[winswitch] Xpra GL and gtkglext woes

Ben Sferrazza bsferrazza at avnera.com
Mon Mar 5 22:02:34 GMT 2018


Figured I'd add the Xorg log file, with the Modeline lines stripped out.
Not sure why it's trying to load libinput when it's using the dummy_mouse
and dummy_keyboard.

[3017180.958] X Protocol Version 11, Revision 0
[3017180.958] Build Operating System: Linux 2.6.18-398.el5 x86_64
[3017180.958] Current Operating System: Linux l-server-13 2.6.18-398.el5 #1
SMP Tue Sep 16 20:50:52 EDT 2014 x86_64
[3017180.958] Kernel command line: ro root=LABEL=/ rhgb quiet
[3017180.958] Build Date: 05 March 2018  12:19:47AM
[3017180.958]
[3017180.958] Current version of pixman: 0.34.0
[3017180.958]   Before reporting problems, check http://wiki.x.org
        to make sure that you have the latest version.
[3017180.958] Markers: (--) probed, (**) from config file, (==) default
setting,
        (++) from command line, (!!) notice, (II) informational,
        (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
[3017180.958] (++) Log file: "/home/bsferrazza/.xpra/Xvfb-10.log", Time:
Mon Mar  5 21:55:36 2018
[3017180.961] (++) Using config file: "/home/tools/etc/X11/xorg.conf"
[3017180.961] (==) Using config directory: "/home/tools/etc/X11/xorg.conf.d"
[3017180.961] (==) ServerLayout "dummy_layout"
[3017180.961] (**) |-->Screen "dummy_screen" (0)
[3017180.961] (**) |   |-->Monitor "dummy_monitor"
[3017180.961] (**) |   |-->Device "dummy_videocard"
[3017180.961] (**) |-->Input Device "dummy_mouse"
[3017180.961] (**) |-->Input Device "dummy_keyboard"
[3017180.961] (**) Option "DontVTSwitch" "true"
[3017180.961] (**) Option "AllowMouseOpenFail" "true"
[3017180.961] (**) Option "AutoAddDevices" "false"
[3017180.961] (**) Option "AutoEnableDevices" "false"
[3017180.961] (**) Not automatically adding devices
[3017180.961] (**) Not automatically enabling devices
[3017180.961] (==) Not automatically adding GPU devices
[3017180.961] (==) Max clients allowed: 256, resource mask: 0x1fffff
[3017180.961] (==) FontPath set to:
        ${prefix}/share/fonts/X11/misc/,
        ${prefix}/share/fonts/X11/TTF/,
        ${prefix}/share/fonts/X11/OTF/,
        ${prefix}/share/fonts/X11/Type1/,
        ${prefix}/share/fonts/X11/100dpi/,
        ${prefix}/share/fonts/X11/75dpi/
[3017180.961] (**) ModulePath set to "/home/tools/lib/xorg/modules/"
[3017180.961] (II) Loader magic: 0x80db00
[3017180.961] (II) Module ABI versions:
[3017180.961]   X.Org ANSI C Emulation: 0.4
[3017180.961]   X.Org Video Driver: 23.0
[3017180.961]   X.Org XInput driver : 24.1
[3017180.961]   X.Org Server Extension : 10.0
[3017180.971] (--) PCI:*(0:10:0:0) 102b:0534:1028:0600 rev 1, Mem @
0x90000000/16777216, 0x91800000/16384, 0x91000000/8388608
[3017180.971] (II) Open ACPI successful (/var/run/acpid.socket)
[3017180.971] (II) LoadModule: "glx"
[3017180.976] (II) Loading /home/tools/lib/xorg/modules/extensions/libglx.so
[3017181.002] (II) Module glx: vendor="X.Org Foundation"
[3017181.002]   compiled for 1.19.6, module version = 1.0.0
[3017181.002]   ABI class: X.Org Server Extension, version 10.0
[3017181.002] (II) LoadModule: "dummy"
[3017181.008] (II) Loading /home/tools/lib/xorg/modules/drivers/dummy_drv.so
[3017181.009] (II) Module dummy: vendor="X.Org Foundation"
[3017181.009]   compiled for 1.19.6, module version = 0.3.8
[3017181.009]   Module class: X.Org Video Driver
[3017181.009]   ABI class: X.Org Video Driver, version 23.0
[3017181.009] (II) LoadModule: "void"
[3017181.010] (WW) Warning, couldn't open module void
[3017181.010] (II) UnloadModule: "void"
[3017181.010] (II) Unloading void
[3017181.010] (EE) Failed to load module "void" (module does not exist, 0)
[3017181.010] (II) DUMMY: Driver for Dummy chipsets: dummy
[3017181.010] (WW) Falling back to old probe method for dummy
[3017181.010] (WW) VGA arbiter: cannot open kernel arbiter, no multi-card
support
[3017181.010] (II) DUMMY(0): Chipset is a DUMMY
[3017181.010] (++) DUMMY(0): Depth 24, (--) framebuffer bpp 32
[3017181.010] (==) DUMMY(0): RGB weight 888
[3017181.010] (==) DUMMY(0): Default visual is TrueColor
[3017181.010] (==) DUMMY(0): Using gamma correction (1.0, 1.0, 1.0)
[3017181.010] (**) DUMMY(0): VideoRAM: 192000 kByte
[3017181.010] (--) DUMMY(0): Max Clock: 300000 kHz
:
:
[3017181.012] (++) DUMMY(0): DPI set to (96, 96)
[3017181.012] (II) Loading sub module "fb"
[3017181.012] (II) LoadModule: "fb"
[3017181.013] (II) Loading /home/tools/lib/xorg/modules/libfb.so
[3017181.013] (II) Module fb: vendor="X.Org Foundation"
[3017181.013]   compiled for 1.19.6, module version = 1.0.0
[3017181.013]   ABI class: X.Org ANSI C Emulation, version 0.4
[3017181.013] (II) Loading sub module "ramdac"
[3017181.013] (II) LoadModule: "ramdac"
[3017181.013] (II) Module "ramdac" already built-in
[3017181.013] (--) Depth 24 pixmap format is 32 bpp
[3017181.013] (II) DUMMY(0): Using 1904 scanlines of offscreen memory
[3017181.013] (==) DUMMY(0): Backing store enabled
[3017181.013] (==) DUMMY(0): Silken mouse enabled
[3017181.013] (WW) DUMMY(0): Option "ConstantDPI" is not used
[3017181.013] (==) RandR enabled
[3017181.014] (II) AIGLX: Screen 0 is not DRI2 capable
[3017181.014] (EE) AIGLX: reverting to software rendering
[3017181.020] (II) IGLX: enabled GLX_MESA_copy_sub_buffer
[3017181.021] (II) IGLX: Loaded and initialized swrast
[3017181.021] (II) GLX: Initialized DRISWRAST GL provider for screen 0
[3017181.131] (II) LoadModule: "void"
[3017181.133] (WW) Warning, couldn't open module void
[3017181.133] (II) UnloadModule: "void"
[3017181.133] (II) Unloading void
[3017181.133] (EE) Failed to load module "void" (module does not exist, 0)
[3017181.133] (EE) No input driver matching `void'
[3017181.133] (II) Falling back to input driver `libinput'
[3017181.133] (II) LoadModule: "libinput"
[3017181.134] (WW) Warning, couldn't open module libinput
[3017181.134] (II) UnloadModule: "libinput"
[3017181.134] (II) Unloading libinput
[3017181.134] (EE) Failed to load module "libinput" (module does not exist,
0)
[3017181.134] (II) LoadModule: "void"
[3017181.136] (WW) Warning, couldn't open module void
[3017181.136] (II) UnloadModule: "void"
[3017181.136] (II) Unloading void
[3017181.136] (EE) Failed to load module "void" (module does not exist, 0)
[3017181.136] (EE) No input driver matching `void'
[3017181.136] (II) Falling back to input driver `libinput'
[3017181.136] (II) LoadModule: "libinput"
[3017181.137] (WW) Warning, couldn't open module libinput
[3017181.137] (II) UnloadModule: "libinput"
[3017181.137] (II) Unloading libinput
[3017181.137] (EE) Failed to load module "libinput" (module does not exist,
0)


On Mon, Mar 5, 2018 at 1:25 PM, Ben Sferrazza <bsferrazza at avnera.com> wrote:

> On Mon, Mar 5, 2018 at 11:50 AM, Antoine Martin via shifter-users <
> shifter-users at lists.devloop.org.uk> wrote:
>
>> On 06/03/18 00:38, Ben Sferrazza via shifter-users wrote:
>> > Hi. I've built Xpra 2.2.4 on a Linux server at work which runs an old
>> > distribution, CentOS 5. The kernel is 2.6.18, but I've created a fully
>> > bootstrapped toolchain/binary/library environment that runs the latest
>> > version of glibc (2.19) for the kernel, gcc 7.3.0, and binutils 2.30,
>> > following most of the Linux From Scratch book. Everything I've thrown
>> at it
>> > works well, even having built Xorg, GTK+2/3, and GUI applications. I
>> made
>> > sure to build all the Xpra dependencies and built it with
>> > --prefix=/home/tools, which is where my updated environment is.
>> Impressive!
>>
>> I am considering using a similar approach (maybe using snap / flatpack /
>> appimage) to continue to support CentOS 6 in the future.
>>
>
> Thanks. Certainly wasn't easy at first, and had to make quite few tweaks
> and workarounds along the way.
>
>
>>
>> > I have built Mesa, and am using the Xdummy xorg.conf. All seems well,
>> but
>> > the Xpra client never showed the --start-child app of konsole. It had
>> > errors like this.
>> >
>> > 2018-03-05 17:04:40,440 x_event_filter event=('xpra-create-event',
>> > None)/CreateNotify window=0x299
>> > 2018-03-05 17:04:40,440 cannot get gdk window for <gtk.gdk.Display
>> object
>> > at 0x2ae82f5ea050 (GdkDisplayX11 at 0x1b3be1f0)>, 12582913
>> > 2018-03-05 17:04:40,441 XError: XError: 3 processing CreateNotify
>> > Traceback (most recent call last):
>> >   File "xpra/x11/gtk2/gdk_bindings.pyx", line 1062, in
>> > xpra.x11.gtk2.gdk_bindings.parse_xevent
>> >   File "xpra/x11/gtk2/gdk_bindings.pyx", line 914, in
>> > xpra.x11.gtk2.gdk_bindings._gw
>> > XError: XError: 3
>> > 2018-03-05 17:04:40,442 Some window in our event disappeared before we
>> > could handle the event 16/CreateNotify using ('xpra-create-event',
>> None);
>> > so I'm just ignoring it instead. python event=<X11:CreateNotify
>> > {'delivered_to': '0x299', 'send_event': 0, 'serial': '0x36a', 'type':
>> 16,
>> > 'display': ':100'}>
>> Error 3 is BadWindow. This usually means that the window got destroyed
>> before we could get hold of it. (X11 is asynchronous)
>> I assume that you got this error in your server log with some debug
>> flags enabled, right?
>>
>
> correct. I start xpra using '-d all'
>
> xpra start :100 --start-child=xterm --socket-dirs=~/.xpra -d all
>
>
>>
>> > So I then went ahead and installed gtkglext and the Python bindings,
>> > thinking it was somehow related to OpenGL.
>> What makes you think that?
>> What vfb are you using? Are you using Xdummy or Xvfb? Have you tried
>> switching to the other option? Does your vfb support opengl?
>> Is using VirtualGL an option?
>> (there's quite a bit on the wiki about all these options)
>>
>
> It was just a hunch. Perhaps a poor one. I believe to be using Xdummy. I
> have the latest Xvfb as part of the xorg-server-1.19.6 tar ball. I placed
> the following line in my /home/tools/etc/xpra/xpra.conf
>
> xvfb=Xorg -dpi 96 -noreset -nolisten tcp \
>                         +extension GLX +extension RANDR +extension RENDER \
>                         -logfile ${HOME}/.xpra/Xvfb-10.log -config
> /home/tools/etc/X11/xorg.conf
>
> I have built and installed the latest xf-video-dummy, am using the
> xorg.conf provided by Xpra. Is this sufficient to be using Xdummy?
>
> I'll have to give VirtualGL a try. These servers at work are headless
> multi-Xeon CPU systems without much of a video card. An lspci -v | grep VGA
> returns
>
> 0a:00.0 VGA compatible controller: Matrox Electronics Systems Ltd. G200eR2
> (rev 01) (prog-if 00 [VGA controller])
>
> The are completely modern systems, so I find it odd that it would have a
> video card from the late 90s in it.
>
>
>> > However, passing --opengl=no on
>> > the command line when starting xpra appeared to do nothing.
>> Correct, it doesn't do anything to the server. This is a client command
>> line option for enabling the OpenGL accelerated rendering backend.
>>
>> > But now I get
>> > this mysterious ImportError
>> >
>> > 2018-03-05 17:04:38,809 err(xpra opengl)=xpra main error:
>> > Traceback (most recent call last):
>> >   File "/home/tools/lib/python/xpra/scripts/main.py", line 175, in main
>> >     return run_mode(script_file, err, options, args, mode, defaults)
>> >   File "/home/tools/lib/python/xpra/scripts/main.py", line 1517, in
>> run_mode
>> >     return run_glcheck(options)
>> >   File "/home/tools/lib/python/xpra/scripts/main.py", line 2723, in
>> > run_glcheck
>> >     from xpra.client.gl.gtk_base.gtkgl_check import check_support
>> >   File "/home/tools/lib/python/xpra/client/gl/gtk_base/gtkgl_check.py",
>> > line 35, in <module>
>> >     from xpra.client.gl.gtk_base.gtk_compat import MODE_RGBA,
>> MODE_ALPHA,
>> > MODE_RGB, MODE_DOUBLE, MODE_SINGLE, MODE_DEPTH
>> >   File "/home/tools/lib/python/xpra/client/gl/gtk_base/gtk_compat.py",
>> line
>> > 59, in <module>
>> >     from gtk import gdkgl, gtkgl
>> >   File
>> > "/home/tools/lib/python2.7/site-packages/gtk-2.0/gtk/gdkgl/_
>> _init__.py",
>> > line 21, in <module>
>> >     from _gdkgl import *
>> > ImportError:
>> > /home/tools/lib/python2.7/site-packages/gtk-2.0/gtk/gdkgl/_gdkgl.so:
>> > undefined symbol: gdk_window_is_gl_capable
>> >
>> > When I run strings on that _gdkgl.so, I get the following
>> >
>> > bash-4.4$ strings _gdkgl.so | grep gdk_window_is_gl_capable
>> > gdk_window_is_gl_capable
>> > _wrap_gdk_window_is_gl_capable
>> > _wrap_gdk_window_is_gl_capable
>> >
>> > and the function is defined in the following include file
>> >
>> > /home/tools/include/gtkglext-1.0/gdk/gdkglwindow.h
>> >
>> > bash-4.4$ grep gdk_window_is_gl_capable gdkglwindow.h
>> > gboolean     gdk_window_is_gl_capable       (GdkWindow   *window);
>> Looks like your build of pygtkgl is not quite right.
>> This is unrelated to the missing window problem.
>>
>
> OK, the latter is reassuring. Using both the latest sources from the git
> repositories of gtkglext and pygtkglext, I had to use the patch file you
> created as part of https://www.xpra.org/trac/ticket/226 to get it to even
> build correctly. The only change I made to the patch was removing the OSX
> dependent paths in configure.in. $TOOLS = /home/tools.
>
> +CFLAGS="$CFLAGS -I$TOOLS/include/gtkglext-1.0 -I$TOOLS/lib/gtkglext-1.0/include
> -I$TOOLS/include/"
>
>
>> During startup the xpra server probes the vfb's OpenGL capabilities by
>> running "xpra opengl" in a subprocess (so it won't crash the server if
>> opengl is somehow badly misconfigured to the point of causing crashes)
>> The message you are seeing just means that the probing fails.
>>
>> > Perhaps the issue of not getting a window drawn is unrelated to the
>> > ImportError, but I'd like to fix both of these issues. Thank you for any
>> > help in solving this!
>> Is the window not being drawn, or not being shown at all?
>>
>> Are you sure that the application (konsole in this case) is still
>> running and hasn't just crashed?
>> The best way to debug this is to launch an xterm instead, then run your
>> applications from there. Or maybe even with gdb if it does crash.
>>
>
> konsole is just the KDE terminal program. But good point. Changing it to
> xterm seems to suggest I'm not even connecting. I am prompted for my
> password and the Xpra client (Windows) disappears. Seems to be running as
> per taskmgr, but the tray icon goes away.
>
> Here's where the child process is executed.
>
> 2018-03-05 20:25:44,790 exec_start_commands() start=[],
> start_child=['xterm']
> 2018-03-05 20:25:44,790 start_child('xterm', 'xterm', False, None, True,
> None, {})
> 2018-03-05 20:25:44,791 threaded_init() start
> 2018-03-05 20:25:44,796 add_process(<subprocess.Popen object at
> 0x2ba85fca5cd0>, xterm, ['xterm'], False, False) pid=7293
> 2018-03-05 20:25:44,797 pid(['xterm'])=7293
> 2018-03-05 20:25:44,797 started command 'xterm' with pid 7293
> 2018-03-05 20:25:44,797 poll() procinfo list: [ProcInfo({'returncode':
> None, 'name': 'xterm', 'process': <subprocess.Popen object at
> 0x2ba85fca5cd0>, 'pid': 7293, 'dead': False, 'ignore': False, 'callback':
> None, 'command': ['xterm'], 'forget': False})]
> 2018-03-05 20:25:44,797 guess_session_name() commands=['xterm']
> 2018-03-05 20:25:44,797 <bound method XpraServer.init_sockets of
> <XpraServer object at 0x2ba85a41b460 (xpra+x11+server+XpraServer at
> 0xc9849a0)>>([('unix-domain', <socket._socketobject object at
> 0x2ba857935d00>, '/home/bsferrazza/.xpra/l-server-13-100')])
> 2018-03-05 20:25:44,798 init_sockets([('unix-domain',
> <socket._socketobject object at 0x2ba857935d00>, '/home/bsferrazza/.xpra/l-server-13-100')])
> will add unix-domain socket <socket._socketobject object at 0x2ba857935d00>
> (/home/bsferrazza/.xpra/l-server-13-100)
> 2018-03-05 20:25:44,798 added unix socket path: /home/bsferrazza/.xpra/l-
> server-13-100
> 2018-03-05 20:25:44,798 local sockets we can use for printing:
> ['/home/bsferrazza/.xpra/l-server-13-100']
> 2018-03-05 20:25:44,798 <bound method XpraServer.init_when_ready of
> <XpraServer object at 0x2ba85a41b460 (xpra+x11+server+XpraServer at
> 0xc9849a0)>>([])
> 2018-03-05 20:25:44,798 running <bound method XpraServer.run of
> <XpraServer object at 0x2ba85a41b460 (xpra+x11+server+XpraServer at
> 0xc9849a0)>>
> 2018-03-05 20:25:44,799 xpra X11 version 2.2.4-r18312 64-bit
> 2018-03-05 20:25:44,801  uid=10345 (bsferrazza), gid=1001 (eng)
> 2018-03-05 20:25:44,801  running with pid 6769 on Linux CentOS 5.11 Final
> 2018-03-05 20:25:44,802  connected to X11 display :100 with 24 bit colors
> 2018-03-05 20:25:44,802 do_run() calling <function gtk_main at
> 0x2ba84d7a52a8>
> 2018-03-05 20:25:44,802 poll() procinfo list: [ProcInfo({'returncode':
> None, 'name': 'xterm', 'process': <subprocess.Popen object at
> 0x2ba85fca5cd0>, 'pid': 7293, 'dead': False, 'ignore': False, 'callback':
> None, 'command': ['xterm'], 'forget': False})]
> 2018-03-05 20:25:44,803 check() alive=[ProcInfo({'returncode': None,
> 'name': 'xterm', 'process': <subprocess.Popen object at 0x2ba85fca5cd0>,
> 'pid': 7293, 'dead': False, 'ignore': False, 'callback': None, 'command':
> ['xterm'], 'forget': False})], quit callback=None
> 2018-03-05 20:25:44,803 x_event_filter event=('xpra-property-notify-event',
> None)/PropertyNotify window=0x400001
> 2018-03-05 20:25:44,803 parse_event(..)=<X11:PropertyNotify
> {'delivered_to': '0x400001', 'send_event': 0, 'window': '0x400001', 'atom':
> '_NET_WM_NAME', 'serial': '0x3a', 'type': 28, 'display': ':100'}>
> 2018-03-05 20:25:44,804 x_event_filter event=('xpra-property-notify-event',
> None)/PropertyNotify took 0.8ms
> 2018-03-05 20:25:44,804 x_event_filter event=('xpra-property-notify-event',
> None)/PropertyNotify window=0x400001
> 2018-03-05 20:25:44,804 parse_event(..)=<X11:PropertyNotify
> {'delivered_to': '0x400001', 'send_event': 0, 'window': '0x400001', 'atom':
> 'WM_NAME', 'serial': '0x3b', 'type': 28, 'display': ':100'}>
> 2018-03-05 20:25:44,804 x_event_filter event=('xpra-property-notify-event',
> None)/PropertyNotify took 0.6ms
> 2018-03-05 20:25:44,805 x_event_filter event=('xpra-property-notify-event',
> None)/PropertyNotify window=0x400001
>
> The bottom of the log file, before and even after attempting a connection
> still just shows this.
>
> 2018-03-05 21:16:09,763 sigchld(17, <frame object at 0x2acc79900238>)
> 2018-03-05 21:16:09,763 poll() procinfo list: [ProcInfo({'returncode':
> None, 'name': 'xterm', 'process': <subprocess.Popen object at
> 0x2acc8e6a9d50>, 'pid': 27371, 'dead': False, 'ignore': False, 'callback':
> None, 'command': ['xterm'], 'forget': False})]
> 2018-03-05 21:16:09,763 reap() calling os.waitpid(-1, 'WNOHANG')
> 2018-03-05 21:16:09,763 reap() waitpid=0
> 2018-03-05 21:16:09,764 add_listen_socket(unix-domain,
> <socket._socketobject object at 0x2acc8633ad00>)
> info=/home/bsferrazza/.xpra/l-server-13-100
> 2018-03-05 21:16:09,764 reset_server_timeout(True) server_idle_timeout=0,
> server_idle_timer=None
> 2018-03-05 21:16:09,764 xpra is ready.
> 2018-03-05 21:16:09,764 org.xpra.Server.Event(ready, [])
> 2018-03-05 21:16:09,764 server-event: ('ready',)
>
> So it's as though it's not even aware of the attempt client connection?
>
> It's probably unrelated but I receive this error immediately when starting
> Xpra. This again is run as non-root. I have XDG_RUNTIME_DIR set to a path
> in my home directory. So I'm not sure why it's still attempting to access
> /run/xpra/system
>
> 2018-03-05 21:16:04,755 get_enabled_encoders(['rencode', 'bencode',
> 'yaml']) enabled=['yaml', 'rencode', 'bencode']
> 2018-03-05 21:16:04,799 netifaces loaded sucessfully
> 2018-03-05 21:16:04,799 successfully loaded socket C library from libc.so.6
> 2018-03-05 21:16:04,804 failed to connect using <bound method
> _socketobject.connect of <socket._socketobject object at
> 0x2b9f9143e520>>/run/xpra/system
> Traceback (most recent call last):
>   File "/home/tools/lib/python/xpra/scripts/main.py", line 1910, in
> _socket_connect
>     sock.connect(endpoint)
>   File "/home/tools/lib/python2.7/socket.py", line 228, in meth
>     return getattr(self._sock,name)(*args)
> error: [Errno 2] No such file or directory
> Warning: cannot use the system proxy for 'start' subcommand,
>  failed to connect to '/run/xpra/system':
>  [Errno 2] No such file or directory
>
>
>
>> As for the gdkgl error, this is a build issue. I would just ignore it
>> and remove the package. Unless you also intend to use this same system
>> as an xpra client, in which case you do want accelerated rendering...
>>
>
> I do not intend to use this system as a client. Thank you for your help!
>
>
>>
>> Cheers
>> Antoine
>>
>
>



More information about the shifter-users mailing list