Re: Making the transport layer more robust

[Mark Wielaard <mjw at redhat dot com>]

On Tue, 2011-07-19 at 10:58 +0200, Mark Wielaard wrote:
> I haven't yet tested on really old kernels, like those used on rhel4,
> where we are using an almost completely different transport layer
> implementation (even using different message types in some cases).
> I'll try and find such an old machine and test a bit (does anybody
> still use such an old beast with systemtap git trunk?). But in general
> I think we should deprecate this part of the transport layer. It makes
> things really complex.

I found a RHEL4-U8 x86_64 install which uses a 2.6.9 (!) kernel. I fixed
things up so that it could be build and run at least the evil transport
layer testcase, and that seems to also work fine on such an old system.

But it would be nice to know whether people really expect systemtap
trunk to work on such old systems. Looking at some of the test failures
there is a lot that just doesn't work. And I was actually able to crash
the system (unrelated to the transport layer, but our kprobe entry code
seems flaky on such an old system). Is it really worth it to keep
supporting it? Dropping such old kernels would certainly make it
possible to clean up some tricky code.

Cheers,

Mark