[Bug testsuite/18154] New: let's be more forgiving on ppc64le

[Bug testsuite/18154] New: let's be more forgiving on ppc64le

[mcermak at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

            Bug ID: 18154
           Summary: let's be more forgiving on ppc64le
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: testsuite
          Assignee: systemtap at sourceware dot org
          Reporter: mcermak at redhat dot com

Created attachment 8205
  --> https://sourceware.org/bugzilla/attachment.cgi?id=8205&action=edit
proposed patch

Commit bf952a76444755f3bf0b8f7deb808e2b6b2ce7ac removed --skip-badvars from
syscall.exp. Given the suboptimal quality of debuginfo on this architecture,
this causes many testsuite failures. Let's consider conditionally putting
--skip-badvars back.

For the same reason it might be helpful to use prologue searching by default on
this architecture in the testsuite.

Proposed patch attached. Thoughts?

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mcermak at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

Martin Cermak <mcermak at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dsmith at redhat dot com

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[dsmith at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

David Smith <dsmith at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mjw at redhat dot com

--- Comment #1 from David Smith <dsmith at redhat dot com> ---
Here's what I get when compiling syscall.exp on a ppc64le system (using -P):

====
semantic error: unhandled DW_OP_GNU_entry_value in DWARF expression [0] at 0
(0xf3: 1, 70366814935779): identifier '$which_clock' at
/usr/local/share/systemtap/tapset/linux/syscalls.stp:525:49
        source:         argstr = sprintf("%s, %p",
_get_wc_str(__int32($which_clock)), $tp)
====

This looks like an instance of bug #16596 (Support DW_OP_GNU_entry_value in
location expressions). We probably need to get Mark to look at this and see
what he can find.

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mcermak at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

--- Comment #2 from Martin Cermak <mcermak at redhat dot com> ---
Yup, that's what I saw too:
https://bugzilla.redhat.com/show_bug.cgi?id=1204249.

After checking with FChE I checked the -P hunk in as
958356595c659111a5494b05feea125444de2ecc and got bugzilla mid-air collision
when attempting to report it here..

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mjw at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

--- Comment #3 from Mark Wielaard <mjw at redhat dot com> ---
(In reply to David Smith from comment #1)
> This looks like an instance of bug #16596 (Support DW_OP_GNU_entry_value in
> location expressions). We probably need to get Mark to look at this and see
> what he can find.

I am not sure prologue searching is such a good idea. I haven't analyzed many
cases yet. But at least for asmlink functions it seems -P causes us to put a
probe "too late". When we don't probe at the function start there is a
probability of the code already having clobbered the registers that hold the
function arguments. In which case we are too late and all DWARF gives us is a
location description using DW_OP_GNU_entry_value. Basically telling us that the
value is only available earlier.

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mjw at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

Mark Wielaard <mjw at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|                            |17638

--- Comment #4 from Mark Wielaard <mjw at redhat dot com> ---
I have been testing with the proposed patches of
https://sourceware.org/bugzilla/show_bug.cgi?id=17638 and removed the the -P
hunk change added in commit 958356595c659111a5494b05feea125444de2ecc. Looks
like that produces much better results. With those patches probes are places at
locations that give stap a location description that makes it possible to
resolve much more make installcheck RUNTESTFLAGS=syscall.exp:

Running ./systemtap.syscall/syscall.exp ...

        === systemtap Summary ===

# of expected passes        133
# of untested testcases        2
# of unsupported tests        5

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mjw at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154
Bug 18154 depends on bug 17638, which changed state.

Bug 17638 Summary: Symbol resolution broken for PPC64 ABIv2
https://sourceware.org/bugzilla/show_bug.cgi?id=17638

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

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

[Bug testsuite/18154] let's be more forgiving on ppc64le

[mjw at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18154

Mark Wielaard <mjw at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #5 from Mark Wielaard <mjw at redhat dot com> ---
commit 11d578d28e52068e4b81f36ea84f217467037e09
Author: Mark Wielaard <mjw@redhat.com>
Date:   Wed Apr 22 15:55:38 2015 +0200

    Revert "Turn prologue searching on by default for ppc64le in the
testsuite."

    This reverts commit 958356595c659111a5494b05feea125444de2ecc.

    No longer necessary after Bug 17638 - Symbol resolution broken for PPC64
    ABIv2 was fixed with commit b4c6a4.

Tested on ppc64le 3.10.0, gcc 4.8.3, elfutils 0.160, glibc 2.17

Running ./systemtap.syscall/syscall.exp ...

        === systemtap Summary ===

# of expected passes        138
# of untested testcases        2
# of unsupported tests        5

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