On Sat, Jan 09, 2016 at 05:32:46PM -0500, Cole Robinson wrote:
On 01/09/2016 12:36 PM, Martin Kletzander wrote:
> For some reason we are not setting the driver with memset() to zeros.
> But since commit 74abc3deac6e14ffa9151e425c6e6cd2b075aac5
> driver->securityManager is being accessed and qemuagenttest started
> crashing due to that.
>
> Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com>
> ---
> I chose to just clear the pointer instead of clearing the whole driver
> with memset() and opening another can of worms. But I would rather do
> memset() to zeros. however, that can be done after a discussion.
> This needs to be pushed under the build-breaker rule... Done!
>
> tests/testutilsqemu.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/tests/testutilsqemu.c b/tests/testutilsqemu.c
> index 8a4f567b0b2a..f2eacdded53d 100644
> --- a/tests/testutilsqemu.c
> +++ b/tests/testutilsqemu.c
> @@ -558,6 +558,8 @@ int qemuTestDriverInit(virQEMUDriver *driver)
> if (virMutexInit(&driver->lock) < 0)
> return -1;
>
> + driver->securityManager = NULL;
> +
> driver->config = virQEMUDriverConfigNew(false);
> if (!driver->config)
> goto error;
Hmm, I wonder why I wasn't hitting this...
I don't think it's because I am building without SELinux, but it will
probably be one of the compiler options. It shouldn't be caused just by
using gcc 5.3.0, but rather one of:
lv_cv_static_analysis=no CFLAGS='-O0 -ggdb -fno-profile-arcs
-std=gnu11 -ftest-coverage'
or maybe just a coincidence =)
Martin