summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
* display: fix palette regressionMarc-André Lureau2013-10-031-3/+3
| | | | | | | | | | | | palette_get() used to return a ref, and palette_release() used to release that ref. Since ed877341, the palette is no longer refcount'ed, since its usage is exclusively local in common/canvas code. palette_release() shouldn't remove the palette from the cache. https://bugzilla.redhat.com/show_bug.cgi?id=1011936
* mono cursor: increase contrast, better looking putty cursor (rhbz 998529)Alon Levy2013-10-021-8/+8
| | | | Signed-off-by: Alon Levy <alevy@redhat.com>
* spice-channel: Fix usbredir being broken since commit 159c6ebfHans de Goede2013-09-301-0/+1
| | | | | | | | | | The usbredir channel uses spice_msg_in_raw to get its data, which uses in->dpos to determine the msg size and that was no longer being set for non sub-messages. https://bugs.freedesktop.org/show_bug.cgi?id=69935 Signed-off-by: Hans de Goede <hdegoede@redhat.com>
* win32: fix alt-tab grab regressionMarc-André Lureau2013-09-261-1/+1
| | | | | | | Since 5f67178c, alt-tab is no longer grabbed by the client. The keyboard hook still needs to handle WM_SYSKEY{DOWN,UP} messages. https://bugzilla.redhat.com/show_bug.cgi?id=873341
* build-sys: do not compile python binding with windows by defaultMarc-André Lureau2013-09-261-14/+10
|
* Update spice-commonMarc-André Lureau2013-09-261-0/+0
|
* build-sys: add --with-audio=autoMarc-André Lureau2013-09-261-2/+6
| | | | Defaults to GStreamer on Windows, and PulseAudio otherwise atm.
* build-sys: fix --with-saslMarc-André Lureau2013-09-261-11/+5
| | | | | The default configure sasl auto setting will error out if SASL is missing. Instead, silentely skip sasl configure if detected missing.
* build-sys: fix g_clear_pointer usage on old glibMarc-André Lureau2013-09-201-0/+1
| | | | | As pointed out and verified on the ML after 0.21 release by Klaus Hochlehnert.
* Release v0.21v0.21Marc-André Lureau2013-09-181-0/+11
|
* acl helper: Use ruid of invoker rather than looking up euid in /procColin Walters2013-09-181-1/+2
| | | | | | | | This way we avoid a race condition if the parent execve()s a setuid program (possibly this program). This is the same as the fix for pkexec which is CVE-2011-1485: See: https://bugzilla.redhat.com/show_bug.cgi?id=692922
* build-sys: fix bad cast on win64Marc-André Lureau2013-09-181-1/+1
| | | | | spice-widget.c:814:50: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast] g_return_if_fail(SystemParametersInfo(SPI_SETMOUSESPEED, 0, (PVOID)d->win_mouse_speed, 0));
* display: unshare the palette cacheMarc-André Lureau2013-09-133-8/+3
| | | | | | | As pointed out by Yonit on the ML: > (1) the palette cache shouldn't be shared among the display channels. I.e., > there should be one instance per channel, and not one instance in > spice-session.
* channel: unref incoming migrate dataMarc-André Lureau2013-09-131-3/+6
|
* gtk: simplify spice_channel_recv_msgMarc-André Lureau2013-09-133-37/+17
| | | | | | Use of coroutines allow to simplify spice_channel_recv_msg(), it doesn't need to keep current reading state, it can rely on the coroutine stack for that.
* gtk: use slices for frequently allocated objectsMarc-André Lureau2013-09-132-6/+6
|
* gtk: use GHashTable in display_cacheMarc-André Lureau2013-09-135-220/+102
| | | | | The cache code isn't very quick, it shows up in profilers. Using GHashTable allows to simplify the code while making it faster.
* channel: do not reenter the mainloop at every iterationMarc-André Lureau2013-09-131-31/+28
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The current coroutine channel_iterate() creates a GSource for the socket reenters the mainloop waiting for IO condition. This is really heavy usage of mainloop showing up in system time and user space profiling (10% of CPU time spent in mainloop overhead). Instead flush all pending input messages before switching context and reentering main loop. This is the kind of results I get with a replay: before: 2179,440455 task-clock # 0,629 CPUs utilized 3 580 context-switches # 0,002 M/sec 207 cpu-migrations # 0,095 K/sec 48 909 page-faults # 0,022 M/sec 7 362 442 301 cycles # 3,378 GHz 5 256 957 520 stalled-cycles-frontend # 71,40% frontend cycles idle <not supported> stalled-cycles-backend 5 460 266 981 instructions # 0,74 insns per cycle # 0,96 stalled cycles per insn 982 749 523 branches # 450,918 M/sec 20 745 453 branch-misses # 2,11% of all branches 3,463422087 seconds time elapsed after: 1741,651383 task-clock # 0,522 CPUs utilized 5 093 context-switches # 0,003 M/sec 137 cpu-migrations # 0,079 K/sec 49 033 page-faults # 0,028 M/sec 5 874 567 974 cycles # 3,373 GHz 4 247 059 288 stalled-cycles-frontend # 72,30% frontend cycles idle <not supported> stalled-cycles-backend 4 419 666 346 instructions # 0,75 insns per cycle # 0,96 stalled cycles per insn 776 972 366 branches # 446,112 M/sec 17 862 170 branch-misses # 2,30% of all branches 3,337472053 seconds time elapsed
* util: avoid calling getenv for every SPICE_DEBUGMarc-André Lureau2013-09-131-3/+10
| | | | | That doesn't seem to really improve performance so much, but that' s what we should do probably.
* display: keep a reference to the primary surfaceMarc-André Lureau2013-09-131-13/+15
| | | | | | | | | | | | | | | | | Avoid hard-coding surface 0 as being primary, although in practice it always is so far. Also a lot of lookups are primary, so add a shortcut for this special case (~30% apparently), it shows some small lookup speedup. before: real 0m5.008s user 0m3.253s sys 0m2.015s after: real 0m4.930s user 0m3.133s sys 0m2.027s
* display: make the hashtable to destroy the surfaceMarc-André Lureau2013-09-131-7/+9
| | | | Improve a bit the code by using hashtable ownership.
* display: replace ring with hashtableMarc-André Lureau2013-09-132-22/+12
| | | | | | | | | | | | | | | | | | With a Spice replay (a tool not yet merged, but available in dev branches), the following commit improves a little bit performance by not spending so much CPU time in looking up surfaces. I found initially hotspot with "perf", and get a consistant ~200ms speedup with "time spicy-stats" after replacing the ring. before: real 0m5.147s user 0m3.506s sys 0m1.949s after: real 0m5.008s user 0m3.253s sys 0m2.015s
* channel: add SPICE_DISABLE_CHANNELSMarc-André Lureau2013-09-132-0/+9
| | | | | Allow to disable selectively channels, mainly used for testing, ex: SPICE_DISABLE_CHANNELS=display spicy-stats -p 12345
* channels: use spice_channel_set_handlers()Marc-André Lureau2013-09-1311-276/+155
| | | | | | This allows to simplify a little bit derived class (no need to override handle_msg), and allows the base class more flexibility (for example for filtering messages, as in the following patch)
* channel: add spice_channel_set_handlers()Marc-André Lureau2013-09-133-1/+43
| | | | | | This function will allow to set base handlers and specific channel handlers in a common way, instead of each channel having to override the base channel virtual handle_msg().
* display: use bitfields for surface flagsMarc-André Lureau2013-09-131-1/+2
| | | | | | Checking by value make the flag fields useless. Unfortunately, when adding more flags, the server will have to ensure it can safely send it, by checking some of new client caps (for some features).
* Update spice-commonMarc-André Lureau2013-09-131-0/+0
|
* Fix _FORTIFY_SOURCE redefine errorDunrong Huang2013-09-101-1/+1
| | | | | | | | | | | | If the _FORTIFY_SOURCE has been already defined, we shouldn't redefine it, or it will raise a build error as below: In file included from spice-audio.c:36:0: ../config.h:12:0: error: "_FORTIFY_SOURCE" redefined [-Werror] spice-audio.c:1:0: note: this is the location of the previous definition Suggested-by: Christophe Fergeau <cfergeau@redhat.com> Signed-off-by: Dunrong Huang <riegamaths@gmail.com>
* Revert "tmp: debug MIGRATE_DATA"Marc-André Lureau2013-08-281-15/+0
| | | | This reverts commit f670e0197c37b2f12518f6d216642f67cdf5de84.
* gtk/channel-cursor: copy spicec hack, RHBZ #998529Alon Levy2013-08-281-5/+18
| | | | | | | flip -> unsupported by x11, since XCreatePixmapCursor has no invert functionality, only a mask, shape, background and foreground colors. Use this checkerboard hack to get some contrast for cursors in the guest that relied on invert for the same contrast.
* gtk/channel-cursor.c: add cursor_type_to_string for debuggingAlon Levy2013-08-281-2/+20
|
* gtk/channel-display: only use SysV shm for x11Alon Levy2013-08-281-1/+1
|
* tmp: debug MIGRATE_DATAAlon Levy2013-08-281-0/+15
|
* Remove incorrect option context from spicy-statsJonathon Jongsma2013-08-281-1/+1
| | | | | | | The argument given to g_option_context_new() was apparently copy/pasted from spicy-screenshot and therefore inaccurate for spicy-stats. Since the actual description of the program (from _set_summary()) is displayed on the next line anyway, simply make the context NULL.
* channel-main: Fix minor documentation issueJonathon Jongsma2013-08-281-1/+1
| | | | | Documentation for spice_main_clipboard_selection_request() refers to deprecated signal 'main-clipboard' instead of 'main-clipboard-selection'
* tests: add some dos2unix testsMarc-André Lureau2013-08-245-1/+110
| | | | This is probably not exhaustive enough, but better than nothing.
* gtk: handle clipboard CRLF conversionMarc-André Lureau2013-08-242-5/+53
| | | | | | | | | | | | | | gtk+ internal text/utf8 is using LF conversion, on all platforms. Even though the toolkit may only handle a single line ending type, we may want to avoid the conversion for Spice use cases, gtk+ could learn a new native utf8 target type, see also: https://bugzilla.gnome.org/show_bug.cgi?id=706657 In the meantime, the only thing we need to convert, is to/from crlf guest (from/to lf). This is what this change is about. It has been tested successfully with the various guest/client OS combinations.
* util: add unix2dos and dos2unixMarc-André Lureau2013-08-242-0/+104
| | | | Convert line endings from/to LF/CRLF.
* Revert "channel-main: Convert text line-endings if necessary (rhbz#752350)"Marc-André Lureau2013-08-231-69/+4
| | | | | | | | | | | | | | | | This implementation assumes that the toolkit doesn't do any conversion on its own. This is actually not the case, gtk+ converts line endings already on windows. It would be pretty ugly to do conversions back and forth at different levels. So the channel implementation shouldn't try to do conversion, and implementation must be done at higher level, where the actual system clipboard implementation is known (at gtk-session for instance) Since this code hasn't been officially released, let's revert it and implement a better crlf conversion handling in gtk-session in the following commits. This reverts commit e45a446a9981ad4adaeff9c885962a8c6140333e.
* gtk: remove dead codeMarc-André Lureau2013-08-231-2/+0
|
* win-usb-dev: ignore devices with addr 1 for LIBUSBX_API_VERSIONUri Lublin2013-08-201-0/+3
| | | | | | | | | Starting libusbx version 1.0.13 (commit 8cd30bb7066f785ee78cf6c3dccafdbc4b957b50) windows device enumeration changed and root hubs address number is 1 intead of 0xff. This patch uses LIBUSBX_API_VERSION which was introduced after 1.0.13 release (commit 9d368fc4774344d81ab02840f3a8478301bfb6fa).
* win usb: request for driver uninstall only for driver that were installedUri Lublin2013-08-201-1/+13
| | | | | | | | | | | | | | | | | | | | | | Currently when spice_usb_device_manager_remove_dev() is called, for windows clients a request to uninstall the driver is sent to usbclerk. This cause problems when 2 instances (A and B) of the client are running: - Both A and B get notified about a new USB device when it's plugged in - A is requested to usbredir the USB device. - A requests usbclerk to install WinUSB driver for that device. - usbclerk starts installing the driver - Windows sends to both A and B - device removal events, which may cause B to call spice_usb_device_manager_remove_dev. - usbclerk completes installing the driver and reply to A - B requests usbclerk to remove the WinUSB driver for that device. To prevent that, spice-gtk now requests usbclerk to remove the driver only if it was the instance that requested the driver to be installed. This is done with the help of a new device SPICE_USB_DEVICE_STATE_INSTALLED state. rhbz#919166
* Use g_slist_foreach_full glib implementationChristophe Fergeau2013-08-191-11/+4
| | | | | For the compat case (glib < 2.28), we were using our own implementation instead of directly reusing glib code.
* glib-compat: g_slist_free_full: pass the right ptr to destroy (rhbz#997893)Hans de Goede2013-08-191-1/+1
| | | | | | | The destroy function passed to g_slist_free_full should be passed the elements data pointer, not the element itself. Signed-off-by: Hans de Goede <hdegoede@redhat.com>
* Only use GOBJECT_INTROSPECTION_CHECK when availableChristophe Fergeau2013-08-131-5/+7
| | | | | | | | | | | | | | Current el6 systems don't have gobject-introspection packages, so the GOBJECT_INTROSPECTION_CHECK m4 macro is not available on such systems. This makes it difficult to run autogen.sh on these systems. This commit uses m4_ifdef to check if this macro is available before trying to use it, and makes sure introspection is disabled when GOBJECT_INTROSPECTION_CHECK could not be found. This allows to compile spice-gtk from git on el6 systems with: ./autogen.sh --with-gtk=2.0 --disable-gtk-doc --disable-vala --disable-controller --disable-werror
* usb-device-manager: fix compilation error when usbredir is disabledYonit Halperin2013-08-061-1/+1
|
* usb-widget: fix gtk2 Python bindingsMarc-André Lureau2013-08-061-10/+2
| | | | | | | | The Python bindings generator failed to bind the USB widget, because of the object/class declaration. The declaration was circumventing the deprecated errors when compiling with GTK_DISABLE_DEPRECATED. We used to need that because of broken gtk+ headers, but it is no longer necessary since 15bd7ceba1434b5d710bfd16078044f30693467b.
* session: copy "ca" property in copy ctorMarc-André Lureau2013-07-221-0/+1
| | | | | This fixes the GSpice-WARNING **: no cert loaded, when doing a seamless migration (when using the "ca" property).
* smartcard: Only init manager onceChristophe Fergeau2013-07-111-17/+29
| | | | | | The SpiceSmartcardManager is a singleton, so it does not make a lot of sense to try to init it multiple times. This commit adds a GOnce to ensure the manager is only init'ed once.
* smartcard: Handle VCARD_EMUL_INIT_ALREADY_INITEDChristophe Fergeau2013-07-111-1/+4
| | | | | | | | | | | | | When initializing a software smartcard, vcard_emul_init() can report success, error, or indicate that initialization has already been done. In this last case, we would assume that an error occurred instead of behaving as if the initialization succeeded. vcard_emul_init() can end up being called multiple time if the smartcard channel gets destroyed and recreated during the lifetime of the application Fixes rhbz#815639
generated by cgit (git 2.34.1) at 2025-09-30 00:32:49 +0000