Hi Seyeong,
first, I commend you for going to the trouble of fixing tests in
lp1316970_trusty_glib2.0.debdiff, but I think those are unrelated to
this actual bug and so the patch is not required for this. I'll focus
only on the first patch.
In that patch, I did a quick review of the changes, and I'm not entirely
sure they are correct? LP bug comments are not a great format for
reviewing patches, so please excuse formatting/line-wrapping:
> --- a/lib/services/upstart.c
> +++ b/lib/services/upstart.c
> @@ -70,6 +70,7 @@
> if (error) {
> crm_err("Can't connect obtain proxy to %s interface: %s", interface,
> error->message);
> g_error_free(error);
> + g_object_unref(proxy);
this doesn't seem right - if error is non-NULL, then proxy should be NULL,
according to the docs:
https://developer.gnome.org/gio/stable/GDBusProxy.html#g-dbus-proxy-new-for-bus-sync
> proxy = NULL;
> }
> return proxy;
> @@ -107,7 +108,9 @@
> /*
> com.ubuntu.Upstart0_6.GetJobByName (in String name, out ObjectPath job)
> */
> - GVariant *_ret = g_dbus_proxy_call_sync(proxy, "GetJobByName",
> g_variant_new("(s)", arg_name),
> + GVariant *_ret = NULL;
> +
> + _ret = g_dbus_proxy_call_sync(proxy, "GetJobByName",
> g_variant_new("(s)", arg_name),
> G_DBUS_CALL_FLAGS_NONE, -1,
> cancellable, error);
unless i'm missing something, there is no need for this change?
>
> if (_ret) {
> @@ -200,6 +203,7 @@
>
> g_variant_iter_free(iter);
> g_variant_unref(_ret);
> + free(path);
this does not seem correct - per the docs, 'path' will be freed by each call in
the while loop to g_variant_iter_loop(), and only needs to be manually freed
after the while loop if the code breaks out of the loop manually. That doesn't
appear to be the case, so path should not need freeing here.
https://developer.gnome.org/glib/stable/glib-GVariant.html#g-variant-iter-loop
> return units;
> }
>
> @@ -224,7 +228,7 @@
> } else if (pass) {
> crm_trace("Got %s", path);
> }
> - /* free(path) */
> + free(path);
technically this should be freed only if !error - the freeing should go
into the else if (pass) block above, although free(NULL) will just do
nothing
however, i believe this should use g_free() instead of free()
> return pass;
> }
>
> @@ -272,6 +276,8 @@
>
> g_object_unref(proxy);
> g_variant_unref(_ret);
> + g_variant_unref(value);
> + g_variant_unref(asv);
asv is returned from g_variant_get_child_value, which does state in its docs
that the returned value should be unref'ed, so this looks correct for 'asv'.
https://developer.gnome.org/glib/stable/glib-GVariant.html#g-variant-get-child-value
However value is returned from g_variant_lookup_value, which returns a value
from a key-value pair in asv (as best i can tell from the docs for the
function), and it does not indicate that the value is newly allocated or that
it should be freed. So it does not appear that 'value' should be unref'ed here.
https://developer.gnome.org/glib/stable/glib-GVariant.html#g-variant-lookup-value
> return output;
> }
>
> @@ -299,11 +305,14 @@
> GVariant *tmp2 = g_variant_get_child_value(tmp1, 0);
>
> instance = g_variant_dup_string(tmp2, NULL);
> + g_variant_unref(tmp2);
> }
> + g_variant_unref(tmp1);
these look correct
> }
>
> crm_info("Result: %s", instance);
> g_variant_unref(_ret);
> + g_object_unref(proxy);
this looks correct
> return instance;
> }
>
> @@ -338,6 +347,7 @@
> }
>
> crm_info("%s is%s running", name, pass ? "" : " not");
> + free(job);
technically this is only needed in the else {} block, but as it should
still be NULL the free will do nothing
again this should be g_free() instead of free() i think
> return pass;
> }
>
> @@ -400,6 +410,7 @@
> crm_info("Call to %s passed: type '%s' %s", op->action,
> g_variant_get_type_string(_ret),
> path);
> op->rc = PCMK_EXECRA_OK;
> + free(path);
looks good, but probably use g_free()
>
> } else {
> crm_err("Call to %s passed but return type was '%s' not '(o)'",
> op->action, g_variant_get_type_string(_ret));
> @@ -501,6 +512,7 @@
> crm_info("Call to %s passed: type '%s' %s", op->action,
> g_variant_get_type_string(_ret),
> path);
> op->rc = PCMK_EXECRA_OK;
> + free(path);
looks good, but probably use g_free()
>
> } else {
> crm_err("Call to %s passed but return type was '%s' not '(o)'",
> op->action, g_variant_get_type_string(_ret));
--
You received this bug notification because you are a member of Ubuntu
Desktop Bugs, which is subscribed to glib2.0 in Ubuntu.
https://bugs.launchpad.net/bugs/1316970
Title:
g_dbus memory leak in lrmd
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/glib2.0/+bug/1316970/+subscriptions
--
desktop-bugs mailing list
desktop-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/desktop-bugs
