Date: Sun, 6 Nov 2016 11:30:55 +1100 From: Andrew Reilly <areilly@bigpond.net.au> To: svn-src-head@freebsd.org, ed@freebsd.org Subject: Re: svn commit: r307227 - in head: include lib/libc/stdlib lib/libc/tests/stdlib Message-ID: <B5ED025A-9F83-43A9-A379-639A4C013225@bigpond.net.au>
next in thread | raw e-mail | index | archive | help
Hi there, Re: = https://lists.freebsd.org/pipermail/svn-src-head/2016-October/092694.html In a rebuild of stable/11 today (and yesterday), the build stops in = lib/libc/tests/stdlib: /usr/src/lib/libc/tests/stdlib/tsearch_test.c:37:20: error: unknown type = name 'posix_tnode' tnode_assert(const posix_tnode *n) ^ /usr/src/lib/libc/tests/stdlib/tsearch_test.c:82:2: error: use of = undeclared identifier 'posix_tnode' posix_tnode *root =3D NULL; ^ /usr/src/lib/libc/tests/stdlib/tsearch_test.c:82:15: error: use of = undeclared identifier 'root' posix_tnode *root =3D NULL; ^ /usr/src/lib/libc/tests/stdlib/tsearch_test.c:89:30: error: use of = undeclared identifier 'root' ATF_CHECK(tdelete(&key, &root, compar) = !=3D NULL); etc This commit seems related, although I wonder if something else is going = on with the build? I would expect that make buildworld in /usr/src would use = /usr/src/include, instead of the system one, but this failure seems to = relate to the use of the system /usr/include/search.h (which doesn't = include the change above) instead of /usr/src/include/search.h, which = does. When the buildworld started, there was a comment to the effect that the = system clang looked to be the same as the source clang, so it would use = that instead. Is clang somehow hanging onto the system include = directories, when it shouldn't? After copying /usr/src/include/search.h to /usr/include and restarting = make buildworld, the lib/libc/tests/stdlib (all) passes OK... Cheers, --=20 Andrew Reilly
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?B5ED025A-9F83-43A9-A379-639A4C013225>