Re: OpenGL forwarding via ssh session: works on jessie, does not work on stretch

[Étienne Mollier <etienne.mollier@mailoo.org>]

Simon, on 2019-06-02:
> The following test setup:
> (A) linux machine running ubuntu (Nvidia GPU)
> (B) BBB running jessie
> (C) BBB running stretch
>
> I connect from (A) -> (*) by ssh -X x@host
>
> When I connect from (A) to (B):
> > export LIBGL_ALWAYS_INDIRECT=1;  glxinfo|head
> name of display: localhost:10.0
> display: localhost:10  screen: 0
> direct rendering: No (LIBGL_ALWAYS_INDIRECT set)
> server glx vendor string: NVIDIA Corporation
> server glx version string: 1.4
> server glx extensions:
>
> When I connect from (A) to (C):
> > export LIBGL_ALWAYS_INDIRECT=1;  glxinfo|head
> Error: couldn't find RGB GLX visual or fbconfig
> name of display: localhost:10.0

Good Day Simon,

At first I was thinking that could have to do with remote GLX
support of your X server, which requires to be explicitly
enabled on Debian Stretch using Xorg option +iglx for instance.
However, your setup and your symptoms do not match this
situation.  I vaguely recall having seen this "couldn't find RGB
GLX visual" message a couple of times when working on hardware
accelerated remote direct rendering though, but can't recall the
exact situation.


Here is an attempt to reproduce your issue, except that my setup
is:
(A) Amd running Sid
(C) Eeepc running Stretch

If I start X on (A) without "+iglx", then ssh -X to (C), I get:

	$ glxinfo | grep '^direct rendering:'
	direct rendering: Yes

	$ LIBGL_ALWAYS_INDIRECT=1 glxinfo | grep '^direct rendering:'
	X Error of failed request:  GLXBadContext
	  Major opcode of failed request:  152 (GLX)
	  Minor opcode of failed request:  6 (X_GLXIsDirect)
	  Serial number of failed request:  34
	  Current serial number in output stream:  33

	$ glxdemo
	/* Redrawing the window is clunky. */

	$ glxgears
	50 frames in 5.1 seconds =  9.875 FPS

If I start X on (A) with "+iglx", then ssh -X to (C), I get:

	$ glxinfo | grep '^direct rendering:'
	direct rendering: Yes

	$ LIBGL_ALWAYS_INDIRECT=1 glxinfo | grep '^direct rendering:'
	direct rendering: No (LIBGL_ALWAYS_INDIRECT set)

	$ glxdemo
	/* Redrawing the window is clunky. */

	$ LIBGL_ALWAYS_INDIRECT=1 glxdemo
	/* Redrawing the window is smooth. */

	$ glxgears
	35 frames in 5.0 seconds =  6.953 FPS

	$ LIBGL_ALWAYS_INDIRECT=1 glxgears
	/* Can't get performances output because my X server...
	 * erm...
	 * dies instantly of a segmentation violation.
	 */

Anyway, error messages are not the same as yours, so there may
be something else in play in your setup that might have been
missed.  For instance my architecture was different from yours,
the X server is run on Debian instead of Ubuntu, the graphic
card is an Amd instead of an Nvidia, etc.

Perhaps someone else with a better knowledge of X internals
might pop-up with better insights.

Kind Regards,
-- 
Étienne Mollier <etienne.mollier@mailoo.org>


Off-topic, for the curious mind, the Xorg log says:

	Backtrace:
	0: /usr/lib/xorg/Xorg (OsLookupColor+0x139) [0x55a034a102c9]
	1: /lib/x86_64-linux-gnu/libpthread.so.0 (funlockfile+0x50) [0x7fa380c9b77f]
	2: /usr/lib/xorg/Xorg (ResetCurrentRequest+0xb) [0x55a034a0f34b]
	3: /usr/lib/xorg/Xorg (DRI2WaitSwap+0x56) [0x55a0349dc766]
	unw_get_proc_name failed: no unwind info found [-10]
	4: /usr/lib/xorg/modules/extensions/libglx.so (?+0x0) [0x7fa3803e1a50]
	unw_get_proc_name failed: no unwind info found [-10]
	5: /usr/lib/xorg/modules/extensions/libglx.so (?+0x0) [0x7fa3803da040]
	unw_get_proc_name failed: no unwind info found [-10]
	6: /usr/lib/xorg/modules/extensions/libglx.so (?+0x0) [0x7fa3803d5ff0]
	7: /usr/lib/xorg/Xorg (dri3_send_open_reply+0xe0a) [0x55a0349e112a]
	8: /usr/lib/xorg/Xorg (_CallCallbacks+0x34) [0x55a0348b6c24]
	9: /usr/lib/xorg/Xorg (CloseDownClient+0x5f) [0x55a0348b0d0f]
	10: /usr/lib/xorg/Xorg (SendErrorToClient+0x39b) [0x55a0348b1a0b]
	11: /usr/lib/xorg/Xorg (InitFonts+0x3b6) [0x55a0348b5956]
	12: /lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main+0xeb) [0x7fa380aea09b]
	13: /usr/lib/xorg/Xorg (_start+0x2a) [0x55a03489f67a]

	Segmentation fault at address 0x8

Not sure if I should open a bug about this; I'm running a kernel
with a custom enough config to wonder if the issue came from the
thing between the chair and the keyboard in the first place.