Em Mon, Sep 24, 2018 at 11:16:32AM -0300, Arnaldo Carvalho de Melo escreveu: > Em Mon, Sep 24, 2018 at 04:55:23AM -0400, Steven Rostedt escreveu: > > On Sun, 23 Sep 2018 20:56:26 +0200 > > Jiri Olsa <jolsa@xxxxxxxxxx> wrote: > > > is there still something left to rename, or we can start moving it into public rpm? > > We're testing it as a external library, and we did find something. Did > > you get this patch? > > https://lore.kernel.org/lkml/20180921152037.1e23c7f4@xxxxxxxxxxxxxxxxxx > I'm finishing tests with the patchset before this last patch, will push > to Ingo soon, then will check if the patch with the private copy of the > strerror_r() wrapper passes on the systems that made that wrapper came > to be, i.e. systems using the musl libc (https://www.musl-libc.org/). Ok, it passes on these systems: 1 alpine:3.4 : Ok gcc (Alpine 5.3.0) 5.3.0 2 alpine:3.5 : Ok gcc (Alpine 6.2.1) 6.2.1 20160822 3 alpine:3.6 : Ok gcc (Alpine 6.3.0) 6.3.0 4 alpine:3.7 : Ok gcc (Alpine 6.4.0) 6.4.0 5 alpine:3.8 : Ok gcc (Alpine 6.4.0) 6.4.0 6 alpine:edge : Ok gcc (Alpine 6.4.0) 6.4.0 Which puzzled me, but then I saw that you guys are not using strerror_r(), but strerror() that is not thread-safe, and then the pitfalls with strerror_r() XSI/_GNU_SOURCE problems are not hit. This same problem happened with tools/lib/bpf/, and I fixed it in by having a separate file to create the wrapper, where we make sure _GNU_SOURCE is not defined and continue using strerror_r(): https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/commit/?h=perf/urgent&id=6d41907c630d3196be89c9ed5a7f8258486b3eaf I think it would be interesting to use something similar so that libtraceevent becomes thread safe for this specific point. For now I've applied the patch to my local perf/core branch, so that it gets tested in more systems. - Arnaldo
|