Intel bug? Developers broken?

[BOT Space]

I remember a little while ago there was a rant on here, that the puddle doesn't show up ingame. It seemed consense, that that was an bug only occuring on Intel Cards (and I think also only on Linux too). Shortly thereafter intel HD was removed from the system requirements on the product page. Now it's back up there now, even though the puddle still doesn't show up.

Do I miss-remember something here? Or is it really the was it seems to be?

[ringerc]

I'm certainly unable to see the liquid on my Fedora 18 x64 system, though the game otherwise functions fine.

The game produces no output on the console by default. If I set LD_DEBUG=libs,symbols when I launch Steam, the 250MB log file contains:

4998: /lib/libGL.so.1: error: symbol lookup error: undefined symbol: glActiveStencilFaceEXT (fatal)
4998: /lib/libGL.so.1: error: symbol lookup error: undefined symbol: glBlendEquationSeparateEXT (fatal)
4998: /lib/libGL.so.1: error: symbol lookup error: undefined symbol: glProgramEnvParameters4fvEXT (fatal)
4998: /lib/libGL.so.1: error: symbol lookup error: undefined symbol: glProgramLocalParameters4fvEXT (fatal)

from Puddle (pid 4998 in this run).

This makes me wonder if Puddle is violating the GLX ABI and trying to link directly to driver-exported symbols.

When I launch Puddle then examine its virtual address mappings to get the loaded libraries - including runtime dlopen()ed libs `ldd` can't see - I find:

$ grep -v Steam /proc/5333/maps | grep \.so | awk '{ print $6; }' | sort -u
/usr/lib/dri/i965_dri.so
/usr/lib/ld-2.16.so
/usr/lib/libc-2.16.so
/usr/lib/libdl-2.16.so
/usr/lib/libdricore9.1.0.so.1.0.0
/usr/lib/libdrm_intel.so.1.0.0
/usr/lib/libdrm.so.2.4.0
/usr/lib/libglapi.so.0.0.0
/usr/lib/libGL.so.1.2.0
/usr/lib/libm-2.16.so
/usr/lib/libnsl-2.16.so
/usr/lib/libnss_files-2.16.so
/usr/lib/libpciaccess.so.0.11.1
/usr/lib/libpthread-2.16.so
/usr/lib/libresolv-2.16.so
/usr/lib/librt-2.16.so
/usr/lib/libxcb-dri2.so.0.0.0

I'm an experienced C programmer and I'm comfortable with gdb etc, so I'm very happy to help the team with testing debug-enabled builds if I can do anything that'll help you out.

Going to do some runs with various GLX debug options; I'll follow up with more info.

[ringerc]

This github issue suggests that the issue is fixed in Fedora 19's Mesa.

https://github.com/ValveSoftware/steam-for-linux/issues/1990

As I noted there you can force software rendering to see if you're experiencing this driver-specific issue... though you'll get a pretty slide-show.