Blurring not working with fglrx

[alextsits]

I have a hybrid intel/amd laptop, which means I also have Intel and AMD GPUs.
When using intel, i3lock -f works as expected.
However when on AMD (using proprietary fglrx) it does not work. It blurs for a moment and then it probably crashes or something, as my screen is not locked and unblurred.
I don't know how to debug this, so any suggestion is appreciated.

[karulont]

Hi,
There are some stuff in file blur.c, which can be enabled by defining DEBUG_GL.
This can be done by adding -DDEBUG_GL to CPPFLAGS in makefile.

Find line:
CPPFLAGS += -DVERSION="${GIT_VERSION}"
And replace it with this one:
+CPPFLAGS += -DVERSION="${GIT_VERSION}" -DDEBUG_GL

After recompiling (make clean; make), i3lock should print some additional info to stdout.

[alextsits]

Ok, here is what I get in stdout with ati:
V Shader: 1
F Shader: 1
Program: 1

And this is what I get with intel (where it works as expected):
V Shader: 1
shader_infolog:
F Shader: 1
shader_infolog:
Program: 1
shader_infolog:
program_infolog:
V Shader: 1
shader_infolog:
F Shader: 1
shader_infolog:
Program: 1
shader_infolog:
program_infolog:

Don't see how this could help, but I hope it will.
I'm available you need any more info/tests.

[karulont]

I looked at my code a little found out that actually there was a programming error on my part, which might trigger the crash by using ati driver.

I did a commit, which might fix it.

[alextsits]

Just did git pull and tested. It didn't work :/
However using LD_DEBUG=libs i get the following errors:
./i3lock: error: symbol lookup error: undefined symbol: __glXSetTexBufferInfo (fatal)
./i3lock: error: symbol lookup error: undefined symbol: __glXgetGLXPixmapInfo (fatal)

Note that I also get those errors using LD_DEBUG=libs with other programs when I use ati card, (although the programs seem to work fine, like fgl_glxgears).

[alextsits]

@karulont I just found out that for some reason ./i3lock -f --nofork works fine for me even with catalyst drivers. That is weird as --norfork option is not required with i915 driver.

[karulont]

Probably the driver does something funky with the process ID. And after a fork the process ID changes.

To be more accurate: parent process will open up a connection to Xserver and does some stuff. Then a fork happens and the child remains. However child has a different Process ID than the parent, which opened the connection to Xserver.