Why OSADL's "Latest Stable"?
Posted Sep 13, 2012 7:14 UTC (Thu) by
cemde (subscriber, #51020)
Parent article:
LPC: The realtime microconference
Hi Darren,
thanks a lot for writing this report! In case someone still is asking "Why is OSADL's Latest Stable needed in addition to Steven's releases?", I would explain it with a comparison of the mainline vs the PREEMPT_RT release strategy.
Mainline:
A kernel with the -rcX suffix is not yet released; a kernel without this suffix is released. People then consider a released kernel as stable (although they may need to apply subsequent stable release patches).
PREEMPT_RT:
When adaption of the PREEMPT_RT patches starts for a given kernel, the -rtX
suffix is appended to its version, e.g. -rt1, but this only says that the patches are applied. The -rtX number then continuously increases and anywhere between, for example, 20 and 40, the thing becomes good enough so it can be used. But this very moment is not visible, since we do not use the -rcrtX suffix that eventually becomes -rtX.
Why OSADL's "Latest Stable"?
OSADL's "Latest Stable" simply was intended to make it visible when a PREEMPT_RT patch along with a given kernel version is suitable for use in all our test systems. Adding more and more systems to our QA farm for testing and requesting that the kernel must not only be suitable for use but for integration into an industrial product made it somewhat more difficult. But the principle still is the same, i.e. that we would like to mark the transition between the development period and the release period of a PREEMPT_RT patched kernel.
For the time being, crashes, regressions and other problems of the farm systems are collected at our farm but not sent out. I promise to send all of them on a regular basis to the new PREEMPT_RT Bugzilla tracker - everybody can then check out the open bugs and determine whether they affect a particular system or not. And OSADL's "Latest Stable" web page will be closely synchronized to this list. So we may have "particular stability" as opposed to "general stability". Integrators can then decide to use a PREEMPT_RT kernel version that apparently is stable on a particular system, while this version cannot be generally recommended. This hopefully will reduce the number of complaints that the OSADL's "Latest Stable" PREEMPT_RT kernel version is so much behind.
(
Log in to post comments)
