From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ezequiel Garcia Date: Sun, 31 Aug 2014 18:13:59 -0300 Subject: [Buildroot] Analysis of build results In-Reply-To: <5403544E.6010506@frajasalo.de> References: <20140825063011.07B50101AB1@stock.ovh.net> <20140825181614.3f5c22a0@free-electrons.com> <20140828125217.GA3389@arch.cereza> <5403544E.6010506@frajasalo.de> Message-ID: <20140831211358.GA3134@arch.cereza> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net On 31 Aug 06:58 PM, Frank Bergmann wrote: > On 28.08.2014 Ezequiel Garcia wrote: > >(Ccing Frank) > > > >On 25 Aug 06:16 PM, Thomas Petazzoni wrote: > >>> nios2 | util-linux-2.24.2 | NOK | http://autobuild.buildroot.net/results/fd30974bf2f976c600717268ce340f73d48dda34/ > >> > >>sys-utils/prlimit.o: In function `main': > >>prlimit.c:(.text.startup+0x440): undefined reference to `prlimit64' > >>prlimit.c:(.text.startup+0x5d0): undefined reference to `prlimit64' > >>collect2: error: ld returned 1 exit status > >> > >>Ezequiel, an idea? > >> > > > >This is probably the same missing fallocate64 deal. The function is simply not > >available, yet detected (Frank: please correct me if I'm wrong). > > > >By doing something similar to what we do for e2fsprogs: > > > >+ifeq ($(BR2_nios2),y) > >+UTIL_LINUX_CONF_ENV += ac_cv_func_prlimit=no > >+endif > >+ > > > >I can avoid the SYS_prlimit64 usage. However, we then get conflicting types: > > > > CC sys-utils/prlimit.o > > CC sys-utils/lscpu.o > >sys-utils/prlimit.c:139:12: error: conflicting types for 'prlimit' > >In file included from /home/zeta/buildroot/buildroot-nios2/output/host/usr/nios2-buildroot-linux-gnu/sysroot/usr/include/sys/resource.h:25:0, > > from sys-utils/prlimit.c:28: > >/home/zeta/buildroot/buildroot-nios2/output/host/usr/nios2-buildroot-linux-gnu/sysroot/usr/include/bits/resource.h:313:12: note: previous declaration of 'prlimit' was here > > > >I wonder if anyone has an idea to disable the prlimit build, and all the > >commands that depend on it. > > > >Notice that we have a similar issue on other packages, but the proposed > >solution's have been too hacky to be accepted: > > > >http://lists.busybox.net/pipermail/buildroot/2014-May/096203.html > > I see a new toolchain on the mentor site: 2014.05-47 from August. Maybe Thanks a lot for the hint! > this release fixes some of the issues we have with the current > toolchain. I will make some tests in the next week(s). > Indeed, the new toolchain seem to be working much better now. Packages that have issues with the 2013.05 toolchain build fine with the new one, such as util-linux, e2fsprogs (reverting the fallocate64 modifications), fio, and dhcpd. As far as I can see, they have fixed the issues with fallocate64, prlimit64 and the broken linux headers. As agreed with Peter on the IRC, I'll push a patch adding the new toolchain and leave the old one around for a while. Any further test is welcome. -- Ezequiel Garc?a, Free Electrons Embedded Linux, Kernel and Android Engineering http://free-electrons.com