On Wed, Aug 11, 2004 at 03:01:23PM -0700, Mary Edie Meredith wrote: > ..snip > > So, what kind of regression testing do you all have planned for the > > in-kernel code? That would be quiet useful to have done and welcomed by > > the kernel community (but it's still not a justification to create a SIG > > just to work on regression tests...) > > Ok, to answer this I need to clarify something about this > SIG and OSDL SIGs in general. > > OSDL created SIGs so that we (the OSDL members) could > focus our attention on specific areas - so far Storage, > Security, and now HotPlug. I still strongly object to the title "Hotplug". Remember, that now encompasses every virtual and real device in the entire kernel. It's no different from calling it a "Device" SIG. :) Please strongly consider changing it. > Yes, of course, there are other mail lists. We didn't > really intend for SIGs to be the center of the > universe for community conversation about these > areas. Our intent is not to replace those, in > fact we expect that active developer-type participants > on this list will need to be signed up for the > hotplug-relevant lists in addition to this one. So again, you are asking people to come to you, instead of going to them. Please see my previous emails in this thread about why that is not a good idea, and comes across rudely. > We will post on those lists when those communities are needed. This hotplug SIG was not posted on any of those lists that I saw. Why not? It's not like the knowledge of those lists are hidden anywhere... > So, this Hotplug SIG was requested at the last DCL > TWG face to face in June by the members because up > to now we hadn't been looking into the subject. You hadn't? I remember many different conversations about hotplug-this and hotplug-that over the previous years of CGL and DCL meetings. The meeting notes back this up. Hotplug pci, hotplug storage, hotplug usb, hotplug ieee1394, hotplug ide, hotplug pcmcia, and hotplug scsi have all been discussed in the past at meetings that I have attended. I'm sure there were others as I did not attend them all. Now, looking into memory-hotplug or cpu-hotplug or io-node-hotplug is a different thing. I suggest that if this group is to restrict themselves to those types of groups, you form 3 different SIGs since the topics covered is so varied, and hold nothing in common besides a single word. > We are just getting started. So we have done no > planning in the dark, nothing done before now to > determine what makes sense for OSDL. This is the > start. I'm not claiming you did any planning in the dark or light, that's pretty obvious due to the confusion by everyone involved here :) How about you disband this SIG, go back to the working groups and ask specifically what they want a SIG for, and then create that. Doing some planning upfront will help prevent this kind of problem in the future. > That said, I believe that regression testing will > be on the list of activities. Since you remarked > in a previous comment that "there is code present > to test CPU hotplug on machines that can not electrically > remove CPUs", is there any reason that code cannot > run today on STP? That would make a mighty fine > regression test, it seems to me. I admit ignorance > about where to get our hands on the code. If you > could be so kind as to enlighten us, I know our > testdev group would be very interested in investigating. I don't know where such code is, as I'm not actively involved in that development effort. I learned of such code by attending Rusty's talk at OLS. I suggest you read his OLS paper on the topic, and look a the code in the kernel, and anywhere else he refers to. It shouldn't be that difficult to find. Good luck, greg k-h
