systemtap on arm and prelink

systemtap on arm and prelink

[William Cohen]

Looks like Fedora 14 for ARM had a relatively old version of systemtap in it. There is a newer version being built right now (but it is going to take a while) http://arm.koji.fedoraproject.org/koji/buildinfo?buildID=30277

The new version of systemtap-testsuite will still have the dependency on prelink which isn't available for arm. Maybe remove that dependency for arm versions. Any thoughts on the prelink dependency?



-Will

-------- Original Message --------
Subject: [fedora-arm] ARM F-14 Branched report: 20110823 changes
Date: Tue, 23 Aug 2011 14:18:56 +0000
From: ARM Branched Report <rawhide@fedoraproject.org>
To: arm@lists.fedoraproject.org, secondary@lists.fedoraproject.org

Compose started at Tue Aug 23 12:53:27 UTC 2011

Broken deps for arm
----------------------------------------------------------

	systemtap-testsuite-1.4-6.fc13.armv5tel requires prelink

Re: systemtap on arm and prelink

[David Smith]

On 08/23/2011 02:44 PM, William Cohen wrote:
> The new version of systemtap-testsuite will still have the dependency on prelink
> which isn't available for arm. Maybe remove that dependency for arm
versions. Any
> thoughts on the prelink dependency?

We've got a "hard" dependency on prelink (explicitly listed in the
specfile).  If prelink isn't there, they only bad result that should
happen is extra failures in systemtap.exelib/exelib.exp.

So, it seems reasonable to remove that dependency for arm.  I'll take a
whack at exelib.exp and make sure it fails gracefully when prelink isn't
there.

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)

Re: systemtap on arm and prelink

[David Smith]

On 08/23/2011 03:25 PM, David Smith wrote:
> I'll take a whack at exelib.exp and make sure it fails gracefully when prelink isn't
> there.

exelib.exp fails pretty gracefully when prelink isn't there.  It just
causes 3 extra failures.

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)

Re: systemtap on arm and prelink

[William Cohen]

On 08/23/2011 04:44 PM, David Smith wrote:
> On 08/23/2011 03:25 PM, David Smith wrote:
>> I'll take a whack at exelib.exp and make sure it fails gracefully when prelink isn't
>> there.
> 
> exelib.exp fails pretty gracefully when prelink isn't there.  It just
> causes 3 extra failures.
> 

Thanks, David. Sounds like the prelink dependency could be removed for systemtap-testsuite requirements.

-Will

Re: systemtap on arm and prelink

[David Smith]

On 08/23/2011 04:09 PM, William Cohen wrote:
> Thanks, David. Sounds like the prelink dependency could be removed
> for systemtap-testsuite requirements.

Hmm, personally I'd vote for just removing the prelink dependency for
arm.  If we remove it for all arches, we're going to be confused by the
testcase regression (assuming the system doesn't have prelink installed).

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)

Re: systemtap on arm and prelink

[Mark Wielaard]

On Tue, 2011-08-23 at 16:19 -0500, David Smith wrote:
> On 08/23/2011 04:09 PM, William Cohen wrote:
> > Thanks, David. Sounds like the prelink dependency could be removed
> > for systemtap-testsuite requirements.
> 
> Hmm, personally I'd vote for just removing the prelink dependency for
> arm.  If we remove it for all arches, we're going to be confused by the
> testcase regression (assuming the system doesn't have prelink installed).

Yes, please only remove it for architectures that really don't have
prelink available. The exelib prelink tests have shown some tricky
problems in the past, I would like them to PASS as much as possible.

Thanks,

Mark

Re: systemtap on arm and prelink

[William Cohen]

On 09/06/2011 11:12 AM, Mark Wielaard wrote:
> On Tue, 2011-08-23 at 16:19 -0500, David Smith wrote:
>> On 08/23/2011 04:09 PM, William Cohen wrote:
>>> Thanks, David. Sounds like the prelink dependency could be removed
>>> for systemtap-testsuite requirements.
>>
>> Hmm, personally I'd vote for just removing the prelink dependency for
>> arm.  If we remove it for all arches, we're going to be confused by the
>> testcase regression (assuming the system doesn't have prelink installed).
> 
> Yes, please only remove it for architectures that really don't have
> prelink available. The exelib prelink tests have shown some tricky
> problems in the past, I would like them to PASS as much as possible.
> 
> Thanks,
> 
> Mark

Hi Mark,

Might be able to have this enabled for future ARM rpms. Jakub made some fixes for prelink on arm for https://bugzilla.redhat.com/show_bug.cgi?id=733089 and I have prelink working on my arm machine. These fixes have been rolled into the rawhide rpms:

http://koji.fedoraproject.org/koji/buildinfo?buildID=260584

However, not yet in the arm koji.

-Will