On Sat, Mar 03, 2012 at 10:10:06AM +0100, Sami Kerola wrote: > it seems 'license: update FSF addresses' is missing, and required by Fedora. > See I've got Jim's patch in there now. > Another patch which is missing without reason is 'w: use utmp > definitions to determine array sizes'. This patch, though, is > unimportant clean up. I attached both patches to this email to ease > merging. Not sure what happened there, i remember seeing that go in at one stage because I was curious about %zu anyhow the fixes are in. > The abort() is there for reason. One should never reach > catastrophic_failure(), but if that happens core dump is desirable. We > should look why 'impossible' condition happen, and fix the reason. Yes that's correct. The problem is why would changing something inside a function mean the function is called or not? exit isn't perfect but it means it works. > How about having libprocps redesign mail thread, from which the best > bits would be wrote to wiki, see git address below? For example I've That sounds good. > thought how I would like to use a lib, and I came to conclusion that > nearly always I want to see delta values of measurements e.g. PID -> > usr cpu in n usec. Doing that for each process separately is horrible > idea, so the lib should do native selecting what is being measured. It You mean select the pids to get the information? > is probably good idea to split functions per selection type, such as > kernel, processes, disks, network, [...], measurements. Fetching They can be broken up basically around what proc file they open. They're pretty isolated to each other this way and opening another file you don't need is expensive. Probably should try to look at one of the simpler ones than processes first to get a good feel for what is needed to be changed. - Craig -- Craig Small VK2XLZ http://enc.com.au/ csmall at : enc.com.au Debian GNU/Linux http://www.debian.org/ csmall at : debian.org GPG fingerprint: 5D2F B320 B825 D939 04D2 0519 3938 F96B DF50 FEA5