|
Suspicious Dish posted:seriously the dbus here is just a serialization protocol. it was standardized before msgpack and if there was a json parser in pid1 you would be complaining about that too dbus is just Mach messaging reimplemented poorly outside the kernel by people who don’t know their history like you can certainly do a better job than mach_port_t but not if you don’t actually understand it (the top of the list would be to make their refcounting sane, Mach port leaks suck and are very easy to write, but I’m sure people like nbsd will show up to say “yeah make it not slow!!!!” showing they don’t understand the problem space at all)
|
# ? May 5, 2020 22:36 |
|
|
# ? Jun 4, 2024 00:49 |
|
Sapozhnik posted:Next what we need is some way to create a new process ex nihilo and load memory mappings, threads, seccomp policy, and fds into it. posix_spawnattr_t is your friend here it’s no mach_task_t but you can turn it into that if you want just by piling on the extensions
|
# ? May 5, 2020 22:40 |
|
oh right there’s also posix_spawn_file_actions_t just to make things more confusing it does suck that all the sane extensions are currently _np though like specifying what file descriptors to preserve across the spawn rather than close (since the latter is racy in any multithreaded process, which is virtually every process) or specifying what the new process’s initial working directory should be once you have enough stuff atop it though posix_spawn is a huge boon to things like build systems for large projects because you don’t need eg trampoline tools to reduce intermediate process image overhead eschaton fucked around with this message at 22:58 on May 5, 2020 |
# ? May 5, 2020 22:51 |
|
1989 is the year of mach on the desktop
|
# ? May 5, 2020 22:52 |
|
pram posted:1989 is the year of mach on the desktop 31 years later I think we can call it a mild success
|
# ? May 5, 2020 23:14 |
|
eschaton posted:dbus is just Mach messaging reimplemented poorly outside the kernel by people who don’t know their history that's not what any of this lol
|
# ? May 5, 2020 23:36 |
|
The Aérospatiale/BAC Concorde (/ˈkɒŋkɔːrd/) is a British–French turbojet-powered supersonic passenger airliner that was operated until 2003. It had a maximum speed over twice the speed of sound, at Mach 2.04
|
# ? May 6, 2020 00:34 |
|
eschaton posted:dbus is just Mach messaging reimplemented poorly outside the kernel by people who don’t know their history patches welcome
|
# ? May 6, 2020 02:27 |
|
ratbert90 posted:? Python works fine in the embedded Linux world. Why wouldn't Python work? same, I'm shipping python on my next project It's the only way to get free from the monstrous C++ tarpit that our last *~ArChItEcTz~* created
|
# ? May 6, 2020 04:37 |
|
Poopernickel posted:same, I'm shipping python on my next project Seriously, it's so drat awesome. As a side note, here are all the options firewalld requires in the kernel: https://patchwork.ozlabs.org/project/buildroot/patch/20200507231457.2093052-4-aduskett@gmail.com/
|
# ? May 8, 2020 00:39 |
|
so what you're saying is that firewalld requires a firewall?
|
# ? May 8, 2020 01:02 |
|
pseudorandom name posted:so what you're saying is that firewalld requires a firewall? firewalld uses iptables for some of the passthrough stuff. Masquerading wasn’t supported until just recently.
|
# ? May 8, 2020 04:43 |
|
firewalld is a way to write firewall policy declaratively vs iptables,which is more procedural
|
# ? May 8, 2020 06:03 |
|
iptables is possibly the worst interface for a firewall i have ever encountered firewalld also sucks but sucks differently really i just want asa or junos or screenos gently caress this god damned linux poo poo
|
# ? May 8, 2020 18:40 |
|
Notorious b.s.d. posted:iptables is possibly the worst interface for a firewall i have ever encountered What's wrong with Firewalld's interface?
|
# ? May 8, 2020 18:42 |
|
ratbert90 posted:What's wrong with Firewalld's interface? 1. gently caress polkit forever 2. the command line is dogshit and very obviously meant not to be used basically firewalld is a desktop firewall and it shows. they never cared about a use case other than gnome
|
# ? May 8, 2020 19:53 |
|
firewall-cmd is too long to type and i have to look at the man pages every time to distinguish get- commands from the list- ones.
|
# ? May 8, 2020 19:56 |
|
whatever happened to nftables
|
# ? May 8, 2020 20:10 |
|
Sapozhnik posted:whatever happened to nftables https://fedoraproject.org/wiki/Changes/firewalld_default_to_nftables should be working right now
|
# ? May 8, 2020 20:47 |
|
iptables was a direct interface to the kernel and is very good at what it does... a very Linux solution not to say that it is good, but it is adequate firewalld has a learning curve but the command line is not terrible. I also have to look up the command line options as well. Now that you have many services jockeying for control over the firewall (from fail2ban to libvirt) it makes more sense to have something like this that everyone can interface with in a safer way. The alternative is risking stepping on someone's feet. ufw seems the most user friendly, however
|
# ? May 8, 2020 21:18 |
|
who remembers halted firewalls lmao
|
# ? May 8, 2020 21:24 |
|
is there an easy way to give a nonroot user start/stop access over a global systemd unit without sudo? I'm working on an embedded product and want to give an unprivileged user the ability to mount/unmount one partition
|
# ? May 8, 2020 23:16 |
|
Does it have to involve a global systemd unit? You can do that with udisksctl and polkit
|
# ? May 8, 2020 23:29 |
|
polkit is a pretty heavyweight dependency that I'd rather not pull in - got anything else that could do the job? it doesn't need to be a global systemd unit - I just thought that would be an easy way to set the mount up with careful control
|
# ? May 8, 2020 23:34 |
|
3 line C program that calls mount for you and make it setuid
|
# ? May 8, 2020 23:37 |
|
Poopernickel posted:is there an easy way to give a nonroot user start/stop access over a global systemd unit without sudo? if you put a user entry in fstab, unprivileged users can mount it here, under non-superuser mounts. you can similarly configure fstab for group access by adding a group to the system, chown the mount point appropriately, and restrict access to that one user (or group of users) by adding whoever needs permission to it to that group. iirc you may have to specify it in fstab by gid rather than by group name
|
# ? May 8, 2020 23:39 |
|
a Facebook engineer:Behdad Esfahbod posted:That's what we decided not to support anymore. You can try to get used to the "blurry" rendering. Or you can find something else to use, or whatever. You are using Free Software you realize, right?
|
# ? May 9, 2020 20:55 |
|
MrMoo posted:a Facebook engineer: I feel that it's more fair to say "a HarfBuzz developer" rather than "a Facebook engineer" dunno what Facebook has to do with Harfbuzz... I don't get the impression that they have any business developing a component in the text drawing pipeline
|
# ? May 9, 2020 21:24 |
|
after the enabling a genocide or two because they didn’t have good south asia font support they probably hired him to address that
|
# ? May 9, 2020 21:36 |
|
el dorito posted:I feel that it's more fair to say "a HarfBuzz developer" rather than "a Facebook engineer" Don’t need to be fair with Facebook engineering, readers are more likely to be aware of Facebook than HarfBuzz. It also importantly conveys that the developer is actually being paid to be a dick, most open source developers are usually not so lovely* * ha ha ha. quote:dunno what Facebook has to do with Harfbuzz... I get the impression that they have any business developing a component in the text drawing pipeline I read a while ago that the Facebook mobile apps reimplement the kitchen sink, so that probably means they ship HarfBuzz in their app under the pretense that older platforms have broken layout support or simply incomplete support of the continually updating Unicode spec.
|
# ? May 9, 2020 23:11 |
|
MrMoo posted:Don’t need to be fair with Facebook engineering, readers are more likely to be aware of Facebook than HarfBuzz. It also importantly conveys that the developer is actually being paid to be a dick, most open source developers are usually not so lovely* https://arstechnica.com/gadgets/2020/05/a-mistake-at-facebook-broke-spotify-venmo-tiktok-and-other-iphone-apps/ facebook: "move fast and break things"
|
# ? May 9, 2020 23:33 |
|
quoting something out of context with no link to the discussion really makes me think you have an agenda anyway, here's an explanation of the situation: https://github.com/harfbuzz/harfbuzz/issues/2394#issuecomment-626254448 edit: the tldr is that a bunch of abusive halfwits are getting pissy about the how harfbuzz/pango/cairo/etc. in a non-standard configuration render some fonts that can't legally be used on non-Windows platforms pseudorandom name fucked around with this message at 02:05 on May 10, 2020 |
# ? May 10, 2020 01:53 |
|
the facebook guy sounds like a toxic shithead that should just quit
|
# ? May 10, 2020 02:08 |
|
so it sounds like they removed support for a certain kind of hinting that apparently you can request globally in unix and all the people with “above average visual acuity” are really mad about it
|
# ? May 10, 2020 02:10 |
|
it also seems like part of it is that people are using unsupported system configurations because if you use all of the correct versions of the libraries together theres no problem, but distros just toss in whatever junk they found on the shelf and unsurprisingly it doesnt work
|
# ? May 10, 2020 02:18 |
|
I don't get the point of bragging about your contributions to the open source experience when you poo poo on everything that isn't android. If you're going to acknowledge that corporate funding drives development then how stupid are you to not recognize that when the funding dries up and people complain, that's not your problem? really stupid messianic complex about fonts and emojis of all things
|
# ? May 10, 2020 02:24 |
|
Suspicious Dish posted:it also seems like part of it is that people are using unsupported system configurations because if you use all of the correct versions of the libraries together theres no problem, but distros just toss in whatever junk they found on the shelf and unsurprisingly it doesnt work I think part of the problem is also that the stolen Windows fonts require ClearType and other stuff that FreeType can't/doesn't implement and they look like poo poo unless you crank the autohinter blur up to maximum. hifi posted:If you're going to acknowledge that corporate funding drives development then how stupid are you to not recognize that when the funding dries up and people complain, that's not your problem? really stupid messianic complex about fonts and emojis of all things The problem is that the complainants are entitled teenage boys who know your email address.
|
# ? May 10, 2020 02:28 |
|
heated coding moment
|
# ? May 10, 2020 02:30 |
|
https://gitlab.gnome.org/GNOME/pango/issues/463 It’s a complicated mess, driven by changes in Pango breaking rendering because they want to rely on HarfBuzz which has a half-rear end implementation apparently. Microsoft fonts were pretty important historically because everything relies on them. The legality is a bit silly as MS published them many moons ago, however from that RedHat paid for some type of freedom font that matches the metrics but without the licensing issues. Are those affected too? Meh, the main issue here is the lack of civility and poor handling of the situation. MrMoo fucked around with this message at 02:50 on May 10, 2020 |
# ? May 10, 2020 02:43 |
|
|
# ? Jun 4, 2024 00:49 |
|
MrMoo posted:Meh, the main issue here is the lack of civility and poor handling of the situation. recurring problem in freetype and friends, going all the way back to the xfs-tt vs client side dispute
|
# ? May 10, 2020 02:49 |