[SOLVED] Pale Moon crashes when I try to log in to Yahoo

Emerson · 09-27-2017, 11:52 AM

I still believe there is something fishy with your graphics. Does glxinfo tell direct rendering is enabled? Anything in Xorg log that points to rendering problems? Xorg defaults to modesetting with Intel graphics, some Intel chipsets work better with Intel driver. I suggested testing before but you never came back with results.

hazel · 09-28-2017, 01:27 AM

Xorg log shows that the modesetting driver is not used either in Crux or in LFS. It gets loaded, then i915 gets loaded and modesetting gets unloaded again. I looked for significant differences between the Xorg logs of Crux and LFS but couldn't find any. I'm in Crux at the moment. I'll run glxinfo in LFS and post the result. I'm afraid it's all Greek to me; I've never understood graphics.

hazel · 09-28-2017, 06:41 AM

OK. Here are the relevant files for the current (not the old) LFS. This is the one where Pale Moon is stable except for logging in to Yahoo. The first file is the output of glxinfo, the second is my Xorg.log.

btw if someone could find me an entry-level explanation of how a computer does graphics, I'd be grateful. Don't tell me to go to Wikipedia

; the articles there on this subject are already beyond my competence level.

Emerson · 09-28-2017, 07:58 AM

I'm no X expert either. This is all I know:
There is kernel DRI driver, if it is loaded you get /dev/dri/card0 node and can have hardware acceleration.
Then there is Xorg driver.
Plus MESA for 3D.

Here is interesting part from your glxinfo:

Code:

Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Mesa Project (0x8086)
    Device: i915 (chipset: Q33) (0x29d2)
    Version: 17.1.6
    Accelerated: yes
    Video memory: 384MB
    Unified memory: yes
    Preferred profile: compat (0x2)
    Max core profile version: 0.0
    Max compat profile version: 2.1
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 2.0
OpenGL vendor string: Mesa Project
OpenGL renderer string: Gallium 0.4 on i915 (chipset: Q33)
OpenGL version string: 2.1 Mesa 17.1.6

Here is mine, for comparison:

Code:

Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Intel Open Source Technology Center (0x8086)
    Device: Mesa DRI Intel(R) HD Graphics 630 (Kaby Lake GT2)  (0x5912)
    Version: 17.2.1
    Accelerated: yes
    Video memory: 3072MB
    Unified memory: yes
    Preferred profile: core (0x1)
    Max core profile version: 4.5
    Max compat profile version: 3.0
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 3.2
OpenGL vendor string: Intel Open Source Technology Center
OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 630 (Kaby Lake GT2) 
OpenGL core profile version string: 4.5 (Core Profile) Mesa 17.2.1
OpenGL core profile shading language version string: 4.50
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
... 
long list of supported functions
...

Frankly, I do not know what to make of it. Do you have same glxinfo output in Crux?

Emerson · 09-28-2017, 08:23 AM

I did some search regarding Intel and gallium3d. As I mentioned before, Gentoo even does not offer gallium for Intel. And it seems to be the correct choice, I could not find anything encouraging its usage with Intel.

hazel · 09-28-2017, 08:26 AM

Crux doesn't currently have the glx utils but I can install them. They'll be in the mesa-demos package.

The difference in the glxinfo output between my system and yours is striking! I think you're right: if we can find out what the difference means, we will have a handle on this.

I do have the card device you mentioned. In fact I have two devices in that directory:

crw-rw---- 1 root video 226, 0 Sep 28 12:30 card0
crw-rw---- 1 root video 226, 128 Sep 28 12:30 renderD128

I am in the video group so I should have access to both.

Here's what LFS says about gallium drivers:
The available drivers are i915, nouveau, r300, r600, radeonsi, freedreno, svga, swrast, vc4, and virgl. If the Gallium drivers are not specified, the default is 'r300,r600,svga,swrast'. Modify the command below for any desired drivers. The listed drivers below will cover most modern video cards. For help in selecting drivers see http://www.mesa3d.org/systems.html.
GLL_DRV="i915,r600,nouveau,radeonsi,svga,swrast"

Emerson · 09-28-2017, 10:01 AM

Just in case, here is how I load modesetting instead of intel.

Code:

Section "Device"
    Identifier  "Intel Graphics"
    Driver      "modesetting"
    Option      "AccelMethod"    "glamor"
    Option      "DRI"            "3"
EndSection

hazel · 09-28-2017, 10:44 AM

Quote:

Originally Posted by Emerson

Just in case, here is how I load modesetting instead of intel.

Code:

Section "Device"
    Identifier  "Intel Graphics"
    Driver      "modesetting"
    Option      "AccelMethod"    "glamor"
    Option      "DRI"            "3"
EndSection

That partly worked. When I restarted X, it did load the modesetting driver. But then it choked on glamor:

Code:

[ 13572.040] (**) modeset(0): Option "AccelMethod" "glamor"
[ 13572.040] (==) modeset(0): RGB weight 888
[ 13572.040] (==) modeset(0): Default visual is TrueColor
[ 13572.040] (II) Loading sub module "glamoregl"
[ 13572.040] (II) LoadModule: "glamoregl"
[ 13572.040] (II) Loading /usr/lib/xorg/modules/libglamoregl.so
[ 13572.101] (II) Module glamoregl: vendor="X.Org Foundation"
[ 13572.101]    compiled for 1.19.3, module version = 1.0.0
[ 13572.101]    ABI class: X.Org ANSI C Emulation, version 0.4
[ 13572.101] (II) glamor: OpenGL accelerated X.org driver based.
[ 13572.150] (II) glamor: EGL version 1.4 (DRI2):
[ 13572.153] (II) modeset(0): glamor initialized
.....
[ 13572.188] (WW) glamor requires at least 128 instructions (64 reported)
[ 13572.188] (EE) modeset(0): Failed to initialize glamor at ScreenInit() time.
[ 13572.188] (EE) 
Fatal server error:
[ 13572.188] (EE) AddScreen/ScreenInit failed for driver 0
[ 13572.188] (EE) 
[ 13572.188] (EE)

I gather that glamor is part of the X server, which I built (as recommended) with --enable-glamor.

Emerson · 09-28-2017, 11:03 AM

Try without glamor, then. Just curious if this fixes your Yahoo.

hazel · 09-28-2017, 12:06 PM

Quote:

Originally Posted by Emerson

Try without glamor, then. Just curious if this fixes your Yahoo.

Hallelujah! Emmerson, you're a genius!

Glamor is the default accelerator so I had great difficulty getting rid of it. In the end, I specified uxa, because that's what the Intel driver uses. And now I can log into Yahoo.

Just out of curiosity, I'm going to do the same thing on the old LFS and see if I can stabilise PM on that too.

That still doesn't solve the other problem, the incompatibility between that PM header and gcc-7, which forces me to build it in another system. I've asked about that on the LFS mailing list but haven't got an answer yet. It's going to be more of a problem going forward as gcc-6 comes to its end-of-life. Still, this thread was about Yahoo and I'm going to mark it solved.

Emerson · 09-28-2017, 12:29 PM

Glad you got it working.

Although this is just workaround, something is wrong with your video acceleration.

hazel · 09-28-2017, 12:50 PM

Quote:

Originally Posted by Emerson

Glad you got it working.

Although this is just workaround, something is wrong with your video acceleration.

So how do I find out what that could be?

Actually, you're assuming that something is actively wrong. But maybe nothing's wrong, it's just that my aging hardware doesn't support glamor.

Emerson · 09-28-2017, 01:07 PM

I think your glxinfo showing in compat mode indicates a problem.

hazel · 09-29-2017, 01:56 AM

Quote:

Originally Posted by Emerson

I think your glxinfo showing in compat mode indicates a problem.

Why? Please enlarge. What does "compat" mean anyway? Compatible with what? And why is it a bad thing?

Just for fun, I switched the xorg server in my old LFS to use the modesetting driver to see if PM would become stable on that system. But it still crashes immediately after loading. You can't use it for anything.

hazel · 09-29-2017, 05:54 AM

Now here's something interesting. If I run in LFS 8.1 the PM-27.4.0 that I built myself in LFS 8.0 with gcc-6, it behaves much better than it does in its native system. It still segfaults when I try to log into Yahoo (which PM-27.6.0a1 no longer does, thanks to Emerson) but otherwise it seems to be stable. Maybe that's because LFS 8.0 has much older graphics software (half a year older in fact). I'm beginning to think that the interactions between this program and the underlying system are simply too complicated to explain.