I've never used the website notifications, just e-mail notifications. I would rather one e-mail per reply, but it was impractical since the e-mails are not threaded together so had to turn it off.

View this comment - View article - View full comments

Here's an idea. Release a new build of Steam that's amd64-only, but maintain a minimal 32-bit "compatibility" build for as long as distributions keep shipping x86 releases for legacy games and operating systems. Allow both to run simultaneously and seamlessly, if selected from the Preferences.

Games can continue to be sold as 32-bit, but it has to be advertised that it only runs under the compatibility-mode client (which people may not opt to install). This will encourage developers to go 64-bit only, but not forcing them to do so.

By doing this, everybody wins. Valve starts doing something to phase x86 out. People with x86-only hardware can keep using Steam, albeit a minimal version (which is probably more suited to older hardware anyway). People who want to play their old 32-bit games still have the option, but have to install the compatibility build (which integrates nicely with the 64-bit build if both are installed). People who want x86 to die can keep their system clean and free of x86 Steam components!

View this comment - View article - View full comments

I brought it with the intent of running it under Wine (and it does indeed run very well under Wine) because the game looks awesome - but I stopped when I heard a GNU/Linux version will be released and will wait for it.

View this comment - View article - View full comments

I've got a CrossOver subscription (just to support CodeWeavers really), but never use it. Now that it's got 64-bit support, I'll be giving it another look!

View this comment - View article - View full comments

Never heard of Roccat before. Only one store I can find here sells anything by that brand - just one mouse model (which is end of life, Ex-demo model).

I personally switched to Ducky keyboards about a year ago. Everything is controlled by hardware - no special drivers required.

View this comment - View article - View full comments

I built my current rig at the start of the year, so I'm not due for a high-performance upgrade/replacement for a while... but when I am I'll definitely go AMD if it stacks up roughly the same (or somewhat close) to Intel.

View this comment - View article - View full comments

I'm with Kimyrielle. It's quite rare that I'll purchase Windows-only games these days, but there might be one or two a year which I cannot resist, which is why I went back to dual-booting - just for those couple of games. Currently, that's Far Cry Primal and Doom. Last year, it was Grey Goo. However Doom and Grey Goo are looking very close to running nicely under Wine. I also went back to Windows for playing Dying Light for a bit because the GNU/Linux performance on AMDGPU-Pro was just too horrible (and the game doesn't work with Mesa). Probably Dying Light under Wine would run quicker when it is able.

If Wine was able to run those one or two must-play games for me each year, and Vulkan games were out so I could make use of both of my GPUs (since CrossFire under GNU/Linux isn't a thing since AMD dropped fglrx), I wouldn't need a Windows install on my gaming machine. Which in practice might well mean I can delete it again in around 12 months.

View this comment - View article - View full comments

Quoting: edddeduckferal

Quoting: boltronicsIt's way harder to get the game working at all on Machine 1. Could the issue be related to clock speed? I don't really want to undo my overclock, since it took many hours to perfect...

No, unless your CPU is failing in some way I doubt the overclock is a factor in your issue, I don't think you should need to undo any overclocking if everything else works perfectly.

I was thinking along the lines of there being a race condition or some such. But like everything I've considered, it seems a long shot.

Quoting: edddeduckferalSounds more likely something in a subsystem what that could be is a mystery but I've logged your (super detailed) reports internally. If anything crops up that we think could be the cause of your problem we'll let you know but I would say the odds are fairly low due to the extremely low reproduction levels.

No worries. If anyone at Feral comes up with something specific you would like me to test or would like more information on anything, feel free to contact me by e-mail at any point via abolte at systemsaviour.com.

View this comment - View article - View full comments

It's way harder to get the game working at all on Machine 1. Could the issue be related to clock speed? I don't really want to undo my overclock, since it took many hours to perfect...

View this comment - View article - View full comments

Quoting: edddeduckferalWe have a Fiji card here (R9 Nano) it's pretty much the same hardware as the Fury X. We retested just now using running Mesa 13.0.2 (using the Padoka stable PPA Paulo Dias created for us the other day) our distro used is Ubuntu 16.10.

Everything runs perfectly with no issues on the latest release, we've also not had other reports. I suspect this might be related to something else linked to your system given stretch is the testing distribution it's completely possible some other related library in your distro is causing side effects. It's one of the reasons why we can't support rolling releases / testing distributions as things are in a state of constant flux.

Hope this helps you debug your issues further. Do let us know if you find anything we should know about.

Okay... by now I've spent way more hours trying to get to the bottom of this than I care to admit.

I started by ruling hardware issues out of the picture. I have a second PC running an AMD R9 285 TONGA. Still GCN hardware supported by AMDGPU, so the software stack can stay the same. Then I installed Ubuntu 16.10, and the game ran fine. Next I compiled the latest Mesa 13.0.2, and the game failed. Okay, must be one of my build options then! But I think they are reasonable, and they used to work (and indeed do still seem to work with all my other games - I tested Dirt Showdown, glxgears, looked carefully at the glxinfo output of both the x86 and amd64 builds, etc. - everything seemed great), so I wondered if tracking down the commit that introduced the regression could help explain the issue. So started my way down that path.

After 12 or so rebuilds of Mesa (running tests on each build), git bisect eventually narrowed down the problem to... [d9546b0c5d1a5136a92276cdd7c14883f0c62737] i965: Integrate precise trig into configuration infrastructure. WTF?! That didn't look right. Just to be sure, I reverted that commit on top of 13.0.2, and the issue remained. Darn it!

So then I got thinking, the issue might be intermittent? So I re-built and re-tested some of the builds that I had previously marked off as bad, and to my surprise they now appeared to be working. I tested some more builds from around that commit, but with the randomness it was hard to come to any conclusions. It seemed once it stopped working, the game wouldn't start working easily again for the next few minutes. So strange!

Then I figured I'd tackle the problem by comparing my compile flags to the ones used in the Padoka repo. I set up a pdebuild environment, but it seems the builds are not produced in a perfectly clean environment since that always produced a linking error... (I later noticed it would only seem to build with dpkg-buildpackage) however, at least I was able to get the configure flags from the build log file, and I updated my build scripts to match - adjusting the paths to /opt/xorg which I also reference via my /etc/ld.so.conf.d/custom-mesa.conf and /etc/profile.d/custom_xorg.sh configs to enable a clean uninstall (well, I do have dpkg-diverts for /usr/lib/x86_64-linux-gnu/dri and /usr/lib/i386-linux-gnu/dri references too).

So built that, ran the game and... fail! Gah! It's supposed to be the same software! Sure the Padoka has four extra patches (in the source debian/patches/ directory), but those don't look related to radeonsi on x86/amd64 at all.

After some more messing around, eventually it worked - and I was at a loss to explain why. Eventually I stumbled upon the suspicion that it was the way I launched Steam. Sometimes I would just type the "steam' command into the terminal, and sometimes I would just click the Steam icon in my Xfce panel. Maybe Steam was inheriting different environment variables based on where it was launched, which impacted the ability to run? It seemed unlikely, but at this point I was running short on ideas.

So I added "env > ~/ENV ; %command%" to the game launch options, launched Steam from the Xfce panel, launched the game to the menu, quit the game and Steam, moved ~/ENV out of the way and repeated by launching Steam from the command line. Then I looked at the two ENV files I had via the Meld tool, and noticed four obvious additions to the command line run which I added to the launcher options.

ORBIT_SOCKETDIR=/tmp/orbit-abolte COLORTERM=gnome-terminal TERM=xterm IBUS_DISABLE_SNOOPER=1 %command%

I then launched Steam from the Xfce panel, and bam.. the game ran fine! Next I just needed to narrow down which of the environment variables caused the problem.

I tried "ORBIT_SOCKETDIR=/tmp/orbit-abolte IBUS_DISABLE_SNOOPER=1 %command%", and that failed. Then I tried "COLORTERM=gnome-terminal TERM=xterm %command%", and that worked! Then I tried "TERM=xterm %command%" and that worked too! Really?

Just to put to rest any doubts, I used "%command%". That worked too. Uhh...

Next, I left the launch options entirely empty, and the game continued to launch fine.

I also messed with enabling and disabling the Xfce compositor and relaunching Steam from the Xfce panel, but the results there also seemed random.

Now that I've spent my entire weekend on this and not gotten anywhere, I think I'm going to go have a cry in the corner. I'm not even convinced the issue is Mesa-related anymore. There's a ton of ALSA warnings, so maybe it's something related to that. At least I now know that if it doesn't work, it'll probably work if I quit Steam and try again (and *may* be more likely to work if I log out and in again, or reboot). Or I might just stick to Wine which works 100% all the time. Haven't decided.

Since you haven't seen the problem yet, perhaps there is something about my computers that make the issue more likely to show. Here are the specs, in case it helps:

Machine 1:
Intel i7-6700K (overclocked to 4.6GHz, many hours of prime95 Small FFTs torture testing to prove stability)
G.Skill Trident Z F4-3200C15D-32GTZ 32GB (2x16GB) 3200MHz DDR4
Samsung 950Pro M.2 512Gb (for OS, x2 plus other SSDs unused here)
Western Digital 6Tb HDD (for games, although typically cached in RAM after first execution anyway)
AMD Radeon R9 Fury X (Fiji) (x2 actually, although only one used here)
MSI Z170A Gaming M7
Antec HCP-1300W PSU

Machine 2:
Intel i7-5820K 3.3GHz (no OC)
Corsair CMK8GX4M2B3000C15 8GB (2x4GB) 3000MHz DDR4 running at SPD 2133MHz during testing
Corsair CMFSSD-128GBG2D SSD
AMD Radeon R9 285 (Tonga PRO)
ASRock X99E-ITX/ac
Corsair AX860i PSU

Aside from both being Intel and relatively modern, they aren't really that similar. Machine 1 runs Debian Stretch, and Machine 2 runs Ubuntu 16.10. Both show exactly the same issue. Both also run Xfce.

View this comment - View article - View full comments