[Spice-devel] [spice-gtk v1] gstaudio: fix ci due potential null pointer dereference
Frediano Ziglio
fziglio at redhat.com
Tue Jan 22 15:19:59 UTC 2019
>
> Hi,
>
> On Tue, Jan 22, 2019 at 08:01:33AM -0500, Frediano Ziglio wrote:
> > >
> > > From: Victor Toso <me at victortoso.com>
> > >
> > > Can't happen as GstElement *e is not NULL but doing extra G_OBJECT(e)
> > > cast
> > > guarantees this for compiler.
> > >
> > > > spice-gstaudio.c: In function ‘playback_volume_changed’:
> > > > /usr/include/glib-2.0/gobject/gtype.h:2280:70: error: potential null
> > > > pointer dereference [-Werror=null-dereference]
> > > > #define _G_TYPE_IGC(ip, gt, ct) ((ct*) (((GTypeInstance*)
> > > > ip)->g_class))
> > > > ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
> > > > /usr/include/glib-2.0/gobject/gtype.h:525:66: note: in expansion of
> > > > macro
> > > > ‘_G_TYPE_IGC’
> > > > #define G_TYPE_INSTANCE_GET_CLASS(instance, g_type, c_type)
> > > > (_G_TYPE_IGC ((instance), (g_type), c_type))
> > > > ^~~~~~~~~~~
> > > > /usr/include/glib-2.0/gobject/gobject.h:86:38: note: in expansion of
> > > > macro
> > > > ‘G_TYPE_INSTANCE_GET_CLASS’
> > > > #define G_OBJECT_GET_CLASS(object) (G_TYPE_INSTANCE_GET_CLASS
> > > > ((object),
> > > > G_TYPE_OBJECT, GObjectClass))
> > > > ^~~~~~~~~~~~~~~~~~~~~~~~~
> > > > spice-gstaudio.c:373:45: note: in expansion of macro
> > > > ‘G_OBJECT_GET_CLASS’
> > > > } else if (g_object_class_find_property(G_OBJECT_GET_CLASS (e),
> > > > "volume") != NULL) {
> > > ^~~~~~~~~~~~~~~~~~
> > >
> > > Signed-off-by: Victor Toso <victortoso at redhat.com>
> >
> > Why not adding g_return_if_fail(e != NULL) before using "e" instead ?
>
> Why that would be better than a proper G_OBJECT()?
>
It looks less weird and checking for NULL costs less.
G_OBJECT is a bit weird. GObject cast macros check for NULL and if the type
if the type specified assuming is a GObject pointer, however in this case
(beside the NULL check) is checking that is a GObject pointer assuming is
a GObject pointer so is a tautology. For the compiler is just a trick
to make it happy.
The g_return_if_fail is cheaper as the G_OBJECT macro will call a function
in any case while g_return_if_fail will call a function (g_warning) only
if "e" is NULL.
> We already check if playback.sink or record.src are null early on
> in those functions, adding g_return_if_fail() would only check if
> the assignment was done. I don't think it is needed.
>
> Cheers,
>
> > > ---
> > >
> > > Working CI:
> > > https://gitlab.freedesktop.org/victortoso/spice-gtk/pipelines/15885
> > >
> > > src/spice-gstaudio.c | 8 ++++----
> > > 1 file changed, 4 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/src/spice-gstaudio.c b/src/spice-gstaudio.c
> > > index 51ff028..b2d9cd3 100644
> > > --- a/src/spice-gstaudio.c
> > > +++ b/src/spice-gstaudio.c
> > > @@ -370,7 +370,7 @@ static void playback_volume_changed(GObject *object,
> > > GParamSpec *pspec, gpointer
> > >
> > > if (GST_IS_STREAM_VOLUME(e)) {
> > > gst_stream_volume_set_volume(GST_STREAM_VOLUME(e),
> > > GST_STREAM_VOLUME_FORMAT_CUBIC, vol);
> > > - } else if (g_object_class_find_property(G_OBJECT_GET_CLASS (e),
> > > "volume") != NULL) {
> > > + } else if
> > > (g_object_class_find_property(G_OBJECT_GET_CLASS(G_OBJECT(e)),
> > > "volume") != NULL) {
> > > g_object_set(e, "volume", vol, NULL);
> > > } else {
> > > g_warning("playback: ignoring volume change on %s",
> > > gst_element_get_name(e));
> > > @@ -399,7 +399,7 @@ static void playback_mute_changed(GObject *object,
> > > GParamSpec *pspec, gpointer d
> > >
> > > if (GST_IS_STREAM_VOLUME(e)) {
> > > gst_stream_volume_set_mute(GST_STREAM_VOLUME(e), mute);
> > > - } else if (g_object_class_find_property(G_OBJECT_GET_CLASS (e),
> > > "mute")
> > > != NULL) {
> > > + } else if
> > > (g_object_class_find_property(G_OBJECT_GET_CLASS(G_OBJECT(e)),
> > > "mute") != NULL) {
> > > g_object_set(e, "mute", mute, NULL);
> > > } else {
> > > g_warning("playback: ignoring mute change on %s",
> > > gst_element_get_name(e));
> > > @@ -437,7 +437,7 @@ static void record_volume_changed(GObject *object,
> > > GParamSpec *pspec, gpointer d
> > >
> > > if (GST_IS_STREAM_VOLUME(e)) {
> > > gst_stream_volume_set_volume(GST_STREAM_VOLUME(e),
> > > GST_STREAM_VOLUME_FORMAT_CUBIC, vol);
> > > - } else if (g_object_class_find_property(G_OBJECT_GET_CLASS (e),
> > > "volume") != NULL) {
> > > + } else if
> > > (g_object_class_find_property(G_OBJECT_GET_CLASS(G_OBJECT(e)),
> > > "volume") != NULL) {
> > > g_object_set(e, "volume", vol, NULL);
> > > } else {
> > > g_warning("record: ignoring volume change on %s",
> > > gst_element_get_name(e));
> > > @@ -466,7 +466,7 @@ static void record_mute_changed(GObject *object,
> > > GParamSpec *pspec, gpointer dat
> > >
> > > if (GST_IS_STREAM_VOLUME (e)) {
> > > gst_stream_volume_set_mute(GST_STREAM_VOLUME(e), mute);
> > > - } else if (g_object_class_find_property(G_OBJECT_GET_CLASS (e),
> > > "mute")
> > > != NULL) {
> > > + } else if
> > > (g_object_class_find_property(G_OBJECT_GET_CLASS(G_OBJECT(e)),
> > > "mute") != NULL) {
> > > g_object_set(e, "mute", mute, NULL);
> > > } else {
> > > g_warning("record: ignoring mute change on %s",
> > > gst_element_get_name(e));
> >
> > Frediano
>
More information about the Spice-devel
mailing list