Systemtap scripts unable to compile on RHEL4 U2
Vara Prasad
prasadav@us.ibm.com
Wed Sep 28 01:00:00 GMT 2005
William Cohen wrote:
> Keshavamurthy, Anil S wrote:
>
> <...>
>
>> The above procedure is okay for the developers, but not advisable for
>> end users I think.
>> If Systemtap RPM is installed then all its dependencies should be met
>> else I think there is some bug and people get confused.
>
>
> Having a dependency on kernel-debuginfo is not going to work. It
> breaks the way that the distribution set of rpms is computed. Red Hat
> doesn't currently ship the kernel-debuginfo.
>
> Also there is no guarantee that the kernel-debuginfo installed is the
> one for the kernel being instrumented. It is quite possible to get
> messages saying "no match for probe point" even with a
> kernel-debuginfo installed if it doesn't match the kernel being
> instrumented.
>
> SystemTap should be a bit more vocal when it is unable to find any
> debuginfo for the kernel being instrumented, rather than having the
> developer determine whether the debug information is missing or
> whether the developer mis-typed the function name.
>
> -Will
>
This brings up couple of questions.
1) Is it not a good idea as part of the very first check translator can
do a version check between the debug info it is using and the version of
the binary where it is going to insert probes?
2) If a customer upgrades the OS release how do we make sure the
corresponding debug info matches without corresponding debug RPM's shipped.
More information about the Systemtap
mailing list