[Bug translator/2645] New: Need a syntax to specify optional probes

[Bug translator/2645] New: Need a syntax to specify optional probes

[joshua dot i dot stone at intel dot com]

see http://sources.redhat.com/ml/systemtap/2006-q2/msg00286.html

Summary of desired behavior: given

  probe FOO ? { }
  probe alias = BAR ?, BAZ ? { }

If FOO is not found, then that probe should be discarded.  If both BAR and BAZ
are not found, there are two cases:

  probe alias { } // ERROR: probepoint not found
  probe alias ? { } // ok, probe is discarded


Similarly, if a syscall "sys_foobar" isn't found:

tapset:
  probe syscall.foobar = kernel.function("sys_foobar") ? {}

user script:
  probe syscall.foobar {} // error, probepoint not found
  probe syscall.foobar ? {} // ok, probe is discarded

  probe syscall.* {} // ok if all non-optional syscall probes were found,
                     // or if all are optional, then at least one was found

-- 
           Summary: Need a syntax to specify optional probes
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: translator
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: joshua dot i dot stone at intel dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[bibo dot mao at intel dot com]

------- Additional Comments From bibo dot mao at intel dot com  2006-05-09 03:52 -------
what about this kind of tapset, user script should care nothing about whether is
it optional or not.

tapset:
  probe syscall.foobar = kernel.function("sys_foobar") ? {}     optional 
  probe syscall.foobar1 = kernel.function("sys_foobar1") {}     non-optional
 
user script:
  probe syscall.foobar  {}    // ok, probe is discarded
  probe syscall.foobar1 {}   // ERROR: probepoint not found

  probe syscall.* {} // ok if all non-optional syscall probes were found,
                     // or if all are optional, then at least one was found

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-05-09 21:09 -------
(In reply to comment #1)
> what about this kind of tapset, user script should care nothing about 
> whether is it optional or not.

The user doesn't know whether the lower level tapset has marked a binding as
optional or not.  Thus, when the user writes a probe on syscall.foobar, they
will likely be surprised if this probe was automatically removed.  We should
avoid surprising results.

At a minimum this should be an warning, but I still think it's better to be an
error.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2006-05-14 12:01 -------
If I understand the concern correctly, in this scenario,
   probe group.f = pp ? { }  # in tapset
   probe group.* { }         # in user script
and for missing "pp", should this succeed or fail?  There is an ambiguity.
With the current flavouring of the code, it would fail, since the group.*
match would expand to an empty set of derived_probe's.

With a little bit of extra work, we can let the tapset author make this pass.
Let's provide a new probe point "never", which would be similar to "begin"
and "end" but never actually run its handler.  (It doesn't even need to show
up in the translated C code.)  Then, if a tapset author wants to make the
alias succeed with a missing pp, even for non-optional users, he can say:
  probe group.f = pp ? , never { } # in tapset

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-05-14 20:22 -------
(In reply to comment #3)
> If I understand the concern correctly, in this scenario,
>    probe group.f = pp ? { }  # in tapset
>    probe group.* { }         # in user script
> and for missing "pp", should this succeed or fail?  There is an ambiguity.
> With the current flavouring of the code, it would fail, since the group.*
> match would expand to an empty set of derived_probe's.

I think failure is correct here.  With missing "pp", "group.*" should fail, yet
"group.* ?" would be ok.  If there were another alias, "probe group.g = qq ?
{}", then "group.*" will succeed iff at least one of "pp" and "qq" are found.

> With a little bit of extra work, we can let the tapset author make this pass.
> Let's provide a new probe point "never", which would be similar to "begin"
> and "end" but never actually run its handler.  (It doesn't even need to show
> up in the translated C code.)  Then, if a tapset author wants to make the
> alias succeed with a missing pp, even for non-optional users, he can say:
>   probe group.f = pp ? , never { } # in tapset

I think this is a good workaround if a tapset author really wants to write a
probe that could possibly do nothing.  While I personally think this should be
discouraged, for the reasons I gave in comment #2, it may be good to allow this
possibility to cover all use cases.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[joshua dot i dot stone at intel dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |2295
              nThis|                            |


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[fche at redhat dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|systemtap at sources dot    |fche at redhat dot com
                   |redhat dot com              |


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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

[Bug translator/2645] Need a syntax to specify optional probes

[fche at redhat dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
   Last reconfirmed|0000-00-00 00:00:00         |2006-06-02 14:59:54
               date|                            |


http://sourceware.org/bugzilla/show_bug.cgi?id=2645

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