What makes Gentoo so great?
it teaches you how linux actually works. lfs is better though
>>55209893
Arch install teaches you more about the inner workings of Linux.
>>55209904
Gentoo is a meme, there is nothing good about it. Every other distribution does Linux better.
>>55209867
it makes you really question your life priorities as you're waiting for binary #4114990 to compile
>>55209939
LOL
>>55209867
Comfy once you set it up, flags allow you to do whatever you want with your system.
>>55209904
Ever compile a kernel in Arch? Most haven't.
>>55209939
True...
>>55209893
No it's not. The main difference between a Gentoo and a LFS install is a lot of manual tedious compiling.
it contains the source for every other linux so that you don't have to download all of them
>>55209926
this comic applies way more to gentoo than to Arch
>>55209867
No nsa/systemd cancer on my gentoo rig.
Stay cucked systemd faggots
>>55209957
You can apply a software patch just by dropping it in a directory. Tell me of that's not comfy.
>>55209904
Gr8 b8 m8
>>55209994
Don't take the bait.
>>55210035
systemd works fine for me though
>>55210035
do you use open source hardware too?
Isn't Lennart Poeterring openly hostile toward openrc and anti-systemd gentoo folks?
>>55210150
I made my own out of transistors from radio shack
It generally works faster. Compiling your own kernel can speed up boot times also. If you compile with more optimization, it will compile longer but ran faster, something other distributions dont do to save compile time.
tl;dr Its faster
>>55209926
At least the passive aggressive IRC channel is spot on.
>>55209932
nice b8 m8
>CFLAGS
>USE FLAGS
>Portage
>Sources
>Devs are great
>Community is great
What's not to like about Gentoo?
enlightenment
>>55210683
The only negative is the compile times
>>55210798
It's a small price to pay. Updates really aren't that bad.
>>55210798
it really is.
I ran an emerge --emptytree @world this morning.
Still going, 8 hours later. 90 packages left.
>>55210798
>try to install gentoo
>make the error of compiling kde
>on an old thinkpad t60
>mfw
>>55211098
Wow, you're either compiling on a toaster or you have a very bloated system because recompiling my entire source tree when I moved to hardened only took about an hour (+/-)
>>55211234
less than 900 packages. X220 with a 2540m.
make -j5, and compiling in tmpfs.
I'm switching to hardened too.
libreoffice is FUCKING RIDICULOUS BY THE WAY. THREE FUCKING HOURS. gcc is another hour. THings take long.
>>55211234
Don't think i didnt see that
>>55210209
By how much? I hear about the speed increase, but no one has ever provided proofs that it is a worthwhile endeavor. 1%,2%?
How much, fucker?
Made a clean setup from a fresh stage3 few days ago (my oldest install has emerge started first in Feb 2005, two others are year or two younger) and holy shit, emerge is incredibly fast. I've stayed on older version (of portage and python libs) and especially on older python versions, so that must be the reason why my older setups are much slower.
>>55209904
Which is total bullshit, most arch linux users haven't seen a Linux kernel config file.
>>55211280
the speed increase in everyday shit probably isn't noticeable. you do have much less shit on your system because not compiling shit you don't use means pulling in less dependencies you don't use.
boot time is wicked-fast though with a small kernel and no initramfs
>>55211299
portage is faster than it once was but it REALLY needs a rewrite in a C-family language. It's a turtle riding on a snail's back through a river of molasses compared to pacman at resolving dependencies.
gentoo has unnessecary profiles and it doesn't have a curses-based menu to let me set options for programs like FreeBSD, you have to do that by hand by adding flags with a text editor
>>55209939
>4 hours compile time on install
>3 hours compile time every week
It's not THAT bad, also it makes the OS run significantly faster because of compile time processor architecture optimization
>>55211261
Why did you j5 instead of the recommended j3?
>>55211361
ok but someone needs to to some math on how long you need to use a system before the 0.5 second speed gains surpass the time lost compiling the program in the first place. I really don't think you're making up that 3 hours a week
>>55209957
Compiled the Linux libre-kernel shit took hours.
>>55211357
what are "unnecessary profiles"?
setting the flags isn't very tedious, you only have to do it once for everything you install, which, ignoring dependencies (which you shouldn't usually touch) is like 20 or 30 packages (typical world file). once it's set, you're good forever.
>>55211397
2 cores = 4 threads. +1 = 5.
>>55211347
>it REALLY needs a rewrite in a C-family language
One of the older Gentoo devs wrote a 'port' of portage in C, but stopped when pretesting showed no significant performance increase in any use scenario. And sorry, the main part that makes portage 'slow' is the tree system, in which language makes very little difference.
But if it REALLY needs it, then how about volunteering to write it?
>>55211357
>I HAVE TO FIGURE OUT A DEPENDENCY AND TYPE!!!!!!! WAAAAAAAAAAAAAAAAAAAAAAAAA!
>>55211357
i like BSDs but the ncurses menus are seriously fucking annoying
>>55211460
it could at least use multi-threading.
>>55211347
Not trying to be a dick, but I always get this same bullshit answer. I don't believe you, I think it's purely theoretical, something people respond with because that's what they have been told, and they just go with it because they dont have many reasons for Gentoo, they just want to use it because. And that's cool.
>>55211474
Why would I willingly use a distro that makes me do that
>>55211423
That's not at all what the manual/wiki recommends. That's why I'm asking.
>>55211523
You obviously don't have the need, or are just not intelligent enough for complex problems.
>>55211510
There is zero benefit, IO limit comes before there is any use out of multithreading with querying. And of course the compiling itself is multithreaded; make will eat as many cores or threads your system (or systems) have to offer.
>>55211511
What don't you believe?
That I can control what gets built into my binaries?
That this removes the need for dependencies only pulled in for use cases that I'll never have?
>>55211536
"A good choice is the number of CPUs (or CPU cores) in the system plus one"
Each thread is seen as a "cpu" (ex: htop lists 4 cpus). 4 + 1 is 5.
>>55211402
You of course can't compare responsiveness to compile time when you aren't even using your computer. They are different things.
>>55211576
then devise a system that doesn't use as much I/O.
Use a text file instead of recursing through directories. I don't know, I haven't looked into how its put together. All I do know, as a user, is that it's ridiculous to stare at it while it resolves dependencies for 20 seconds before even installing anything.
I totally understand the compile times, and I understand that that takes long, but resolving dependencies SHOULD NOT.
Every other package manager can do it in a fraction of a second, why not portage?
>>55211261
It's only two real cores bro, you're killing it with context switching. Instead you should try something like:MAKEOPTS="-j4 -l2.1"
>>55211787
What does -12.1 do?
>>55211261
>libreoffice
libreoffice-bin exists, use that so you don't have to waste a painful 3 hours...
>>55209867
Denial.
>>55211880
conflicts with packages from ~
>>55210141
>it works so it is instantly good
nice. i bet you use microsoft word through wine inside of virtualbox as a text editor because it works too.
>>55211573
So not making things harder on myself makes me stupid?
>>55211808
-l2.1, l as in the lower case letter L. That full line tells make to run up to 4 jobs and not to launch any new jobs if there is at least one job running and the load average (the -l bit) is above 2.1 i.e it'll launch new jobs only when it makes sense to do so for HT'd two core. The load average could be pumped up a notch to something like 2.5 if it was just one build going on, but I've also got emerge with its own jobs settingEMERGE_DEFAULT_OPTS="--jobs=2 --load-average=2.3 --quiet-build=y"
so the end result is that there is either two makes that each get to use 1-2 jobs, or one make with 3-4. And due to concurrency of the two packages emerging at once it's less likely that processors are waiting for i/o.
>>55211717
>>55211460
I've had my portage tree on a sparse file for nearly a decade now, there's a guide on some old wiki on how to do it for you youngsters. The speed difference from reading it all from one file vs hundreds of tiny files is not even funny.
t. old gentoo fart
>>55212852
I don't do any of that. Nice try though.
>>55209904
The Arch install involves downloading a bunch of binary packages via a single simple shell script.
Arch could be installed via a GUI in just as "customizable" a way, given that despite all of the claims of "digging into the OS to customize it", every Arch install is the exact same.
I started intoing linux with arch because of all of the memes on here, and I thought it was pretty comfy. Then I realized most "hur dur noobuntu" users knew more than I did. Needless to say, I'm not on Arch anymore.
Arch is not hard and Arch is not anything special. It's just a meme OS with a subpar package manager and an install process that makes people who don't know anything about linux feel like hackers.
>inb4 "muh aur"
Most distros just have a big package repository. The way Arch does it is that they have a small repository and then they have an even smaller mostly-unmaintained repository full of four to five year old glorified installation shell scripts.
>>55212883
Just because something works doesn't mean it's instantly good.
If I went and made a text editor that took a half hour to render a character, it would technically work. That wouldn't make it good.
>>55212903
>Page cannot be displayed due to robots.txt.
http://wikigentoo.ksiezyc.pl/TIP_Speeding_up_portage.htm
>>55213039
http://wikigentoo.ksiezyc.pl/TIP_Speeding_up_portage.htm
how about this one
>>55211280
Here for example, a Raspberry Pi GPIO library compiled with more optimization gives more speed.
"Without any optimizations, I got an excellent 14 MHz square wave. Adding -O3 to the compile command (gcc -O3 strobe.c -o strobe) increases the rate to hefty 22 MHz."
Same counts for kernels, programs etc.
>>55209867
Install FreeBSD.
>>55209867
Nothing.
>>55209867
The memes.
>>55213143
I would install a BSD if any of them had support for anything at all.
It's sad. I really like the BSD license when compared to the meme GPL, but Linux won. Get over it.
>>55213289
No, you're just inferior.
>>55212878
Can sparse files work on btrfs?
>>55213143
>>55213200
>>55213285
>newfags have an opinion too
>>55212878
You could also move to the git tree.
I get syncs of like 2-3 seconds with the git tree. It's incredible.
>>55209904
Gentoo is amazing. I can have multiple conflicting versions of packages, or 64 and 32 bit libraries at once. And it's all fine because the programs that depend on specific architectures or specific versions of programs are compiled from source, so portage just links them with the version it knows it depends on.
See https://devmanual.gentoo.org/general-concepts/slotting/ and https://www.gentoo.org/support/news-items/2015-03-28-true-multilib.html
If you've ever had to deal with dependency issues with packages you'll know why this is awesome.
Also being able to manually include (or exclude) certain features from any package or your kernel is extremely useful. For example, I was able to simply remove the SSL heartbeat extension and recompile my packages when the heartbleed vuln came out.
Did I even mention Portage?
>Portage works without any external repo.
>Portage supports using llvm icc etc to build with.
>Portage supports distcc.
>Portage supports slotting of dependencies.(multiple versions of python ruby gtk etc)
>Portage supports multiple kernels BSD Fedora debian etc.
>Portage can thread package installs and downloads
Meanwhile, with Arch
>overzealous autistic fan boys
>"you'll learn how Linux REALLY works!" When it's literally just configuring a package manager and letting scripts do the rest
>offers nothing that minimal net installs already offered for other distros don't
>muh bleeding edge packages!! when you can just install directly from the upstream source in any distro
>only reason to use it is the aur, which is full of broken and unmaintained packages and isn't monitored at all, most "packages" are just a bash script to download the package and it's install script from GitHub
>aur is far worse than Open Build Service, which actually lets you package binaries and programs for multiple distros
It's not the worst distro, but there's nothing it offers that makes it worth using over any other distros and it has the worst fucking user base.
Just a reminder that compiling from source, persistent compiler flags and building you're own kernel are not the exclusive domain of gentoo.
>>55213334
yes
>>55213322
Nice argument.
>>55213387
Avoiding COW?
>>55213356
Thank you. This is the post I was about to make.
Whenever you back an Archfriend into a corner they just reply "AUR AUR AUR AUR" when the AUR is nothing special.
>>55213401
yes
>>55213109
You should really try ffast-math then
>>55213347
That's just for syncing?
>>55213356
>Also being able to manually include (or exclude) certain features from any package or your kernel is extremely useful. For example, I was able to simply remove the SSL heartbeat extension and recompile my packages when the heartbleed vuln came out.
The USE flag system is a bit limited desu, some packages include a huge amount of configuration options that can't really be accessed unless you write your own ebuild for it, same goes with applying custom patches. Imo a better system would be to just ditch the whole global USE flags all together and have the portage pick sane default build options based on your chosen profile for most packages, and when it comes down to those rare few packages you actually want to tweak then that could be done with a more powerful setup tool that would facilitate full access to all make [menu]config options that that particular package has, functionality for applying patches and custom cflags (that last one can already be done via package.env)
>>55213465
Thanks anon.Before
du -sh . 0.31s user 3.42s system 20% cpu 17.872 total
find . > /dev/null 0.30s user 0.44s system 95% cpu 0.772 total
1.2G
After
du -sh . 0.36s user 0.67s system 92% cpu 1.118 total
find . > /dev/null 0.29s user 0.37s system 99% cpu 0.671 total
842M
tmpfs
du -sh . 0.02s user 1.42s system 92% cpu 1.571 total
find . > /dev/null 0.64s user 0.76s system 98% cpu 1.428 total
>>55214062
>portage pick sane default build options based on your chosen profile for most package
This sounds like what the current system allows.
>that would facilitate full access to all make [menu]config options that that particular package
As in an ncurses wrapper around editing package.use?
>>55209867
loosing your time compiling all your packages and have the same performance as other distributions isn't great.
>>55209867
>configure gentoo kernel
>create initramfs file
>it wont boot some shit about fsck and /usr
>fuck\ it.webm
>emerge genkernel
>>55214297
>This sounds like what the current system allows.
That functionality is exactly like it now is, that part of portage behaviour is fine, but
>As in an ncurses wrapper around editing package.use?
The normal global USE flags don't perfectly translate to what configuration options are available in some packages, in those cases there usually are some package specific local USE flag that try to cover most bases, but since it's rather cumbersome to set them up like that there are edge cases that are left out. The biggest functional improvement however would come from the easier application of custom patches, right now there is no other way than writing your own ebuild to do it.
And yes, I am aware that portage as it is today is pretty awesome tool and completely changing it would hardly be justified for the marginal benefits, but the point I was kinda ranting about was that portage is not total freedom to customize packages however you like, it's total freedom to customizing packages as long as you stay inside the box that the .ebuild restricts you in. If what I want is outside that box then I need to make my own box.
>>55214662
Congrats on thinking for yourself. This community needs more hackers like you.
>>55209867
Nothing really
any way I could have portage take source code outside of the package tree and integrate it into the package manager and dependency resolver?
>>55214801
Yes: https://devmanual.gentoo.org/ebuild-writing/
>>55212866
Dude, it makes it easier. That's the fucking point.
>>55211578
Well yeah, it's sort of a panacea. the -j flag is actually for your RAM.
>>55214662
Don't use fucking genkernel. Just find out wtf you did wrong.
So far the only people I know who hate Gentoo are people who fucked it up themselves and blamed Gentoo.
>>55215508
Not him, but why is everyone on /g/ so against genkernel? I used it when I was running gentoo/funtoo and it worked great. What's the issue?
>>55212878
Cute penguin!
>>55215577
>What's the issue?
It's normie shit.
>>55215577
Oh, it's pretty good for the initramfs but I can't recommend it for the actual kernel configuration.
>>55215816
Wow that's a great argument it is so convincing, how didn't I see the obvious?
Any 'install gentoo for fuckheads' tutorials? I tried to boot it from an usb stick but it didn't work/am too retarded to make it work
>>55216156
My SSD wouldn't recognize the Gentoo live medium and so I used SystemRescueCD for the install
>>55216211
I usually use linux mint live cd, so I can have a gui until my entire install finishes compiling.
the gentoo livecd is silly.
>>55209867
It contains all the fags
>>55216245
SystemRescueCD does have a GUI... It's definitely the best live medium and it's based on Gentoo
The only reason to use gentoo is if you want your OS to be highly tailored to your specific machine. Not implying that if you use it without that reason that you're stupid. Just that the tedium of the install + environment isn't worth what you'd get out if it.
>>55209867
Retard Roundup.
>>55212852
That's exactly what I think when people shill systemd.
>>55209867
the meme
gentoo has an ugly LOGO. Very Ugly.
>>55215508
nothing wrong with genkernel, after making your own kernel few times it gets boring and then just use genkernel because its so easy.
>>55215577
genkernel never works for me and generate bloated config that takes fucking long time to compile
>>55221309
sure it can take a long time if you just let it compile without turning stuff off --menuconfig :)
>>55221352
>pass menuconfig to genkernel
Then why even use genkernel?
make nconfig is better though
>>55221377
all my gentoo installs are automated since hardware i run is all the same on servers, i just pass genkernel my config and let it do magic and it reboot and everything is installed the way i want :)
>>55209867
install gentoo
