Gentoo on My Lappy

Gizmo

Gizmo

Chief Site Administrator
Staff member
So, after talking to Kaitain, I've decided to take the plunge. I've downloaded a stage 3 tarball and I'm installing Gentoo.

Recompiling the world right now. 6 hours in and I'm about 3/4 done, I think (about 50 packages left, started with 231).

I've mounted an 8G ramdisk at /var/tmp/portage. That makes a HUGE difference in compile times. Using -j9 since this is a quad-core hyper-threaded CPU.

Just using -O2 for the compiler, and using -march=native.

Using the 17.0 desktop/gnome profile (I'm planning on installing Cinnamon for the DE).
 
Last edited:
... so I'm getting the blame again :p

A little health warning about -march=native, GCC occasionally makes bizarre decisions about what your architecture actually is, and can introduce instructions you don't have, or miss optimizations that you do have. Probably no major harm if you're using, say, a 6th generation Intel i7, but of some concern if you're using something very recent. Not least because Gentoo's "stable branch" GCC is also quite old.

So for better results, refer to https://wiki.gentoo.org/wiki/Safe_CFLAGS and also https://wiki.gentoo.org/wiki/CPU_FLAGS_X86

Welcome to the family :D
 
I've had to abort this. Getting Gentoo running in a desktop environment has turned out to be more of a challenge than I had anticipated. Right now, Cinnamon is refusing to compile, and I've not got the time to track down why, so I've gone to Fedora for the time being. (This is my work computer, and it is imperative that I have it in some semblance of running this evening.)

I'm going to come back to this when I have more time (and a different computer) to futz with this.
 
Do both. Shrink your /home down a little, build Gentoo in that and use Fedora's bootloader for now. When it works you can always mkfs.ext4 && (mount) && cp -ar * over the top is the Fedora installation
 
You know, I recently saw a report from a guy that was having trouble compiling Cinnamon. I can't remember where, and I can't find it; but I do remember that at least part of his solution was to change:

MAKEOPTS=-j[whatever] to MAKEOPTS=-j[something less]

YMMV, [and you may be aware of this already] but if that happens to work you can make the change permanent by creating a 'package.env' file in /etc/portage, and a /etc/portage/env directory like this: https://wiki.gentoo.org/wiki/Knowledge_Base:Overriding_environment_variables_per_package

I've had to do this for a package or two, for example dev-lang/ghc which always seems to fail if it's not compiled with -j2. If I use -j5, my default, it is a predictable mess. dev-libs/boost was another frequent problem for me, and reducing it to -j2 also solved it permanently.

Another thing you might check is if your ramdisk is big enough; I struggled with getting Chromium compiled for ages until I realized that mine was insufficient for the huge memory demands Chromium needs to compile. In the end, I noticed that it was consuming not just all of the ramdisk, but all of the 1GB swap partition I had [Normally, my machine NEVER swaps out, hence the small partition]. I had to create a 6GB swapfile in addition to the swap partition just to get it done.

I doubt that Cinnamon is that demanding, but it's something to check.
 
Last edited:
ThunderRd said:
You know, I recently saw a report from a guy that was having trouble compiling Cinnamon. I can't remember where, and I can't find it; but I do remember that at least part of his solution was to change:

MAKEOPTS=-j[whatever] to MAKEOPTS=-j[something less]
Click to expand...
That MIGHT actually be the problem. It was hanging at 'cp cinnamon-real cinnamon.tmp && mv cinnamon.tmp cinnamon', as if it was waiting on something.

ThunderRd said:
Another thing you might check is if your ramdisk is big enough; I struggled with getting Chromium compiled for ages until I realized that mine was insufficient for the huge memory demands Chromium needs to compile.
Click to expand...

That was the 1st thing I checked, as I had gcc fail to compile for that reason. I had a 4GB ramdisk set up, bumped it to 8GB and solved that problem. In any case, the ramdisk was only at like 10% of capacity or something.
 
Finally got up and running on Fedora. I had to install Fedora 26, because Fedora 27 IS A HOT STEAMING PILE OF POO!!!

SERIOUSLY

What god-forsaken walnut-brained pre-humanoid amoeba decided that it would be a Good Idea to automatically build the nVidia drivers with an experimental feature (KMS) turned on, when nVidia THEMSELVES indicate that it is EXPERIMENTAL, as in NOT TO BE USED!?!?!

That, coupled with Fedora's infatuation with Wayland, meant that F27 was essentially destined to die as soon as I installed it.

Finally, there appears to be some kind of weird bug in VirtualBox 5.2.6 which causes conferencing apps like Skype to just wig out. You get about 2 seconds of audio, and then nothing: no-one can hear you, and you can't hear anyone else. Audio on the host works fine, and OTHER audio on the guest ALSO works. It's just anything that involves both recording and playing at the same time (or even nearly the same time; using sound recorder and then immediately going to play back the audio file in WMP will do the same thing: WMP doesn't produce any playback after about 3 seconds). Downgrading to the 5.1 series resolves the issue.
 
I hate to tell you, but +kms is default for Gentoo as well, and has been for quite some time:

Code: 
root@QX9650: ~# eix nvidia-drivers
[I] x11-drivers/nvidia-drivers
    Available versions:  304.137(0/304)^md 340.96-r1(0/340)^msd[1] 340.104(0/340)^md 340.104(0/340)^md[1] (~)340.106(0/340)^md 375.82(0/375)^md 378.13-r1(0/378)^md 381.22-r1(0/381)^md 384.90(0/384)^md (~)384.98(0/384)^md (~)384.111(0/384)^md 387.22(0/387)^md (~)387.34(0/387)^md (~)390.12(0/390)^md (~)390.25(0/390)^md {+X acpi compat +driver gtk3 +kms multilib pax_kernel static-libs +tools uvm wayland ABI_MIPS="n32 n64 o32" ABI_PPC="32 64" ABI_S390="32 64" ABI_X86="32 64 x32" KERNEL="FreeBSD linux"}
    Installed versions:  390.25(0/390)^md(11:41:43 PM 01/31/2018)(X driver gtk3 kms multilib static-libs tools -acpi -compat -pax_kernel -uvm -wayland ABI_MIPS="-n32 -n64 -o32" ABI_PPC="-32 -64" ABI_S390="-32 -64" ABI_X86="32 64 -x32" KERNEL="linux -FreeBSD")

https://packages.gentoo.org/packages/x11-drivers/nvidia-drivers

Did you actually have problems because of it? I *think* we've had +kms since 360.xx
 
Last edited:
The rpmfusion repo nvidia drivers install with nvidia-drm.modeset=1, which turns on the nvidia KMS support if the kernel is compiled with CONFIG_DRM and CONFIG_DRM_KERNEL_HELPER (which the F27 kernel is).

When the nvidia drivers are enabled, REGARDLESS of the nvidia-drm.modeset parameter setting, I get to the gdm, login, and get a blank screen. The keyboard doesn't work and there's no mouse. The only way I can successfully boot is with the nouveau drivers, and no disrespect intended to the nouveau folks (because they have a done a truly phenomenal job getting drivers for nVidia working, considering they've had essentially no help from nVidia) but the nouveau drivers SUCK; no 3D support, and the 2D support is slow, particularly on HiDPI displays.

In addition, you pretty much HAVE to disable Wayland with the nVidia drivers (via /etc/gdm/custom.cfg), or GDM becomes unusable and you can't even login.

I DID finally figure out that I could fall back to the 340 nVidia drivers and get some semblance of stability, only to then run into the issue with VirtualBox that I mentioned with the sound. Since at that point I was highly suspicious of F27, I just reformatted and went back to F26. When I found that VirtualBox still had the issue, I realized the problem there had to be with VB, so downgraded to the 5.1 series and solved that problem.

Now that I've figured out more of what's going on, I could probably get F27 working stably, but the whole DRM/KMS/nVidia/Wayland thing is just ENORMOUSLY frustrating. And while disabling Wayland appears to be a doable thing, the configuration of F27 with the proprietary nVidia drivers leaves nVidia owners with a brick BY DEFAULT!

IMO Wayland is nowhere CLOSE to being ready for prime time, and for RedHat to just force it on the user community with it in its current state is plain irresponsible IMNSHO.
 
Is your graphics card pre-Fermi? If it is, you'd HAVE to use 340.xx.

I don't think it's that old, though, although I'm not sure what you're running in there.
 
It's a Quadro K2100M, reports 576 CUDA cores, and is supported by the latest drivers. It's based on the GK106 (Kepler) core, roughly equivalent to a GTX 645.
 
add modprobe.blacklist=nouveau to your kernel command line? I don't know whether it's been fixed, but it used to be that nouveau and nv couldn't coexist on a system without the black screen of nothingness...
 
Kaitain said:
add modprobe.blacklist=nouveau to your kernel command line? I don't know whether it's been fixed, but it used to be that nouveau and nv couldn't coexist on a system without the black screen of nothingness...
Click to expand...

That's still true, nouveau has to be blacklisted or Bad Things happen when trying to use the nVidia drivers.
 
Well, blame's gotta go somewhere, and the French are broad-shouldered and mature enough to pay it the attention it deserves*. Not like some of these other oversensitive cultures ;)


* i.e. none whatsoever
 
Gizmo said:
Finally got up and running on Fedora. I had to install Fedora 26, because Fedora 27 IS A HOT STEAMING PILE OF POO!!!
Click to expand...

Fedora 28 is ALSO a HOT STEAMING PILE OF POO.

I'm on Mint 18.3 now. I'm tired of screwing around with Fedora. I realize that Fedora is basically Redhat's testing environment for stuff the gets put into RHEL, but the implementation of Wayland with the nVidia drivers sucks so hard that I'm pretty sure it could pull petroleum sludge straight up through 200 feet of 1/4" pipe!
 
I assume your going to all this trouble because you can't use Mint...Am I correct?

I was thinking of using Mint on Patti's new lap top (should be getting her one this week)...would this be unwise for me to do?
 
You must log in or register to reply here.
Back
Top