gnome-calendar:gnome-3-32

Last commit made on 2019-09-03
Get this branch:
git clone -b gnome-3-32 https://git.launchpad.net/gnome-calendar

Branch merges

Branch information

Name:
gnome-3-32
Repository:
lp:gnome-calendar

Recent commits

dbfa0f4... by Michael Catanzaro <email address hidden> on 2019-07-03

Fix crash in GcalSourceDialog on_source_added

This signal handler can run after the GcalSourceDialog has been
destroyed. We need to use g_signal_connect_object() to fix it.

Also, since this property is writable post-construction, we should
prepare for this to happen and disconnect in set_property() as well.

Probably fixes #418

8cb6448... by Michael Catanzaro <email address hidden> on 2019-07-03

GcalWindow should properly disconnect from its GcalManager

Fixes #392 for the gnome-3-32 branch

5639175... by Michael Catanzaro <email address hidden> on 2019-07-03

Fix crash in source_color_changed()

And opportunistically use g_signal_connect_object() when connecting to
the source object as well, because it's probably needed.

It looks like this is already fixed in a different way for master, so
this is only for gnome-3-32.

See !81 for details.

4326c33... by Michael Catanzaro <email address hidden> on 2019-07-03

Avoid crashing when year view's weather icon is NULL

This is not a proper solution, because I don't see how it could possibly
ever be NULL here. But a huge number of users are reporting this crash,
so we have to try something.

This fixes #299, but I'm not happy with it.

7849e8f... by Michael Catanzaro <email address hidden> on 2019-07-04

Fix buffer overflow in get_icon_name_sortkey()

We can just use strlen() here.

Fixes #425

8f80fdd... by Michael Catanzaro <email address hidden> on 2019-07-21

Don't leave dangling pointers in the GcalEvent cache

The cache has an unowned pointer to the GcalEvent's UID. We need to
update the cache with the GcalEvent's new UID to avoid use-after-free.

Fixes #434

1695bb3... by Michael Catanzaro <email address hidden> on 2019-07-21

Fix invalid ESource unref in on_client_connected()

This patch was written by Milan Crha.

The problem is that on_client_connected() expects the ESource being
references, but it's not. Similarly the self->clients expects the
ESource being inserted is referenced, but it's not. This causes too
early free of the ESource.

While reading the code, the expectation of on_client_connected() that
the source_object argument is an EClient descendant
(source = e_client_get_source (E_CLIENT (source_object));) is also
wrong, especially when this is called before
e_cal_client_connect_finish(). You should get the ESource from the
argument returned by the e_cal_client_connect_finish() and only if it's
not NULL. It can be NULL on error. That's yet another issue here,
client = E_CAL_CLIENT (e_cal_client_connect_finish (result, &error)); -
the typecast suggests that you expect always returned a non-NULL, but
it can be NULL on error, as said above.

Fixes #435

2068fab... by Michael Catanzaro <email address hidden> on 2019-07-02

Try to fix a crash in update_default_calendar_row()

This function is crashing in 3.32 because the manager object is invalid.
I think there are two related bugs:

First, it looks like gcal_quick_add_popover_set_property() is failing to
disconnect its signals from the old GcalManager before setting the new
one. In one backtrace, I see the GcalManager emitting the signal is
different than the GcalQuickAddPopover's current manager, which is
surely unintended.

But that shouldn't be enough to crash on its own, since the
GcalQuickAddPopover should still have a valid manager, even if not the
intended one. So I suspect the GcalQuickAddPopover itself is invalid at
this point. (The crash occured for me after adding an event, so it was
probably just destroyed.) GcalQuickAddPopover is not disconnecting these
signals in dispose, which is unsafe. We can avoid the need to do so by
using g_signal_connect_object().

Hopefully fixes #416 for the 3.32 branch.

https://bugzilla.redhat.com/show_bug.cgi?id=1509551

f9c1931... by Georges Basile Stavracas Neto on 2019-09-03

ci: Remove unnecessary CI stages

0a5a8e6... by Javier Jardón on 2019-07-31

flatpak: build intltool before e-d-s

Needed now that it is not part of the runtime/SDK.