In libusb 1.0.21, UsbDk is enabled at compile time with --enable-usbdk
option. In libusb 1.0.22, the backend is selected at runtime.

This patch avoids a crash when spice client is compiled against libusb
1.0.22 and UsbDk is not installed.

* Enabling dynamic backend selection in Windows
https://github.com/libusb/libusb/commit/54884e84d024e761450287ab56aca

#74


Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

As it is not really used and outdated. We should use Gitlab's issues
tracker instead.

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>

The reason for this commit is that Meson expects all submodules to be
placed in this subdirectory, and since autotools build is more flexible
in this case, we make some small adjustments to configure.ac and
Makefile.am files to accommodate for this change.

The same has been done for spice-server in commit b0e141b3

Signed-off-by: Eduardo Lima (Etrunko) <etrunko@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>

Snir Sheriber authored 6 years ago and Frediano Ziglio committed 6 years ago

Acked-by: Frediano Ziglio <fziglio@redhat.com>

By default, all widgets are double buffered so this patch removes two
calls to gtk_widget_set_double_buffered() that are only enabling it.

The only special case is under egl (virgl) at set_egl_enabled() as
disabling the double buffered seems to prevent glitches. The call to
gtk_widget_set_double_buffered() was kept but wrapped under
G_GNUC_BEGIN/END_IGNORE_DEPRECATIONS to avoid warnings.

Note however that this is only expected to work under X11.

Warnings fixed:
 | spice-widget.c: In function ‘spice_display_init’:
 | spice-widget.c:643:5: warning: ‘gtk_widget_set_double_buffered’ is
 | deprecated
 |
 |     gtk_widget_set_double_buffered(area, true);
 |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 |
 | spice-widget.c:661:5: warning: ‘gtk_widget_set_double_buffered’ is
 | deprecated
 |
 |     gtk_widget_set_double_buffered(area, true);
 |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 |
 | spice-widget.c: In function ‘set_egl_enabled’:
 | spice-widget.c:1290:9: warning: ‘gtk_widget_set_double_buffered’ is
 | deprecated
 |     gtk_widget_set_double_buffered(GTK_WIDGET(area), !enabled);
 |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Some deprecation blocks in the code were there to avoid
warning when using too recent GTK and Glib functions.
Now that we bump GTK version remove these warning avoidance.

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>

As those were deprecated in favor of fetching for information directly
with GdkMonitor.

Warnings fixed:
 | spice-widget.c: In function ‘mouse_wrap’:
 | spice-widget.c:1136:5: warning: ‘gdk_screen_get_width’ is deprecated
 |
 |     xr = gdk_screen_get_width(screen) / 2;
 |     ^~
 |
 | spice-widget.c:1137:5: warning: ‘gdk_screen_get_height’ is deprecated
 |
 |     yr = gdk_screen_get_height(screen) / 2;
 |     ^~

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

As gdk_keymap_get_default() was deprecated in 3.22

Warnings fixed:
 | vncdisplaykeymap.c: In function ‘vnc_display_keyval_set_entries’:
 | vncdisplaykeymap.c:263:4: warning: ‘gdk_keymap_get_default’ is
 | deprecated: Use 'gdk_keymap_get_for_display' instead
 |
 |     gdk_keymap_get_entries_for_keyval(gdk_keymap_get_default(),
 |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 |
 | spice-gtk-session.c: In function ‘spice_gtk_session_init’:
 | spice-gtk-session.c:181:5: warning: ‘gdk_keymap_get_default’ is
 | deprecated: Use 'gdk_keymap_get_for_display' instead
 |
 |     GdkKeymap *keymap = gdk_keymap_get_default();
 |     ^~~~~~~~~
 |
 | spice-widget.c: In function ‘get_scancode_from_keyval’:
 | spice-widget.c:1731:5: warning: ‘gdk_keymap_get_default’ is
 | deprecated: Use 'gdk_keymap_get_for_display' instead
 |
 |      if (gdk_keymap_get_entries_for_keyval(gdk_keymap_get_default(),
 |      ^~

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Fedora 28: 3.22.30
Fedora 27: 3.22.26
Fedora 26: 3.22.21
RHEL 7.6: 3.22.30
RHEL 7.5: 3.22.26
RHEL 7.4: 3.22.10
Debian 10: 3.22.30
Debian  9: 3.22.11
Ubuntu 18.04: 3.22.30
Ubuntu 17.10: 3.22.24

Besides cleanup the code, gtk is under heavy development to better
work under Wayland and reducing the code path based on library version
should be helpful as well.

This patch also removes the check for gdk_event_get_scancode as it is
included in 3.22.

Signed-off-by: Victor Toso <victortoso@redhat.com>
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>

乐义华 authored 6 years ago and Frediano Ziglio committed 6 years ago

The image in cache was not being freed properly leading the client to
crash with OOM after sometime.

When image flag is SPICE_IMAGE_FLAGS_CACHE_REPLACE_ME, and if found it
in hash table, we will increase the refcount but this image will only
be destroyed, so the cache will usage more and more memory. In this
circumstance, we should not increase refcount.

Fixes: #72



Signed-off-by: 乐义华 <yueyihua@os-easy.com>
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Marc-André Lureau <marcandre.lureau@gmail.com>

Avoid the new error message for explicit celt051 option.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

No need to pass a generic pointer.
Avoid SpiceDisplay* -> gpointer* -> gpointer -> SpiceDisplay* cast
chain.

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

Lukáš Hrázký authored 6 years ago and Frediano Ziglio committed 6 years ago

Signed-off-by: Lukáš Hrázký <lhrazky@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Lukáš Hrázký authored 6 years ago and Frediano Ziglio committed 6 years ago

This brings in the following changes:

Christophe Fergeau (20):
      build: Update git.mk to latest version
      build: Remove docs/.gitignore
      build: Add __pycache__/*.pyc to DISTCLEANFILES
      log: Only install glib log handler if SPICE_DEBUG_LEVEL is set
      test-logging: Improve debug level tests
      test-region: Use GTest API
      test-region: Replace direct printf with g_debug() calls
      test-region: Don't call region_dump() by default
      test-region: Remove unneeded printf
      test-region: Replace spice_assert() with g_assert_true()
      test-region: Add g_assert() checks
      snd: Replace spice_printerr() use with g_warning
      swcanvas: Remove canvas_create()
      meson: Remove '(default: xxx)' from option description
      log: Remove SPICE_DISABLE_ABORT
      common: Remove spice_abort()
      build: Remove spice_common.h
      pixman: Use g_error() rather than g_abort()
      build: Define GLIB_VERSION_MIN_REQUIRED/GLIB_VERSION_MAX_ALLOWED
      quic: Wrap declaration/call of quic method in macros

Christophe de Dinechin (1):
      Add SPICE_ATTR_NORETURN on prototype of error functions

Eduardo Lima (Etrunko) (4):
      meson: Enable '-std=c99' build flag
      meson: Cleanup optional dependency checks
      Update gitlab-ci to use meson
      test-region: Decrease loop count by a factor of 10

Frediano Ziglio (36):
      test-quic: Allows to specify multiple images to test
      canvas_base: Rework DUMP_JPEG debugging
      test-quic: Increase test timeout using Meson
      ssl_verify: Fix build for newer LibreSSL
      Revert "ssl_verify: Fix build for newer LibreSSL"
      ssl_verify: Fix build for newer LibreSSL
      test-region: Create proper test for region from source code
      marshaller: Fix a possible leak
      canvas_base: Make sure top_down is a boolean
      marshal: Fix a bug with zero attribute
      ptypes: Improve some attribute documentation
      build: Reuse TEST_MARSHALLERS macro
      build: Reindent macros
      log: Remove spice_printerr macro
      test-marshallers: Check for "zero" attribute
      quic: Call encode from golomb_coding
      canvas_base: Fix minor indentation issues
      canvas_base: Avoid misaligned access decoding LZ4 data
      meson: Remove -std=c99
      quic: Add UPDATE_MODEL_COMP macro to iterate over channels
      quic: Rework PIXEL_A/PIXEL_B macros
      quic: Add SAME_PIXEL macro
      quic: Make {UN, }COMPRESS_xx macros closer
      quic: Add missing #undef SET_a/GET_a
      quic: Call directly encode_state_run from templates.
      quic: Add CORRELATE*/DECORRELATE* macros
      quic: Add APPLY_ALL_COMP macro to iterate over channels
      quic: Move all golomb decoding macros in a single place
      quic: Add DECLARE_*_VARIABLES macros
      quic: Introduce COPY_PIXEL macro
      quic: Reorder macro declarations
      quic: Unify rgb/non-rgb macro declarations
      quic: Remove duplicate file
      canvas_base: Check for overflows decoding LZ4
      canvas_base: Change spice_warning to g_warning
      tests: Join test-overflow and test-marshallers

Snir Sheriber (1):
      protocol: Add support for h265 video codec

Signed-off-by: Lukáš Hrázký <lhrazky@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Decoded frames are not necessarily dropped, update related debug
messages and comment

Signed-off-by: Snir Sheriber <ssheribe@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Snir Sheriber authored 6 years ago and Frediano Ziglio committed 6 years ago

When GstVideoOverly is used there is no appsink, playing of a frame is
done according the time value of its buffer

Acked-by: Frediano Ziglio <fziglio@redhat.com>

Snir Sheriber authored 6 years ago and Frediano Ziglio committed 6 years ago

Acked-by: Frediano Ziglio <fziglio@redhat.com>

display_queue is queued with decoded frames ready to be displayed.
However current code can insert a timeout before displaying and
removing the queued frames. As the frames are not compressed the
usage of memory by this queue can became in some cases quite
huge (in the order of multiple gigabytes).
To prevent this do not queue all the frames but just one and count the
pending frames.
This way GStreamer will stop queuing after a while.

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Snir Sheriber <ssheribe@redhat.com>

The remaining occurrences of spice_printerr() are warnings when
something unexpected happens, they can be replaced with g_warning() so
that users of spice-server can redirect them with
g_log_set_default_handler().

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

LibreSSL 2.7 hidden some definition, like OpenSSL 1.1.
This fixes https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=229044

.
Original patch from Piotr Kubaj.

Reported-by: Jack L. <xxjack12xx@gmail.com>
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

Old git (e.g. on rhel 7.5) is failing to init/update spice-common
submodule if .git is missing from relative url.

Signed-off-by: Snir Sheriber <ssheribe@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

This requires spice protocol v0.12.15

Signed-off-by: Snir Sheriber <ssheribe@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

With spice-common commit 72b0d603e12, SPICE_CHECK_CELT051 will error
out if celt051-devel is installed, but neither --enable-celt051 nor
--disable-celt051 are specified. This could be a problem when running
make distcheck, so this commit adds --disable-celt051 so that we never
hit that error.

Signed-off-by: Victor Toso <victortoso@redhat.com>

Christophe Fergeau (21):
      quic: Remove configurable RLE_PRED
      quic: Remove configurable PRED
      quic: Get rid of QUIC_RGB #define
      quic: Get rid of RLE_STAT #define
      quic: Get rid of RLE #define
      quic: Factor common code
      quic: Introduce CommonState *state variable in templates
      quic: s/decorrelate_drow/correlate_row
      quic: Add macros to make quic_tmpl.c much closer to quic_rgb_tmpl.c
      quic: Remove unused argument in uncompress_row{0, }
      quic: Use channel->correlate_row in macros
      quic: Add test case for compression/decompression
      build: Ensure we link with -lm if needed
      build: Disable celt 0.5.1 by default
      build: By default, error out if Opus is missing
      build: Use AM_COND_IF
      build: Remove checks for functions which are never called
      build: Remove bitops.h
      build: Move client sources to libspice_common_client_la_SOURCES
      meson: Support auto/true/false for optional dependencies
      meson: Remove check for vfork

Frediano Ziglio (21):
      protocol: Add some documentation for inval_all_pixmaps message
      Add --enable-extra-checks option
      codegen: Add some comments
      codegen: Removed unused get_type methods
      protocol: Use a typedef to specify stream_id type
      lz: Move ENCODE_PIXEL for RGB24 and RGB32 to a common place
      Fix integer overflows computing sizes
      Write a small test to test possible crash
      Avoid integer overflow computing image sizes
      Fix generation of Smartcard channel
      test-overflow: Remove a leak in the test
      marshaller: Remove initial underscore from static function
      codegen: Remove duplicate client and server code from ChannelType::resolve
      Check for messages with duplicate names inside a channel
      Check for messages with duplicate values inside a channel
      lz: Optimise SAME_PIXEL for RGB16
      lz: Inline GET_{r,g,b} macros
      quic: Remove 'no-inline' hack
      quic: Remove some too strict asserts in hot paths
      quic: Fix endianness encoding
      quic: Use __builtin_clz if available

Eduardo Lima (Etrunko) (8):
      build: Remove FIXME_SERVER_SMARTCARD hack
      Fix demarshaller code generator
      Fix field names for Smartcard protocol structures
      Fix cast to spice_marshaller_item_free_func function
      Bump glib requirements to 2.38
      test-quic: Fix -Wsign-compare warning
      Add support for building with meson/ninja
      Add .gitlab-ci.yml file

Jonathon Jongsma (2):
      Remove extra self parameter from member function
      miLineArc(): initialize edge1, edge2

Victor Toso (1):
      messages: document limitation of id in StreamCreate

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

By adding a guard to not handle channel-up on SpiceUsbredirChannel in
case struct usbredirhost wasn't initialized yet. Same guard is in
place for the generic usbredir_handle_msg() function to avoid handling
Server's message while Client's initialization is not done.

As mentioned in commit 291f3e44, this isn't a problem for
graphical clients as some initialization is done to present the
shareable usb devices to user.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1399838


Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

This reverts commit 291f3e44.

Follow up patch should address rhbz#1399838 mentioned in commit above
in a different way.

Major reason to revert is that the SpiceUsbDeviceManager object is
kept in SpiceSession as an easy way to share it between different
SpiceUsbredirChannel while SpiceSession itself does not use it. This
causes problems on migration as we start a new session with a new
SpiceUsbDeviceManager object while the previous one still exists.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1558043


Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>

G_ADD_PRIVATE was added in 2.38 and older functions are getting deprecated:
https://gitlab.gnome.org/GNOME/glib/merge_requests/7/commits



Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>

With touchpads and similar devices, scroll events are not emitted on
Wayland, but only smooth-scroll events.

There is some discussion about handling smooth-scroll events in Spice
(see "Add horizontal mouse wheel support" thread), but it will mean
support from various components. For compatibility reasons, in the
meantime, let's synthetize the smooth-scroll events to regular scroll
buttons to fix scrolling.

Tested on f28, gnome-shell under X and wayland.

Related to:
https://bugzilla.redhat.com/show_bug.cgi?id=1386602



Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

This is mostly needed form client side optimizations on streaming
mode. The following change introduced dependency of latest
spice-protocol:

 | commit 42439754
 | Author: Snir Sheriber <ssheribe@redhat.com>
 | Date:   Thu Mar 8 15:57:31 2018 +0200
 |
 | Gstreamer: Use GstVideoOverlay if possible

The spice-protocol commit for reference:

 | commit 4250216bafd37905afe0e9ad8cb9beb4a0c89254
 | Author: Frediano Ziglio <fziglio@redhat.com>
 | Date:   Tue Nov 7 10:45:20 2017 +0000
 |
 | Allow to specify a surface will be streamed
 |
 | This flag will allow the client to perform some optimisations
 | on output and buffering processing.
 | Old clients will ignore this additional flag.

Note that without bumping the requirement, the code would fail to
build with:

 | channel-display.c:1809:50: error: ‘SPICE_SURFACE_FLAGS_STREAMING_MODE’
 | undeclared (first use in this function); did you mean ‘SPICE_DISPLAY_STREAMING_MODE’?
 |      surface->streaming_mode = !!(create->flags & SPICE_SURFACE_FLAGS_STREAMING_MODE);
 |                                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>

Previously, an application was required to determine whether a session
had been disconnected by listening for the SpiceSession::channel-destroy
signal for each individual channel. The Application had to keep track of
whether all channels had been destroyed.

An additional difficulty is that when a SpiceSession::channel-destroy
signal is emitted, the channel object has not been freed yet, so any
shutdown/de-allocation actions have not yet happened. This became
significant since virt-viewer exits the application in response to the
last 'channel-destroy' signal, which means that the channel objects are
never properly freed. This is particularly problematic for the usbredir
channel, which might need to disconnect USB devices asynchronously, and
if the application exits before that process has completed, the USB
devices may not be available on the client machine.

With this patch, the SpiceSession still emits the 'channel-destroy'
signal for each channel when it is disconnected. But it also internally
tracks which channels have been destroyed, and waits for them to be
freed. When all channels are freed, it emits an additional
'disconnected' signal.  An application can be confident that all
channels have been freed by this point, and the application can exit
safely.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Marc-André Lureau <marcandre.lureau@gmail.com>

A GAsyncReadCallback function should not unref the GAsyncResult passed
to it as an argument since it does not hold a reference. This could
cause a double-free error (or a glib warning that it is not a valid
GObject)..

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Marc-André Lureau <marcandre.lureau@gmail.com>

With gstreamer audio backend, the following criticals are printed:
(virt-viewer:27631): GStreamer-CRITICAL **: gst_segment_to_running_time: assertion 'segment->format == format' failed

Because the default appsrc segment format is byte-based, although an
audio sink want a time-based segment.

Fixes:
#39



Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>

Tested with a ppc64 machine.

Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>