1 |
Hi! |
2 |
|
3 |
On Wed, 22 Aug 2018 20:33:00 +0200 Davyd McColl wrote: |
4 |
> The other day I installed Celestia for the entertainment of my son, who is |
5 |
> delighted with anything stellar / planetary. Celestia wouldn't start up, |
6 |
> and, long-story-short, I tracked down the issue to the symlinks: |
7 |
> |
8 |
> /usr/lib64/libGL.so |
9 |
> /usr/lib64/libGL.so.1 |
10 |
> |
11 |
> which ultimately point to |
12 |
> |
13 |
> /usr/lib64/libGL.so.1.2.0, |
14 |
> |
15 |
> provided by media-libs/mesa. Naturally, I assumed I'd made a mistake with |
16 |
> `eselect` at some point, so I checked with `eselect opengl list` and found |
17 |
> that, as expected, my selected opengl implementation was nvidia. Just in |
18 |
> case, I switched over to xorg-x11 (mesa) and back again, but this didn't |
19 |
> fix the problem. |
20 |
> |
21 |
> Manually redirecting these to /usr/lib64/opengl/nvidia/lib/libGL.so |
22 |
> (provided by x11-drivers/nvidia-drivers) works, however, of course, portage |
23 |
> doesn't know anything about this, so the update I received today for |
24 |
> media-libs/mesa reverted these symlinks back to pointing at mesa libs. |
25 |
> |
26 |
> So the questions I have are these: |
27 |
> 1) Am I reasonable in expecting `eselect opengl` to maintain these |
28 |
> symlinks? I feel like it's a reasonable expectation, but perhaps there's |
29 |
> just yet another thing I have to learn / understand. |
30 |
|
31 |
No, eselect opengl works differently. It uses /etc/env.d to alter |
32 |
LDPATH and OPENGL_PROFILE environment variables. It also changes |
33 |
xorg.conf. |
34 |
|
35 |
So you may need to restart your X server and source /etc/profile in |
36 |
active shells for changes to take effect. |
37 |
|
38 |
> 2) Should I be logging a bug (against eselect, or perhaps celestia, since |
39 |
> this is the only app which seems to have suffered this fate -- games like |
40 |
> Torchlight 2 and utils like glxgears work just fine; glxinfo reports NVIDIA |
41 |
> extensions), or is there just something I've fundamentally missed or messed |
42 |
> up here? |
43 |
|
44 |
If glxinfo reports correct data and glxgears works fine, then this |
45 |
may be a bug and please report it. You may CC both celestia and |
46 |
opengl since right now it is not obvious which is the culprit. |
47 |
|
48 |
Best regards, |
49 |
Andrew Savchenko |