<nekobot>
[haiku/haiku] d97ad7c4dc19 - termios: New ioctl: TIOCOUTQ
dqk has joined #haiku
<x512[m]>
Why Ethernet devices use regular read/write? It adds additional memory copy.
<jessicah>
instead of?
<x512[m]>
Or maybe it is already possible to create Ethernet drivers as "network/devices/*/v1" modules without devfs node?
<x512[m]>
jessicah: Directly using net_buffer and implementing net_device_module_info.
<jessicah>
I pretty much take the view that it was overlooked, or just needs somebody to do the work :)
<jessicah>
it's usually a sufficient why for a lot of things in Haiku :p
<x512[m]>
I need to create new Ethernet driver for VisionFive 2.
<Skipp_OSX>
sounds hard
<x512[m]>
U-Boot driver is ~2000 lines of code.
<x512[m]>
So seems not so hard.
mmu_man has quit [Ping timeout: 480 seconds]
<waddlesplash>
x512[m]: well the FreeBSD drivers have to use mbufs
<waddlesplash>
which will be a copy anyway
<x512[m]>
memcpy may hurt a lot on non x86 boards.
<x512[m]>
If write new native driver, *BSD mbuf should be not relevant.
<x512[m]>
Where and how Ethernet devices are enumerated?
<x512[m]>
It seems that network devices are enumerated by net_server in /dev/net file system path.
<x512[m]>
Also using separate thread to receive packets is suspicious design.
<x512[m]>
It should probably enqueue packets to stack from interrupt handler.
<andreasdr[m]>
Good morning
tuaris has quit [Quit: Leaving.]
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
B2IA has quit [Remote host closed the connection]
B2IA has joined #haiku
Begasus has joined #haiku
<Begasus>
g'morning peeps
<OscarL>
morning Begasus.
<Begasus>
Ola OscarL ! still up? :) (or already up)
<OscarL>
Still up, but I don't think for much longer :-D
<Begasus>
korli already did the fix for gtk_doc I see
<Begasus>
one less to worry about :P
<OscarL>
yeah, saw it too. When I was writting my comment on your PR, I almost left an "But I'm almost sure korli has something else in mind" :-D
<Begasus>
Not going to rebuild llvm9 any time soon to check on the python variables OscarL :P (although it's still pretty common), it's pretty stressfull to build :)
<OscarL>
guess then I shouldn't even think about trying it on my old PC then :-D
<Begasus>
could have a look at fltk (going through the issue from waddlesplash )
<Begasus>
takes about a full day here to build (still less then texlive)
<OscarL>
ouch.
<Begasus>
+4k objects to build there :)
<OscarL>
I tried to give another go to the lua recipes... mostly doing some clean up for the 5.4 one, and trying to understand them all, to see if we can end up avoiding the CONFLICTS for the _devel packages (AFAIK, it should be possible to have "parallel" installs)...
<OscarL>
but ran out of concentration again :-/ (guess I'll try again tomorrow).
Skipp_OSX has quit [Ping timeout: 480 seconds]
<Begasus>
OscarL, how are you going to do that?
<Begasus>
they all ship "liblua.so" as a symlink
<OscarL>
guess they shoudn't then :-) (will have to see how's done in other OSes)
<Begasus>
hence we end up with other recipes that carry CONFLICTS for devel packages
<Begasus>
ah, didn't go there yet :)
<OscarL>
but AFAIK, lua tends to mostly be used as a static lib so... will have to double check regarding the .so. Shouldn't be so discimilar to the Python situation, really.
<Begasus>
even then, only cmd:zipcmp should work in BUILD_PREREQUIRES?
<Begasus>
SONAME doesn't seem to change for the latest zlib (maybe time to update that as well?) :)
<OscarL>
right, mgba recipe should be explicit about cmd:zipcmp on BUILD_REQUIRES (instead of assuming it will be given by devel:libzip).
<OscarL>
not sure if we really need/want to keep cmd:zipcmp being from GCC2 instead of switching it to x86, but... that's a different issue :-)
HaikuUser has joined #haiku
HaikuUser has quit []
Diver has joined #haiku
orealis has quit [Read error: Connection reset by peer]
orealis has joined #haiku
<Begasus>
so far it has been a policy if library ship cmd's and can be build for both arch's the cmd's will only be provided in the primary package (didn't came up with that myself) :P
<Begasus>
but it makes sense, if you ship it for secondary arch only, primary won't be able to use them
<OscarL>
IIRC, that list was around 250 last year.
<Begasus>
that's another list :)
Diver has joined #haiku
orealis has quit [Read error: Connection reset by peer]
orealis has joined #haiku
<OscarL>
ah, "Potentially vulnerable" only. still 200 < 250 :-D
<OscarL>
progress!
Blendie has joined #haiku
<Begasus>
no, that was less :)
* OscarL
tickles checkboxes trying to find some positive stats then.
<Begasus>
been looking into bumping rizin a few times to bring cutter up to date, but somehow I can't make cutter detect rizin at configure time :/
Begas_VM has joined #haiku
<Begasus>
maybe I should push my recipe for ICU73 at one point :P
<OscarL>
Easy to forget things after a while... I forgot I had an "itstool" branch with some fixes post libxml2 changes. Saw that korli commited his version, so I just deleted mine :-D
<botifico-c849d97b>
[haikuports/haikuports] korli a52ed6c - mgba: trigger build on x86
<Anarchos>
hello
<Begasus>
Hi Anarchos
orealis has quit [Read error: Connection reset by peer]
orealis has joined #haiku
Raposa has quit [Ping timeout: 480 seconds]
pvalue has joined #haiku
Anarchos has quit [Quit: Vision[]: i've been blurred!]
pvalue has quit [Quit: Vision[]: i've been blurred!]
vdamewood has joined #haiku
OscarL has quit [Quit: Page closed]
Anarchos has joined #haiku
nephele has joined #haiku
freddietilley has quit [Quit: WeeChat 3.8]
freddietilley has joined #haiku
nexus6 has joined #haiku
nexus6 has quit []
Anarchos has quit [Ping timeout: 480 seconds]
<nephele>
Maybe a silly question but does anyone know how to connect to quakenet irc? I've been really trying, and for months i can't figure it out. I suspect the port i want to use is wrong, but there is no public info anywhere on what the correct port is for quakenet
<Begasus>
nice, can change the language for TuxPaint with tuxpaint-config :D
<Begasus>
doesn't like Chinese though
Forza has joined #haiku
Anarchos has quit [Quit: Vision[]: i've been blurred!]
Anarchos has joined #haiku
<Anarchos>
i wonder how just disabling 'acpi' can solve my 'unhandled pheader type' errors in syslog and the 'did not find boot partition' fatal error !
mmu_man has quit [Ping timeout: 480 seconds]
dqk has quit [Remote host closed the connection]
dqk has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<waddlesplash>
Begasus: don't use mmuman's port for fltk
<waddlesplash>
it's not complete I think, and using Xlibe is a totally different strategy
<waddlesplash>
Begasus: I guess my patch may only apply to 1.4. I'll make one for 1.3 to fix the fluid crash a different way
CPYou has joined #haiku
<Anarchos>
is fltk a modern BeBuilder ?
nephele has quit [Quit: Vision[]: i've been blurred!]
floof58 is now known as Guest1135
floof58 has joined #haiku
dqk has quit [Read error: Connection reset by peer]
<Begasus>
doesn't it make sense to build fltk after xlibe was finished at buildmaster waddlesplash ?
<waddlesplash>
no, it doesn't matter
<Begasus>
ah ok :)
<waddlesplash>
the xlibe change doesn't change API/ABI
<Begasus>
then it's about time to close down here, dogschool in a bit (and exams on Sunday) :)
Begas_VM has quit [Quit: Vision[]: i've been blurred!]
<Begasus>
need to cleanup on tuxpaint before a PR there
<Begasus>
cu peeps! (and thanks waddlesplash :) )
Begasus has quit [Quit: Leaving]
<waddlesplash>
... and Haiku has FLTK now
<waddlesplash>
(well, we've had mmu_man's incomplete port for years, but now we have the full thing)
dqk has quit [Read error: Connection reset by peer]
dqk has joined #haiku
braewoods has joined #haiku
HaikuUser has joined #haiku
<braewoods>
Is there anyway to unload a kernel driver or force it not to attach to a specific USB device? I was exploring the possibility of writing a USB driver in userspace via libusb or native APIs.
<braewoods>
Specifically usb_serial driver.
HaikuUser has quit []
mmu_man has quit [Ping timeout: 480 seconds]
<waddlesplash>
unfortunately I don't think so. you can prevent it from loading altogether if needed
<waddlesplash>
but really if the kernel driver doesn't work right it should just be fixed
<braewoods>
waddlesplash: Actually that wasn't really the problem I was trying to solve but for a project I am researching, it would really help my auto-detection if I could map the serial device back to its USB device.
<braewoods>
The USB device(s) emulate a USB ACM / CDC device that most OS can use as a serial port but most don't provide any way to find out what USB device it was created for.
<braewoods>
I have two real reasons for wanting to do that. To auto-detect supported USB devices but also to map to specific devices due to how device enumeration can end up being.
<braewoods>
Ironically though, in their quest to make these devices work without a special driver in windows, they made it harder to write one elsewhere in some cases.
<x512[m]>
trungnt2910: Also I think that reserve address range syscall can be used for .NET. Wrapper function without _kern_ prefix likely be needed.
humdinger has joined #haiku
<trungnt2910[m]>
x512[m]: Problem here is how to _commit_.
<trungnt2910[m]>
On Windows, MEM_COMMIT can be called on already committed pages. They're simply skipped.
higen6 has joined #haiku
<trungnt2910[m]>
If we want to do that on Haiku we'll have to use get_next_area_info and/or area_for to check for already existing mappings. Then we can use mmap/create_area.
<botifico-c849d97b>
[haiku/website] pulkomandy 437c839 - Fix a dead link
CPYou has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<zdykstra>
that's a really cool bit of history there
flag has joined #haiku
_orealis has joined #haiku
orealis has quit [Ping timeout: 480 seconds]
_orealis is now known as orealis
zard has quit [Quit: leaving]
orealis has quit [Read error: Connection reset by peer]
diver1 has joined #haiku
Diver is now known as Guest1153
diver1 is now known as Diver
orealis has joined #haiku
Guest1153 has quit [Ping timeout: 480 seconds]
jmairboeck has joined #haiku
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
B2IA has quit []
B2IA has joined #haiku
gouchi has joined #haiku
mbrumbelow has quit [Read error: Connection reset by peer]
DKnoto has quit [Ping timeout: 480 seconds]
Skipp_OSX has joined #haiku
DKnoto has joined #haiku
mbrumbelow has joined #haiku
dqk has quit [Read error: Connection reset by peer]
dqk has joined #haiku
mbrumbelow has quit []
mbrumbelow has joined #haiku
Diver has quit [Read error: Connection reset by peer]
Diver has joined #haiku
dqk has quit [Read error: Connection reset by peer]
Diver has quit [Read error: Network is unreachable]
dqk has joined #haiku
Diver has joined #haiku
<trungnt2910[m]>
<waddlesplash> "trungnt2910: no I don't think so..." <- That only happens if the whole range being mapped is available. Mmap with MAP_FIXED displaces memory while MEM_COMMIT on Windows leaves currently mapped pages unchanged.
<waddlesplash>
?
<waddlesplash>
create_area doesn't displace memory. I think? or maybe it does actually
jmairboeck has quit [Quit: Konversation terminated!]
<trungnt2910[m]>
It doesn't.
<trungnt2910[m]>
But create area will fail if there's an existing mapping at that spot.
<trungnt2910[m]>
For the dotnet port this code path actually gét called.
dqk_ has joined #haiku
CPYou has quit [Ping timeout: 480 seconds]
dqk has quit [Ping timeout: 480 seconds]
<x512[m]>
trungnt2910: Conceptually it can be implemented with one big area with initially no mapped physical pages. And then do separate call to allocate and map physical pages if not arleady.
<x512[m]>
Area is a mapping from linear offset to physical pages.
<x512[m]>
waddlesplash: Isn't it already possible to create an area with no mapping?
<waddlesplash>
with no protection? yeah
<waddlesplash>
and then it won't have any mapping if nothing accesses it
<x512[m]>
It should also do not increment used memory counter.
<x512[m]>
So creating 256 GB area should success.
mbrumbelow has quit [Quit: Leaving]
mbrumbelow has joined #haiku
gouchi has quit [Remote host closed the connection]
_orealis has joined #haiku
orealis has quit [Ping timeout: 480 seconds]
_orealis is now known as orealis
HaikuUser has joined #haiku
orealis has quit [Read error: Connection reset by peer]
<trungnt2910[m]>
It's a framework after all, furthermore inside dotnet itself is a lot of subsystems, some with their own virtual memory handling logic.
<x512[m]>
If some single subsystem will work with memory in reserved range using exact addresses, it should be fine?
<x512[m]>
Is there any documentation for net_buffer?
Diver has quit [Quit: Leaving.]
orealis has quit [Read error: Connection reset by peer]