A lot of pain, bugs and dissapointment in an attempt to get a functional desktop

[John Marino (DragonFly)]

Okay, so there is a lot to digest here and a lot of misinformation.  I'm
not going to address all of it, only things relating to Ravenports.

> Would it be possible to compile ravenports with clang, flang and
gnat-llvm? The only reason I think it will fail is, again, a lack of
gnat for gnat-llvm. Also,
https://sourceforge.net/projects/hacadacompiler/ might be of use to
you... if you have an ada compiler. It's under an MIT license and might
work too, I guess.

Where is this question coming from?
By design, Ravenports has a specific ports compiler for all platforms,
which is currently gcc 11.2.0.  This includes gnat and fortran.  The ports
compiler is a key design feature of Ravenports, so the question above
doesn't make a lot of sense to me.  Some ports do use clang though, but
it's used a port dependency in those specific cases.

> The docs are not finished, I get it, but the help command + quickstart
guide were sufficient to figure out basic usage.
1. The last pre-built image I was able to find for ravenadm is for
Dragonfly 5.8.
2. When I try to compile any package with ravenadm build, only the error
message about needing to rebuild ravenadm or update the ravenports
package pops up.
3. So I decided to build ravenports from source. And... I can't do that
because gnat is not available in repos. I am stuck. I have to compile
gnat, but since gnat requires an older version of gnat to work, I can't
compile it.

So 6.2 packages were produced recently, but I reviewed the quickstart guide
and realized that there is no way for somebody to realize that.  I inserted
a new step,
https://github.com/Ravenports/Ravenports/wiki/quickstart-df#quickstep-3-determine-latest-available-repository
, to show how to determine the latest repository and how to set ravensw to
use it.
Ravenports is essentially a binary repository and there's almost no reason
to build from source.
The only real reason is to use some non-default build options.    All the
questions above are based on some misunderstanding.  Hopefully my wiki
addition helps clear that up.

John



On Mon, Nov 21, 2022 at 11:09 AM <unix-enjoyer at tilde.institute> wrote:

> Hello. I decided to try out DragonflyBSD as a desktop system, and had a
> lot of problems. If you want to know why I decided to try Dragonfly
> specifically, read the next paragraph, otherwise, skip.
> Windows is not an answer for me because the last reasonable version was
> 7. After that, the design was unbearable. 7 is now unsupported by most
> software. Besides, I'm not willing to fight proprietary software anymore
> after I learned about Linux. Linux distributions also have a share of
> design problems, which led to me switching from Linux Mint to Debian,
> Garuda and finally Void. What happened in all cases: my programs just
> broke after unfortunate package management. I wasn't surprised since I
> knew that most of this code was not written to be good and stable. It
> was to get paid by Redhat/Google/Microsoft/any other corporate Linux
> contributor. By that point I was sick of the same corporations that
> control Windows controlling my Linux boxes and decided to trade Wine and
> inconsistent Linux interfaces for security, cli minimalism, stability
> and documentation of OpenBSD. It works just fine. I was a little sad
> that virtualisation is limited to single-core with serial console.
> FreeBSD also looked like an interesting alternative, though I have heard
> horror stories about desktop usage + I'm paranoid about OpenSSL
> security, in light of Heartbleed and this recent bug which also affects
> everybody. Once I learned more about Dragonfly, I figured that this
> might just be a perfect fit for my usecase. I live in an area where
> short power outages just happen, and you have to cope with it. Hence why
> I use an additional battery-powered power supply just in case. However,
> my main PC power supply  sometimes fails, soon to be replaced. The
> usefulness of CoW hammer2 (or ZFS/btrfs/bcachefs on Linux) is obvious.
> Program snapshots would allow me to put simple programs like mupdf to
> disk to easily continue my reading/watching/editing session when I want
> to. I tend to prefer very basic programs after all of the bad experience
> I had with complicated ones. Virtualisation will allow me to play some
> of my favourite games, I'm not interested in AAA. A little sad that Wine
> isn't ported yet, but not a big deal. LibreSSL in base provides at least
> some network security, even though ports still use OpenSSL (surprised
> that ravenports have an option of compiling everything with LibreSSL).
> So, what's up with the title?
> I have two devices I tried Dragonfly on. Both attempts failed to meet my
> requirements of being a stable desktop. A reasonably good browser that
> doesn't spy on me like chromium or default firefox, mpv video player, a
> terminal emulator, and a vis text editor is all I need for my everyday
> existense. I prefer vis to (neo)vim or regular vi because it's just much
> less complicated for me to use. Structular regular expressions are more
> comfortable than macros for me. Anyway, let's start with the first
> attempt, which I assumed to be a success at first.
> Dell Inspiron 3542 with 2 core Haswell Pentium 1.4 GHz, 4GB RAM,
> 1366x768 IPS screen (replacement for the old dead one), 128 GB SSD in
> the main slot + 512 GB HDD on rails in the CD slot. No dGPU, dead
> battery.
> The installation went ok. Although the installer is not very flexible,
> e.g. I had to mount my HDD in my home directory after installation, but
> that's a non-issue for me.
> 1. Docs (https://www.dragonflybsd.org/docs/handbook/X/) don't explain
> that I need to install and enable appropriate graphics drivers in
> /etc/rc.conf (the "outdated" page does half of the job, explaining the
> need for installing drivers, but failing to mention rc config
> https://www.dragonflybsd.org/docs/how_to_get_to_the_desktop/#index7h3).
> They also don't explain how to enable mouse, and that my user should be
> in the video group. They don't point me to the appropriate manpages or
> FreeBSD documentation.
> 2. At least both the touchpad and the built-in (PS/2?) keyboard worked.
> However, USB mouse & keyboard didn't work in Xorg regardless of moused,
> devd, or udevd being enabled. In console, USB keyboard works just fine,
> however.
> 3. Xorg has a long startup (about 5-8 seconds). I stare on a black
> screen for a while instead of doing something. I still have no idea why
> this is the case. I see no error messages in logs.
> 4. I can't switch to the console with the usual Ctrl+Alt+n, where n is
> the tty I wish to switch to.
> 5. Qutebrowser starts slowing down after one or two restarts, and
> crashes randomly.
> 6. Netsurf browser crashes when I go into Edit/Settings.
> 7. Otter browser crashes for unknown reasons.
> 8. xterm starts slowing down after restarts.
> 9. htop reports wrong cpu usage. Over 9000% wrong, actually. It was
> pretty amusing to watch, but it doesn't provide any useful information
> now that the percentages are screwed.
> 10. Something I don't understand in BSD's is why the vis(1) program
> needs to be on my hard drive and not in ports. Not a big deal, it's one
> rm command away from being fixed for me.
> 11. So I tried to compile vis-the-editor
> (https://github.com/martanne/vis) from source. It failed to find curses
> even though ncurses was explicitly installed. So I built it without
> curses. I receive a lot of epileptic flashing when I edit anything in
> xterm, but overall the program works ok.
> After declaring the initial testing on this laptop a relative success
> (at the time I thought I fixed qutebrowser by fiddling with it's config
> file), I decided to install Dragonfly on my main PC: Intel Ivybridge
> Xeon, 4 cores/8 threads, 12 GB RAM, Radeon RX 580, 512 GB SSD + 1 TB
> HDD, 2 1920x1080 displays.
> 1. In console, USB keyboard worked fine. When I later tried my PS/2
> keyboard, it could only produce some weird sequences (^A, >, etc.) and
> nothing more.
> 2. Xorg said no. No matter what input device I try, nothing can be used
> in Xorg. At first I thought that it just hangs my GPU, but no, that
> wasn't the case. Tried cwm and wmii, and noticed that in wmii, the panel
> still displays stats and time in realtime, so that's how I know.
> 3. Wayland is unusable on both devices. It stays there (just like Xorg
> the first few seconds), but can't start due to some DRM error I can't
> catch because it doesn't go into stderr, so redirecting it to a file is
> not possible. I suppose I could record the process, but I have to have a
> stable desktop to do it first, right? Tried hikari and wayfire.
> 4. Arcan is outdated and also unusable on both devices. The issue seems
> similar to Wayland. I would prefer to use Arcan over Wayland & Xorg if
> it worked. It looks like future technology to me. I think embracing the
> project might even help with some of your code bounties. Later about
> that.
> After experiencing these problems (which forced me to poweroff with a
> button on my motherboard an uncomfortable amount of times) I decided to
> compile my own kernel & packages from dports.
> Kernel upgrade worked fine, and changed nothing. I can post my config if
> you're interested.
> 1. When compiling Dports in console, packages have broken menus that
> hide options. You have to press arrow keys to inspect all of the options
> because they get crammed in one tiny spot, and long descriptions can't
> be read. Sometimes the issue resolves itself and packages display their
> options properly, but most of the time it's broken and unreadable.
> 2. Searching for a particular port proved tedious. On OpenBSD, it was as
> simple as "make search".
> 3. No matter what I tried to compile, be it xorg, wayland or arcan, I
> had to rmconfig-recursive because something broke. Is there a better way
> of handling my unfortunate choices? Reconfiguring my stuff every time is
> very tedious and a lot of packages tend to break.
> With no luck in dports, I tried ravenports.
> The docs are not finished, I get it, but the help command + quickstart
> guide were sufficient to figure out basic usage.
> 1. The last pre-built image I was able to find for ravenadm is for
> Dragonfly 5.8.
> 2. When I try to compile any package with ravenadm build, only the error
> message about needing to rebuild ravenadm or update the ravenports
> package pops up.
> 3. So I decided to build ravenports from source. And... I can't do that
> because gnat is not available in repos. I am stuck. I have to compile
> gnat, but since gnat requires an older version of gnat to work, I can't
> compile it.
> I suspect that building stuff with llvm might resolve some of my
> compilation problems, a more modern compiler ~ better compatibility,
> especially with freebsd ports which are compiled with clang after all.
> After getting angry at constantly crashing browsers, extremely slow
> xterm and epileptic editor, I switched my laptop back to OpenBSD and
> currently write this email on it.
> I have some questions:
> 1. Why an outdated gcc instead of a recent llvm in base? gcc 8.5 was
> discontinued a year ago, and you use 8.3. llvm is not that big of an
> improvement in terms of code size (still millions of loc), but it is an
> improvement in design and licensing compared to historical inconsistency
> of gcc. I suspect that at least some of my problems come from outdated
> toolchain which you use for unknown reasons.
> 2. Would it be possible to compile ravenports with clang, flang and
> gnat-llvm? The only reason I think it will fail is, again, a lack of
> gnat for gnat-llvm. Also,
> https://sourceforge.net/projects/hacadacompiler/ might be of use to
> you... if you have an ada compiler. It's under an MIT license and might
> work too, I guess.
> 3. Is tcplay dead or just resting? Last update on github was 2.5 years
> ago. I would be interested in TrueCrypt compatibility mentioned on your
> website, some of those crypto algorithms seem to hold up well enough for
> decades.
> 4. nouveau port anytime soon? I don't know why only NetBSD bothered to
> do it out of all BSD's, users of old-ish nvidia cards are still common
> in desktop space, at least.
> 5. tickless kernel anytime soon? energy efficiency suffers, performance
> in virtualisation also suffers.
> While waiting for my -j1 llvm15 compilation to fail, I skimmed through
> your code bounties, and have some suggestions:
> 1. usb webcam support & utf8 in console
> both of those are solved in Arcan. See Arcan console
> (
> https://arcan-fe.com/2018/10/31/walkthrough-writing-a-kmscon-console-like-window-manager-using-arcan/
> )
> and libuvc, Arcan's dependency (https://github.com/libuvc/libuvc).
> Durden (http://durden.arcan-fe.com/) is able to emulate the behaviour of
> most traditional desktop environments. Pipeworld
> (https://arcan-fe.com/2021/04/12/introducing-pipeworld/) is a
> spreadsheet-shell-like zoomable window manager. Safespaces
> (https://arcan-fe.com/2018/03/29/safespaces-an-open-source-vr-desktop/)
> is a 3D & VR environment. These are lua scripts wrapped around an engine
> developed for 15+ years. Letoram (main developer) is very interested in
> userspace drivers, and already implemented cross-platform VR,
> streamdeck(
> https://arcan-fe.com/2019/10/30/interfacing-with-a-stream-deck-device/),
> and USB video drivers. Arcan provides a stable alternative to Xorg and
> Wayland while keeping compatibility with them through arcan-wayland and
> xarcan. What I want to say is that there is an intersection between your
> goals and Arcan:
> - Migration of drivers to userspace
> - Plan 9-like namespaces with your nullfs union mounting & remote
>   hammer2 mounting, Arcan's network transparency, cfgfs & durden
>   filesystem menu
>   - Compatibility through virtualisation with nvmm, arcan's qemu
>     frontend, arcan-wayland, xarcan.
>     An interesting sidenote: the first desktop environment for Arcan was
>     influenced by Amiga's UI (scroll to the bottom of
>     https://arcan-fe.com/videos/). You all know what Dillon was working
>     on in the 80's.
>     Both you and Arcan might benefit from cooperation, and you certainly
>     have similarities.
>
> 2. userspace vkernel on other os's
>     cosmopolitan libc (https://github.com/jart/cosmopolitan) allows C to
>     run on unices & windows. The amount of implemented syscalls is
>     limited, but I think that's the best option you have right now.
>     3. reducing boot time
>     I suggest to move on from single-core synchronious rc boot scripts
>     to an asynchronius service manager: s6-rc
>     (https://www.skarnet.org/software/s6-rc/). It starts services in
>     parallel instead of one after another. It provides dependency
>     management in case some services depend on others. It does not wake
>     up every couple of seconds, so energy efficiency should increase
>     too. Service scripts can be written in any scripting language, so
>     migration shouldn't be a problem for anybody who still uses shell
>     scripts instead of systemd.
>     The rest of skarnet.org software might be of interest to you if you
>     ever needed secure implementations of network utilities,
>     non-interactive alternative to /bin/sh, small portable unix-ish
>     utilities, or a portable systems programming C library that doesn't
>     suck as much as others do.
>     That's all, I wait for help & answers. I can transfer whatever logs
>     you need using USB flash. I would like to have a functional
>     DragonflyBSD desktop on my main PC, but if it's not possible, I will
>     move on with my life.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dragonflybsd.org/pipermail/users/attachments/20221121/3fdbb248/attachment-0001.htm>