GitLab

Fixes #62

task-cache: Prevent reference to a freed task cache

See merge request !58

As the CancelledData doesn't hold a reference to a DzlTaskCache, it is possible
for the GTask holding it to actually finishing after the DzlTaskCache has been
finalized, do not unnecessarily store the DzlTaskCache reference but what we
actually need from it.

The page size needs to be 0 or spin button brings things down from the
value selected if its > upper-page_size.

css-provider: Apply changes immediately

See merge request !57

While queueing may save updates if multiple things are being changed
at the same time, those occasions are fairly rare, and scheduling the
update causes it to happen one frame later than needed. If the app
takes a while to redraw (e.g. Epiphany), potentially a lot later.

See epiphany#1607

Fix link errors with GCC7 and older macOS

See merge request !56

dzl-dock-widget: Don't set can-focus to True

See merge request !55

Currently when moving the focus with the keyboard (tab, arrow keys), the
focus always moves in the Tab direction, without ever following what
focus move direction was requested.

This is because the DzlDockWidget tells it supports receiving the
keyboard focus, which is not the case: it is not e.g. a button or entry,
instead it's more a dock container than anything else.

So this commit fixes it by dropping the line that set the can-focus
property to True.

Followup to !54

Not really all that helpful, and just causes signed warnings on some
platforms.

Fixes #58

Fix build on Windows

See merge request !54

Doing so messes up the installation on Windows, where the dlls go into
the bindir, where the import libraries go into the libdir

These flags break the build on Windows, where __GNUC__ is not defined and
checked for in the glib headers

build: set PLATFORM_OSX on macOS

See merge request !53

Since this will add some objective-c code to the build, compilation must
then occur with the -xobjective-c flag.

build: fix build on Solaris 10 with GCC 8.4

See merge request !52

munmap on Solaris 10, depending on compile time defines can use caddr_t
instead of void* as its argument failing the build with the following
error message:

error: passing argument 1 of 'munmap' from incompatible pointer type

cast to void*, should be safe.

Include counter headers not for Win32.

See merge request !51

tests: Don't use g_assert

See merge request !50

The g_assert macro can be turned off by defining G_DISABLE_ASSERT.
Therefore, it's not meant to be used in unit tests because the tests
will be ineffective if compiled with G_DISABLE_ASSERT.

In addition to that, g_assert_true and friends emit a more human
readable error message in case the condition is not satisfied.

!50