On 04.10.2019 2:34, William Tu wrote:
On Thu, Oct 3, 2019 at 10:15 AM Ilya Maximets <[email protected]> wrote:
On 03.10.2019 18:13, Ilya Maximets wrote:
On 02.10.2019 20:15, William Tu wrote:
32-bit and 64-bit libunwind can not be installed at the same time.
For 32-bit build, this patch removes the 64-bit libunwind and install
32-bit version.
Tested-at: https://travis-ci.org/williamtu/ovs-travis/builds/592649311
Fixes: e2ed6fbeb18c ("fatal-signal: Catch SIGSEGV and print backtrace.")
Signed-off-by: William Tu <[email protected]>
---
.travis/linux-prepare.sh | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/.travis/linux-prepare.sh b/.travis/linux-prepare.sh
index 70fd98f715ed..164adf7ec4f8 100755
--- a/.travis/linux-prepare.sh
+++ b/.travis/linux-prepare.sh
@@ -14,3 +14,9 @@ cd ..
pip install --disable-pip-version-check --user six flake8 hacking
pip install --user --upgrade docutils
+
+if [[ $BUILD_ENV =~ "-m32" ]]; then
+ # 32-bit and 64-bit libunwind can not be installed at the same time.
+ # This will remove the 64-bit libunwind and install 32-bit version.
+ sudo apt-get install -y libunwind-dev:i386
+fi
Thanks for the new version.
One thing I noticed is that without additional changes[1] this
patch just makes 'configure' script to fail the library check,
because 32bit library can't be linked with 64bit binary it checks:
checking for unw_backtrace in -lunwind... no
But it should work as intended with the patch I posted:
[1] https://patchwork.ozlabs.org/patch/1171259/
Note that I replaced $BUILD_ENV with $M32 variable, so one of the
patches will need to be updated depending on the order of applying.
i.e. following diff should be squashed to one of them:
-if [[ $BUILD_ENV =~ "-m32" ]]; then
+if [ "$M32" ]; then
Other than that,
Acked-by: Ilya Maximets <[email protected]>
Hmm, I tried to apply this patch and check the build and it
failed:
lib/backtrace.c: In function ‘log_received_backtrace’:
lib/backtrace.c:105:23: error: format ‘%lx’ expects argument of type ‘long
unsigned int’, but argument 4 has type ‘unw_word_t {aka unsigned int}’
[-Werror=format=]
VLOG_WARN("0x%016lx <%s+0x%lx>\n",
^
./include/openvswitch/vlog.h:277:41: note: in definition of macro ‘VLOG’
vlog(&this_module, level__, __VA_ARGS__); \
lib/backtrace.c:105:13: note: in expansion of macro ‘VLOG_WARN’
VLOG_WARN("0x%016lx <%s+0x%lx>\n",
^
lib/backtrace.c:105:23: error: format ‘%lx’ expects argument of type ‘long
unsigned int’, but argument 6 has type ‘unw_word_t {aka unsigned int}’
[-Werror=format=]
VLOG_WARN("0x%016lx <%s+0x%lx>\n",
^
./include/openvswitch/vlog.h:277:41: note: in definition of macro ‘VLOG’
vlog(&this_module, level__, __VA_ARGS__); \
lib/backtrace.c:105:13: note: in expansion of macro ‘VLOG_WARN’
VLOG_WARN("0x%016lx <%s+0x%lx>\n",
^
Full log:
https://travis-ci.org/igsilya/ovs/jobs/593132451
Thanks!
For 32-bit libunwind, the unw_word_t is defined as uint32_t, and
for 64-bit libunwind, it is defined as uint64_t. Here actually it is printing
a pointer, so I will change it to use PRIxPTR
diff --git a/lib/backtrace.c b/lib/backtrace.c
index 9347634487c8..2853d5ff150d 100644
--- a/lib/backtrace.c
+++ b/lib/backtrace.c
@@ -102,7 +102,7 @@ log_received_backtrace(int fd) {
if (backtrace[i].func[0] == 0) {
break;
}
- VLOG_WARN("0x%016lx <%s+0x%lx>\n",
+ VLOG_WARN("0x%016"PRIxPTR" <%s+0x%"PRIxPTR">\n",
backtrace[i].ip,
backtrace[i].func,
backtrace[i].offset);
Above fix looks good to me. Now when OVS_CFLAGS fix is merged, you
could send above two changes (for travis and for backtrace.c) based
on current master branch. These could be 2 separate patches or a
single squashed change.
Best regards, Ilya Maximets.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
