On Mon, Aug 27, 2012 at 06:09:21AM +0200, Petr Gajdůšek wrote: > Hello, > > First off all, I beg your pardon for my pure English. > > IMO the only big problem is that libprocps does not export enough > symbols. Currently it doesn't export even the essential freeproc() [1]. > > I filled a bug report against libprocps0-dev [2]. And ss soon as those > symbols begins to be exported, the ulatencyd will be easily patchable. > > I created a branch on github, that uses such external libproc. It > consists of rather trivial code removal and of the cmake system update. > Though, it probably depends on changes I have committed before, these > come to my mind: > > - let readproc/readtask allocate proc_t storage instead of reusing > single buffer (i.e. call them with NULL). This is maybe not necessary > but it allows ulatencyd to more abstract from libproc internals. > > - reimplement and improve u_proc.cgroups_* handling in ulatencyd core, > this is IMHO needed for #629380 [3] fix anyhow. > > I have also temporary disable the "groups" property of u_proc instances > exported tu lua (not used anywhere). > > The branch is https://github.com/gajdusek/ulatencyd/commits/libprocps-ng > with currently one commit: > "Remove embedded libproc and coreutils, use system libprocps (issue > #37)" > https://github.com/gajdusek/ulatencyd/commit/b30c5ef10c63b978e16a2f8addd5ed4a02761944 > > It seems to work fine. > > I haven't tried to backport it to the Poelzi's master yet. I have fixed > some bugs since that, including a try of fixing #629380 [3] and the > patches would overlap. > > Moreover the patches would divert from upstream rather much. > > Any ideas?
Thank you for your work. AFAICT your patch does not look that much invasive, but
it currently does not apply cleanly. Also, what about the FIXMEs, are you going
to fix them?
Cheers
--
perl -E '$_=q;$/= @{[@_]};and s;\S+;<inidehG ordnasselA>;eg;say~~reverse'
signature.asc
Description: Digital signature
