oneshot { x<<<1 x<<<10 x<<<100
println("v: ", @variance(x, 3)) print(@hist_linear(x, 0, 100, 1)) }'
v: 1944
value |-------------------------------------------------- count
    0 |                                                   0
    1 |@                                                  1
    2 |                                                   0
    3 |                                                   0
      ~
    8 |                                                   0
    9 |                                                   0
   10 |@                                                  1
   11 |                                                   0
   12 |                                                   0
      ~
   98 |                                                   0
   99 |                                                   0
  100 |@                                                  1

$ 
=======

In the above example, variance should have been 2997, not 1944:

=======
$ stap --poison-cache -e 'global x probe oneshot { x<<<1 x<<<10 x<<<100
println("v: ", @variance(x, 3)) }'
v: 2997
$ 
=======

One can check the `... -p3 | grep _stp_stat_init` output to see that in the
first case, new histogram rewrites collected stat_ops and bit_shift for given
global.  This problem comes from the elaboration phase.  Attached patch fixes
it.

-- 
You are receiving this mail because:
You are the assignee for the bug.