From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Sat, 10 Feb 2018 18:45:52 +0100 Subject: [Buildroot] [PATCH v2 1/1] LIBLO : bump version to 0.29 In-Reply-To: References: <1518118669-24714-1-git-send-email-alexbaldwinmusic@gmail.com> <20180208230400.06741487@windsurf.lan> Message-ID: <20180210184552.14500165@windsurf.lan> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hello Alex, On Sat, 10 Feb 2018 17:19:16 +0100, Alex Baldwin wrote: > I am not sure the best place to address this so please correct me :) I > noticed that this patch failed with your submission as well as Matt > Webber's, but for different reasons. > > It seems that on yours > http://autobuild.buildroot.net/results/71a6d7e02a50f6cb53c70c38e82e2a2ae8af13ea/build-end.log > there is a warning about having to redirect to in the > file "server.c:54". This could be sorted with a simple patch, is it > acceptable to patch for this? Yes, it is acceptable. In addition to fixing this particular problem, it would be nice to avoid using -Werror. Indeed, the problem here is that this warning is treated as an error, causing a build failure. While -Werror makes sense during development, it doesn't make much sense for releases. > On Matt's > http://autobuild.buildroot.net/results/c8b/c8bed3a3fa7d2b2258f573cbfcb01af07419e0bf/build-end.log > we fail because undefined reference to __atomic_fetch_add_4 . I think a > possible fix could be manually linking to libatomic. I am not entirely sure > why, seems that some compilers are happy to find libatomic on their own and > others aren't. Is this something that buildroot would normally patch for? The libatomic situation depends on the architecture. On most architectures, __atomic_fetch_add_4 is a compiler builtin, so you don't need to link with libatomic. However, on SPARC (which is the CPU architecture on which this build failure occurs), libatomic is needed for __atomic_fetch_add_4. You have two solutions to handle that: (1) Just pass LIBS="-latomic" in LIBLO_CONF_ENV when BR2_TOOLCHAIN_HAS_LIBATOMIC=y (2) Patch the configure.ac script so that it tests if linking with libatomic is needed or not, and links with it if needed. In any case, if you do a patch against liblo, don't forget to submit it upstream. Thanks! Thomas -- Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons) Embedded Linux and Kernel engineering http://bootlin.com