[elrepo-devel] perf-xxx.noarch.rpm

Dag Wieers dag at wieers.com
Thu Feb 10 08:34:53 EST 2011


On Thu, 10 Feb 2011, Dag Wieers wrote:

> On Wed, 9 Feb 2011, Alan Bartlett wrote:
>
>> On 8 February 2011 20:52, Steve Clark <sclark at netwolves.com> wrote:
>>
>>> I can't seem to find the perf-xxx.noarch.rpm for the ml-kernels.
>>
>> They are userland packages, are they not?
>>
>> Are you able to build them? Any success? Or a resounding failure?
>>
>> Sorry for the succession of questions, Steve. This is now getting into
>> a grey area with regards to exactly where the ELRepo Project's
>> provision / support ends . . .
>>
>> I seem to recall that Dag (Wieers) mentioned the idea of perf support
>> a number of months ago (before the birth of his daughter :-) ), so
>> let's see is what Dag's current view on the subject.
>
> A perf binary would indeed be useful, but beware, I remember that the
> binary was specific to a kernel version. So it would mean the binary is
> versioned as well and the alternatives-infrastructure needs to be used.
>
> Hence, the binary needs to be build _with_ the kernel, and probably also
> ship as part of the kernel RPM. Not sure how Fedora is doing this though.
> Probably worth looking into.

Reading up on what was send to the list, here a summary of what would be 
needed:

  - Compile the perf binary and package it with the kernel package
    eg.
        [root at moria media]# rpm -ql kernel | grep perf
        /usr/libexec/perf.2.6.32-71.7.1.el6.x86_64
        /usr/libexec/perf.2.6.32-71.14.1.el6.x86_64

  - Ship the (upstream) perf sub-package, which contains a wrapper shell
    script that picks up the above binary, and a set of man pages.

For RHEL5 the perf sub-package will not be a noarch package (as RHEL5's 
RPM cannot produce noarch sub-packages for arch packages).

If we would produce a RHEL6 kernel-ml package, only the former is needed, 
obviously.

-- 
-- dag wieers, dag at wieers.com, http://dag.wieers.com/
-- dagit linux solutions, info at dagit.net, http://dagit.net/

[Any errors in spelling, tact or fact are transmission errors]



More information about the elrepo-devel mailing list