> So, "This does not work with clang / LLVM" apparently"; looks
> like I'll have to put my hope to gcc. Or am I doing something
> wrong? Anyone want to make a prediction about how that'll turn
> out with the in-tree gcc 10.5.0?
Re-trying with in-tree gcc but with the -fsanitize=thread
argument added:
CCLD cfg_test
ld: /usr/pkgsrc/wip/bind920/work/bind-9.20.3/lib/isc/.libs/libisc.so: undefined
reference to `backtrace_symbols_fd'
% nm -o -p /usr/lib/libexecinfo.so | grep backtrace_symbols_fd
/usr/lib/libexecinfo.so:0000000000001647 T backtrace_symbols_fd_fmt
/usr/lib/libexecinfo.so:00000000000016c8 T backtrace_symbols_fd
%
Hmm... New attempt:
% git diff Makefile
diff --git a/bind920/Makefile b/bind920/Makefile
index d270b440ea..6f706070c5 100644
--- a/bind920/Makefile
+++ b/bind920/Makefile
@@ -29,6 +29,12 @@ GNU_CONFIGURE= yes
CHECK_FILES_SKIP= bin/tests/system/system-test-driver.sh
MAKE_ENV+= WRKDIR=${WRKDIR} PREFIX=${PREFIX}
+# Try to use thread sanitizer
+#PKGSRC_COMPILER= clang # does not work with thread sanitizer(!)
+CFLAGS+= -g
+CFLAGS+= -fsanitize=thread
+LDFLAGS+= -lexecinfo
+
.if ${OPSYS} == "Linux" && !exists(/usr/include/sys/capability.h)
CONFIGURE_ARGS+= --disable-linux-caps
.endif
%
and with that at least the build completed:
=> Creating /usr/pkgsrc/wip/bind920/work/.rc.d/named9
378.961u 78.529s 7:36.63 100.1% 509+78k 3+38245io 209pf+0w
Not yet tested the result, though. That's for ... probably
tomorrow.
Regards,
- Håvard
