<kof673>
lol i consider those the lite version lol
<kof673>
you could do one with env. vars or whatever else lol
<kof673>
or even backtick-like lol run a command lol
<kof673>
they are typically toned down for security
simjnd has quit [Ping timeout: 248 seconds]
<kof673>
it seems to be a fear of implementation -- a symlink of any kind is not going to magically give any privileges that weren't already there, wherever it points. i think people are just afraid of bugs
<kof673>
or that it is too complex
<kof673>
or that people might use them carelessly
simjnd has joined #osdev
<zid`>
why do I know the name andreas kling
<ddevault>
serenity OS
<kof673>
with a command, you could do silly things like round robin load balance a file, points to 1/5 places lol
simjnd has quit [Ping timeout: 260 seconds]
simjnd has joined #osdev
bleb has quit [Ping timeout: 252 seconds]
AtleoS has quit [Ping timeout: 272 seconds]
simjnd has quit [Ping timeout: 276 seconds]
simjnd has joined #osdev
simjnd has quit [Ping timeout: 260 seconds]
xenos1984 has quit [Read error: Connection reset by peer]
simjnd has joined #osdev
xenos1984 has joined #osdev
simjnd has quit [Ping timeout: 252 seconds]
simjnd has joined #osdev
ThinkT510 has quit [Quit: ThinkT510]
ThinkT510 has joined #osdev
the_oz has quit [Quit: Leaving]
simjnd has quit [Remote host closed the connection]
Lucretia has joined #osdev
bleb has joined #osdev
simjnd has joined #osdev
GeDaMo has joined #osdev
<Ermine>
heat: dalias will get mad at this
<Ermine>
gcc 15 breaks linux builds >_<
<heat>
ddevault: hardlinks don't work across filesystems
<heat>
Ermine: i know
Teukka has quit [Quit: Not to know is bad; not to wish to know is worse. -- African Proverb]
pabs3 has quit [Read error: Connection reset by peer]
pabs3 has joined #osdev
\Test_User has joined #osdev
goliath has joined #osdev
\Test_User has quit [Quit: \Test_User]
\Test_User has joined #osdev
simjnd has quit [Ping timeout: 276 seconds]
simjnd has joined #osdev
Matt|home has quit [Quit: Matt|home]
randm has quit [Remote host closed the connection]
randm has joined #osdev
Lucretia has joined #osdev
Lucretia-backup has quit [Ping timeout: 252 seconds]
simjnd has quit [Remote host closed the connection]
simjnd has joined #osdev
simjnd has quit [Ping timeout: 252 seconds]
\Test_User has quit [Quit: \Test_User]
\Test_User has joined #osdev
Matt|home has joined #osdev
\Test_User has quit [Client Quit]
\Test_User has joined #osdev
\Test_User has quit [Client Quit]
\Test_User has joined #osdev
\Test_User has quit [Client Quit]
\Test_User has joined #osdev
\Test_User has quit [Client Quit]
\Test_User has joined #osdev
simjnd has joined #osdev
Turn_Left has joined #osdev
Left_Turn has quit [Ping timeout: 248 seconds]
parabirb has joined #osdev
parabirb has quit [Quit: testing]
parabirb has joined #osdev
simjnd has quit [Ping timeout: 252 seconds]
bleb has quit [Ping timeout: 245 seconds]
bleb has joined #osdev
simjnd has joined #osdev
GeDaMo has quit [Quit: 0wt 0f v0w3ls.]
particleflux has quit [Quit: 418 I'm a teapot]
particleflux has joined #osdev
<sortie>
heat: My strong opinion is that close(2) should not fail with EINTR. posix_close is bullshit, but only because this shouldn't be the case. I didn't know it could happen in Linux though. I'll read up on why. But I really suggest whatever operations are going on happen inside some other syscall like fsync or ioctl or whatever, or that close just returns 0 and you lose the error status (and you should've used another syscall to get it)
<heat>
Why: networked filesystems
<\Test_User>
networked filesystems can just do their own stuff on their own time after the kernel frees up the fd though?
<sortie>
If you wanna know if a write went through, even for networked filesystems, I suggest fsync or such
<heat>
I'm not entirely familiar with the netfs side of things
<heat>
So I can't give an educated comment on that
<bslsk05>
lore.kernel.org: Making sure you're not a bot!
<heat>
Rich is an idiot
<zid`>
makes sense to me
<heat>
Complying to shitty make believe POSIX spec for dead systems
<heat>
Is more important to him than compat
<zid`>
I don't think it impacts the argument at all though
<sortie>
heat: Hey, don't ad hominem
<zid`>
That linux does it one way, and causing a regression in linux is not going to happen, just because posix finally got its act together and specified something
<heat>
I guarantee you no one, literally no one, cares about POSIX
<sortie>
You're supposed to be a professional here
<heat>
sortie: he is an idiot because *argument*
<sortie>
I should note that *I* care about POSIX and I also very much refused to implement posix_close
<heat>
Thus, not ad hominem
<zid`>
I also don't think linux is *correct* here, but regressionless > correctness has been linux's MO
<heat>
Thus, you sir are an idiot!
<sortie>
heat: Nah it's still unprofessional and doesn't fly. Rich is rather intelligent and you just happen to disagree
<sortie>
Which is fine
<zid`>
sortie: dw about heat, he's an idiot
<heat>
Yes I am.
<heat>
Obligatory statement that my opinions do not reflect those of my employer and I only speak for myself
<heat>
However, I disagree with most of his decisions lol
<bslsk05>
lore.kernel.org: Making sure you're not a bot!
<heat>
Most of them *strongly*
<sortie>
heat: Thing is that your opinions may be your one, but this is also literally the area where you're employed, talking about a work mailing list -- your opinions may be your own, but your professionalism also reflects on your employment
<heat>
Ok, lets rephrase it then
<heat>
I think Rich is very very wrong and I wish him good luck in breaking userspace
<sortie>
Yeah that sounds about right
<zid`>
He didn't say he wanted to break userspace
<zid`>
He just said that the man page is not authoritative, and that applications can't rely on close
<sortie>
I think Rich seems to be on my team, he doesn't want the EINTR case to happen, or EINPROGRESS, and just wants close to exit 0
<zid`>
he didn't appear to be advocating for *anything*
<sortie>
And nobody seems to have suggested so far, in the thread, still reading, how these cases can actually happen
<heat>
zid`: he advocates for POSIX compatibility
<zid`>
where?
<heat>
i.e breaking userspace in name of an imaginary spec
<sortie>
POSIX isn't imaginary
Turn_Left has quit [Ping timeout: 252 seconds]
<heat>
This POSIX, is it in the room with us?
<sortie>
But Rich said he returns 0 instead of EINPROGRESS in musl on close, so that's very non-breaking, because EINPROGRESS would be breaking
<sortie>
So far Rich was the only one to point out how this case might happen. Rewinding tape drives, implicitly. I'm not sure that also happens on Linux and could fail with EINTR. NFS is another case, and well, perhaps EINTR can happen there, but it's also not quite a filesystem with those strong promises, is it
<sortie>
close doesn't give you those promises for normal filesystems anyway as writes happen async in the background post close
<sortie>
Unless you sync
<heat>
Sometimes does on ext4
<sortie>
How so?
<heat>
ext4 code has some broken app compat code
<heat>
Where a file you create and rename gets fsynced at close
<heat>
Or something like that
<sortie>
The Linux devs are right though. It appears that POSIX is inventing stuff, or standardizing prior to consensus, without consulting Linux developers, and they have added a few things that seems to get WONTFIX treatment from Linux kernel devs. That is putting POSIX at odds with Linux, and they're gonna have to deal with that, come next edition, when Linux is just ignoring their interfaces
<heat>
The problem with POSIX is that it made sense in a corporate UNIX world
<heat>
It does not make sense in a Linux + a few other weird systems world
<heat>
If all UNIX systems do something, and Linux does something else. Should they standardize the !Linux behavior? Surely, everyone but Linux does it? Oh but Linux significantly outnumbers all deployments of those other systems, combines
<heat>
Btw I figured out today that roaming mobile data ships data to the home country before sending it to the actual destination
<heat>
And that is so stupid
<zid`>
that's why you had to pay!
<sortie>
That explains a lot lol
<zid`>
But yea, the phone network is still the phone network, even though they've gone digital
<heat>
My packets are travelling across europe before going to the actual destination
<zid`>
your phone number is more like an ip address, in that it's basically routing information
<heat>
120ms latency :)
<zid`>
portugal -> balkans -> destination -> balkans -> portugal
<zid`>
"Number is +44? better send the traffic to the UK so they can deal with it"
<heat>
Surely they could think of something better...
<heat>
Of course this is an edgecase but...
<zid`>
They did, it's called voip
<zid`>
but people still use normal ass teffelones, which use the normal ass teffelone system
simjnd has quit [Ping timeout: 276 seconds]
<zid`>
Look at how weird SMS is at some point it's fun
<heat>
It's kinda a problem in the EU because roaming is free, I have no real incentive to get a travel SIM card except bad internet
<zid`>
It's one of those.. evolved systems, if you know what I mean
<zid`>
where the people involved can look at it and go "ofc it works like this, look at how it worked before we added feature x"
<zid`>
until you're back in the 1800s
<heat>
Yes, mature and stable
<zid`>
like x86 encodings
<heat>
Wide compatibility
<zid`>
nobody made a *mistake* but you can probably redesign it with the new bottlenecks catered for :P
simjnd has joined #osdev
tigerbrother has quit [Quit: Ping timeout (120 seconds)]
<heat>
They made kind of a mistake
<heat>
By not having integrated encryption
tigerbrother has joined #osdev
<zid`>
how do you do encryption
<zid`>
in 1980
<zid`>
1980
<zid`>
1890
<heat>
SMS was created in the 1980s
<heat>
In any case, ez caesar cipher
<zid`>
yes but the phone network isn't
<zid`>
wasn't
<zid`>
You have to take what you have, and add features to it *without* rewiring all of europe
<zid`>
if you want a new feature to exist
<heat>
You could probably very trivially devise a scheme where the message plaintext is encrypted
<heat>
Which solves most of the issue
<zid`>
and who decrypts it? with what key?
<heat>
End to end hopefully
<zid`>
bear in mind we added 'encryption' to wifi, that was pretty good, and took a lot of compute
<zid`>
it was called WEP and it got cracked to shit
<heat>
Not sure about the key, I don't know much about SIM card internals
<heat>
AIUI SMS right now is partially encrypted
<heat>
Possible carrier to carrier, or maybe carrier to user
simjnd has quit [Ping timeout: 272 seconds]
<zid`>
have you.. seen a mobile phone from the 90s, before you were born
<zid`>
They used lead acid batteries and were the size of a house, and had transistors the size of paperclips
karenw has joined #osdev
<kof673>
> This POSIX, is it in the room with us? why yes, Steve Posix is sitting right next to Jack Wal-Mart, Doug McDonald, Debbie RedHat, Mark Microsoft
vdamewood has joined #osdev
Turn_Left has joined #osdev
Gooberpatrol66 has quit [Quit: Konversation terminated!]