tuaris has quit [Read error: Connection reset by peer]
tuaris has joined #haiku
bradsco has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<pairisto[m]>
how would I get something like tun.cpp to run as a driver? I think there is enough code for it to run as a driver and have the network/devices/tun/v1 appear when you run ifconfig but even if it doesn't I would still like to know how eventually
HaikuUser has joined #haiku
HaikuUser has quit []
bradsco has quit [Quit: bradsco]
<zdykstra>
welp, fixing #14782 is beyond my BLayoutBuilder knowledge right now
wintp has quit [Quit: Connection closed for inactivity]
<Habbie>
not on 7925 because it's a bump, not a new recipe
<Habbie>
however, on my PR it did not notice the missing PATCHES
<OscarL>
not sure who handles that part of the infrastructure on GitHub. kallisti5[m] or waddlesplash, I would assume.
<OscarL>
Might be worth asking/telling them about that lint check missing things :-D
<Habbie>
i'll see if i can investigate
<Habbie>
also you just mentioned them so perhaps they'l see :)
<Habbie>
+l
<OscarL>
might be too early in the day for them, if I got their timezones right :-), so if they don't react later, creating a new issue on Haikuports about that seems reasonable.
<Habbie>
ack
<Habbie>
i'll check a few things before i do that
tuaris has quit [Read error: Connection reset by peer]
<waddlesplash>
zdykstra: what don't you know about blayoutbuilder?
<waddlesplash>
Habbie: OscarL: CI just runs --lint
<waddlesplash>
of HaikuPorter
<waddlesplash>
easily adjusted if we want to add new things
mmu_man has joined #haiku
<OscarL>
waddlesplash: seems like when Habbie locally ran "hp --lint wdiff" it found a problem that the linter on Github failed to find (missing PATCHSET reference on the recipe).
<win8linux[m]>
Speaking of native clients for federated protocols, have there been any attempts at a native ActivityPub client?
<win8linux[m]>
MS-DOS and Classic Mac OS IIRC have AP clients at this point.
<PulkoMandy>
it's easier when you don't have to "hey I have to write the OS first" :D
<kallisti5[m]>
Hm. Not in haikudepot
<kallisti5[m]>
Doesn't build on modern gcc.. just silly things though
<zdykstra>
waddlesplash: re: BLayoutBuilder - if you have a few minutes some time, can you sanity check my current understanding of how something is done?
<waddlesplash>
srue
<waddlesplash>
sure
<OscarL>
what's the "etiquette" for "resolving" comments on Gerrit? I'm never sure if I'm supposed to mark them as such when addressed by a new patchset, or if that's for the original commenter to decide if it's actually resolved :-/
<waddlesplash>
if you resolved the problem, mark it resolved
<waddlesplash>
if it's not really resolved the original commenter can reopen
<OscarL>
Ok. Thanks. I guess it is more nuanced for comments that are more question that code change suggestions, right?
<waddlesplash>
honestly AddChild should not be used in layout mode
<win8linux[m]>
<PulkoMandy> "it's easier when you don't..." <- Rather sure that SerenityOS has an AP client too.
<zdykstra>
okay, so my understanding is wrong. That's what I thought :)
<win8linux[m]>
* client too. :D
<waddlesplash>
zdykstra: the second parameter is "after"
<waddlesplash>
err sorry. "before"
<waddlesplash>
it's adding the specified view *before* the other one
<win8linux[m]>
In theory, what kits would a native Haiku client use?
<waddlesplash>
zdykstra: so actually your understanding is mostly correct I suppose
<waddlesplash>
there's more items in the group than just those 2 though
<waddlesplash>
and like I said, imho AddChild shouldn't be used in layout mode, instead you should add items to the layout directly
<zdykstra>
so layouts can be appended at runtime?
<waddlesplash>
yes, why wouldn't they be?
<waddlesplash>
layouts aren't anything special
<PulkoMandy>
the same weght would lead to 50% of the group if there are no other constraints. But usually each view has a minimal or a preferred size that the layout system will try to satisfy as well
<PulkoMandy>
and the weight is used when all other constraints were not enough or too much, and the last resort is to distribute the remaining space to all views, wether they want it or not
<zdykstra>
thanks for the input waddlesplash and PulkoMandy
<zdykstra>
you've given me some direction to go on fixing this
bjorkintosh has quit [Quit: Leaving]
freddietilley has quit [Quit: WeeChat 3.8]
nosycat has joined #haiku
bjorkintosh has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
<zdykstra>
so I was thinking for backups that I'd create a ~10GB or whatever image file, put BFS on it, mount it and then rsync key directories in to it, unmount it, and then rsync it to one of my systems with ZFS where it'll land in my snapshots. The followup rsyncs, both in to the image and to the ZFS host, shouldn't be too bad.
<zdykstra>
I'm so used to having ZFS on all of my systems that replicate snapshots to a central host - I feel naked just having a bare nvme with Haiku on it
* OscarL
on the other hand, only ever made a couple of backup CDs from his old BeOS days.
<OscarL>
HDDs failures are met with a "damn it!" followed by a "oh, well... less stuff to organize, at least" :-P
mmu_man has joined #haiku
<zdykstra>
I found my BeOS backup CDs a few years ago - pretty sure that's actually where I pulled the ffmpegGUI source from
<zdykstra>
it might have been in the wild before then, I don't remember
<zdykstra>
I'm less concerned about drive failure with an nvme, and more concerned with accidentally deleting something and having no way to recover it
<zdykstra>
on my workstation, I take ZFS snapshots every 5 minutes of my home directory that are replicated to another machine
mmu_man has quit [Ping timeout: 480 seconds]
BrunoSpr has joined #haiku
<BrunoSpr>
hi all
<nosycat>
Hello!
<OscarL>
Hola
michel has joined #haiku
KapiX has joined #haiku
<Habbie>
so, hp --lint has warn() and sysExit()
<Habbie>
warn() does not cause it to exit 1
<Habbie>
sysExit() does - immediately
<Habbie>
this will need some minor refactoring
<zdykstra>
you know, I wonder if it'd be better to run Haiku in a VM, then rsync my Haiku machine to the VM, and then just snapshot the VM disk
<OscarL>
Habbie: seems like --lint error MO is to give up as soon as it finds an error?
<Habbie>
well, it would be nice to show all that is wrong, if possible
<Habbie>
what you say would work too, but feels a bit like the easy way out
michel has quit []
smalltalkman has quit []
<OscarL>
Is there any distinction on "hard errors", like... "this doesn't even parses as a shell script" (because for those, surely continuing sounds preposterous)
gouchi has joined #haiku
<OscarL>
(LOL @ me using "big words" with my crappy "English" :-P)
<Habbie>
:D
<Habbie>
ah! i thought lint would take multiple package names, but it does not
<Habbie>
so fatal errors already abort early now
<Habbie>
ok, then i'll take the easy way out
ClaudioM has joined #haiku
<Habbie>
waddlesplash, do you want CI to also lint recipe updates? currently it only lints new recipes
<waddlesplash>
yes, it should
<Habbie>
ok
<Habbie>
i'll have a look :)
<waddlesplash>
it should lint anything found in recent commits
<zdykstra>
what's the haiku equivalent of a cron job?
<Habbie>
shall I also move that script into .github? it's no use outside of it
KapiX has quit [Quit: KapiX]
Major_Biscuit has quit [Ping timeout: 480 seconds]
Maya-sama has joined #haiku
OscarL has quit [Remote host closed the connection]
kinkinkijkin is now known as Guest6966
kinkinkijkin has joined #haiku
OscarL has joined #haiku
<OscarL>
zdykstra: haven't used it, but pkgman shows "cronie" as "a cron daemon".
<OscarL>
not sure if Haiku proper has something.
<zdykstra>
yeah - I was hoping there was a Haiku-native way. I haven't seen anything in the Haiku user guide though.
Guest6966 has quit [Ping timeout: 480 seconds]
<zdykstra>
I suppose I could just make my own job, that shouldn't be too hard
<OscarL>
write a deamon that reads its taks from file attributes, so you can edit them with Tracker? :-P
<OscarL>
s/taks/tasks/
<zdykstra>
that's an interesting idea
<zdykstra>
but I'm looking for quick-and-dirty, now that my PoC seems to work
<zdykstra>
rsync to an image, the rsync the image to a remote host. Subsequent rsyncs of a 10GB image file have an effective re-transfer rate of 300MB/sec.
<OscarL>
there was "BeTasks 0.9 - sheduller with GUI-setup", but doesn't has source code :-(
<zdykstra>
looks like cronie will do it
mmu_man has joined #haiku
<kallisti5[m]>
Ok. Submitted a few minor fixes for chat-o-matic. Let's see if students still around
<zdykstra>
nice
<zdykstra>
hmm. I can actually just rsync files straight to ZFS, when extended attributes are enabled. getfattr shows that all of the attributes on a file were copied over, and rsyncing it back preserved them.
tqh has joined #haiku
<zdykstra>
Anybody know off the top of their head the maximum data size of a BFS attribute?
Haikunaut has joined #haiku
chuggy has joined #haiku
<OscarL>
I remember one prank "magic compressor" back in the day... reduced your files to zero bytes... by storing the full content as an attribute. If that's still possible, I would assume same max size of files in BFS, right?
<zdykstra>
Oof. Max data size of extended attributes on Linux is 64k.
Haikunaut has quit [Remote host closed the connection]
<OscarL>
zdykstra: in any case, better ask one of the actual devs (or someone reading with some sources :-D)
kallisti5 has joined #haiku
kallisti5 has quit [Remote host closed the connection]
<pairisto[m]>
how would I get the tun.cpp to run as a driver? I think there is enough code for it to run as a driver and have the network/devices/tun/v1 appear when you run ifconfig but even if it doesn't I would still like to know how when there is enough code eventually
kallisti5 has joined #haiku
kallisti5 has quit [Remote host closed the connection]
<waddlesplash>
Habbie: yes, probably it should be moved, we didn't have .github when I made the script
<waddlesplash>
zdykstra: only 64k on extfs
<waddlesplash>
XFS supports larger attrs
nosycat has quit [Quit: Leaving]
HaikuUser has joined #haiku
HaikuUser is now known as choulth
<choulth>
hi there
<zdykstra>
hola
kallisti5 has joined #haiku
kallisti5 has quit [Remote host closed the connection]
mmu_man has quit [Ping timeout: 480 seconds]
choulth has quit [Quit: Vision[]: i've been blurred!]
OscarL has quit [Remote host closed the connection]
mmu_man has joined #haiku
DKnoto has quit [Ping timeout: 480 seconds]
OscarL has joined #haiku
<OscarL>
pairisto[m]: I'm not of much help (due to my lack of skills) but maybe comparing that tun.cpp file against src/add-ons/kernel/network/devices/ethernet/ethernet.cpp can give you some clues?
DKnoto has joined #haiku
<pairisto[m]>
I am currently trying to add it to the build path and recompiling it to see if it actually appears in Haiku itself and with that I can maybe do something like ifconfig /dev/net/tun create? The documentation is fuzzy on how to actually add a driver to /dev
<pairisto[m]>
unless I am just reading it wrong
<pairisto[m]>
* OscarL: I am currently trying to add it to the build path and recompiling it to see if it actually appears in Haiku itself and with that I can maybe do something like `ifconfig /dev/net/tun create`? The documentation is fuzzy on how to actually add a driver to `/dev`
<zdykstra>
should be able to add the build artifact to /boot/home/config/non-packaged/add-ons/kernel/drivers/dev
<zdykstra>
(that
<zdykstra>
's just a WAG though, I've never done driver development)
<pairisto[m]>
do you know how you would make it a character driver since thats apparently what haiku wants for a network driver? (I ask because the ethernet driver is represented in haiku as a character driver)?
<pairisto[m]>
* character driver?)
<pairisto[m]>
* do you know how you would compile it as a character driver since thats apparently what haiku wants for a network driver? (I ask because the ethernet driver is represented in haiku as a character driver?)
<botifico>
[haikuports/haikuports] korli 13a6cbe - python packages: rebuild to install directly as vendor-packages
<botifico>
[haikuports/haikuports] korli 9dd78e2 - sip: also build with python 3.9
zard has quit [Quit: leaving]
<OscarL>
pairisto[m]: notice that there seems to be a distinction between "drivers" (that publish "devices") and "modules", that can be used by other modules or by the drivers to help them do their job.
<OscarL>
(that's my crude undertanding of that, at least)
<OscarL>
I think that tun.cpp is a "module" and I don't know if those publish things under /dev (maybe under the newer API they can?).
<pairisto[m]>
then wouldn't something like loopback.cpp also be a module?
<OscarL>
that tries to get the PCI kernel module, so you can then access it from that "pci" variable.
<OscarL>
accessing functions from said module like this: pci->read_pci_config() and so and so.
<OscarL>
said modules have their interface available either in a publich header file (like "#include <PCI.h>"), others have them on "private" headers.
<OscarL>
that you can still can access, after massaging a bit the jamfiles :-)
<pairisto[m]>
in tun.cpp it has this:`get_module(NET_STACK_MODULE_NAME, (module_info**)&sStackModule);` but because `NET_STACK_MODULE_NAME` isn't defined, it wouldn't open anything right?
Blendie has quit [Quit: Connection closed for inactivity]
<OscarL>
The aptly named "publish_devices()" function :-D
KapiX has joined #haiku
<OscarL>
As usual when I *dare* to offer some "help" (ha! that's a good joke!) is in hopes that if I err too much, more experienced/knowledgeable folks will chime-in to set me straight, and we can learn in the process.
<pairisto[m]>
lol, so would I not have to exactly use tun.cpp if I were trying to import something like wireguard over and just use tun/driver.c directly?
<pairisto[m]>
that is the ideal but your help is helping me understand a little better bit by bit
<OscarL>
(while also making those experienced folks' blood boil a little for having to deal with my dumbness).
<pairisto[m]>
lol
<OscarL>
I can't really comment on your last question. I feel I've already stretched too much the little I know. Better quit when I'm "ahead" :-D
<pairisto[m]>
fair enough, I thank you for the help though
<botifico>
[haikuports/haikuports] korli fe8f8e1 - pyqt: move to python 3.9
<OscarL>
np!
ClaudioM has quit [Quit: leaving]
<OscarL>
pairisto[m]: just in case... there are plenty of things in the haiku source repo that are not present on the images. Like for example, the mentioned acpi_thermal.c driver (and lots of others).
<OscarL>
that can be because those are considered unfinished, or nothing else uses them (like the acpi_lid driver)...
<pairisto[m]>
I have learned that the hard way and have found it sorta frustrating but I understand why that is
<OscarL>
those you need to "jam -jN <target_you_want>" (eg: jam -jN acpi_thermal) and then move the resulting binary from the "generated/...../" folder, to the appropriate location...
<pairisto[m]>
THANK YOU! I needed that
<pairisto[m]>
* needed that information
<OscarL>
either under "/system/non-packaged/add-ons/kernel/..." or under "~/config/non-packaged/add-ons/kernel/..."
<OscarL>
pairisto[m]: sometimes you'll have a hard time figuring out what the target is called... don't try to guess it, just open the jamfile in the folder of the target you want...
<OscarL>
that file contains the target name.
<pairisto[m]>
got it
<OscarL>
for example, the mentioned "test" driver...
<OscarL>
I doubt using "jam test" will do you any good :-D
<pairisto[m]>
it wouldn't have the extension after it right?
<pairisto[m]>
like .c
<OscarL>
I haven't seen targets with extensions.
<pairisto[m]>
* like .c
<pairisto[m]>
ok just making sure
BrunoSpr has quit [Remote host closed the connection]
KapiX has quit [Quit: KapiX]
<OscarL>
Also of note, some drivers that go under "add-ons/kernel/drivers/bin", and you need to symlink them to the appropriate subdir of "add-ons/kernel/drivers/dev/" (depending on if the are net drivers, graphic drives, etc, etc).
<OscarL>
That was the case for the drivers that use the older API at least. Not sure about the newer one.
<nekobot>
[haiku/haiku] e490b6343abd - kernel/arm64: align interrupt postlude with other architectures
bradsco has quit [Quit: bradsco]
<pairisto[m]>
do you know what "bone_tun.h" is supposed to be? I tried looking it up on OpenGrok but its just defined in the `tun/driver.c` and just isn't a file
<OscarL>
acpi_thermal... I just placed it under add-ons/kernel/drivers/power, with no symlink so... appears that's a difference.
<OscarL>
mmm BONE was an experimental networking API for BeOS.
<OscarL>
That may indicate that that tun.c driver was written for BeOS and not Haiku.
<OscarL>
(I mean... written ON BeOS, with the intention of it working one day on Haiku)
<pairisto[m]>
well then I might want to update that
<OscarL>
Lots of stuff accumulate and get a bit stale in a project of over 20 years :-D
<OscarL>
One more thing that WILL byte you in the rear...
<zdykstra>
I got a good laugh last night - I found an ifdef for DANO
bradsco has joined #haiku
<zdykstra>
I remember when that hit BeShare / IRC back in the day. It was pretty exciting to run that.
<pairisto[m]>
I think I could possibly replace some of the bone_tun.h stuff with the net_ libs
<OscarL>
when working on drivers that use the old API... you can basically overwrite them with a new binary you just compiled... the newer code will be loaded up without the need for a reboot.
<OscarL>
No such luck with the drivers that use the newer API. you HAVE to reboot for the changes to get picked up (otherwise, the old binary still gets used, and you get confused AF).
<pairisto[m]>
makes sense
<OscarL>
pairisto[m]: Dano was a development version of BeOS 5.1 that got leaked. It included the mentioned BONE networking stuff.
<OscarL>
(among other improvements/changes)
<pairisto[m]>
did it ever release?
<pairisto[m]>
nrm
<OscarL>
So, being seen as where BeOS was headed before imploding, some parts of it was taken as inspiration.
<pairisto[m]>
didn't see the word leaked
<zdykstra>
if memory serves, the build date on the kernel was the morning of the last day of Be Inc.
<botifico>
[haikuports/haikuports] korli c80442d - dbuspython: build for python 3.9
<OscarL>
have you configured the, really usefull "inrecipe()" and "inpatch()" helper functions in your bash profile, Habbie? they can be really handy when working with HaikuPorts.
<Habbie>
i took mine from haikuports-sample.conf i think
<OscarL>
I use a slightly modified version of the latter (excluding the haikuports/{repository,packages} dirs to help my slow HDD)
<Habbie>
right
<Habbie>
i bought this SSD a month ago so i'm good ;)
<OscarL>
I even have to use "commit.status=false" on git (for the Haiku and Haikuports repos at least) to help with the slowness :-D (thanks to jessicah for the tip!!!)
<OscarL>
And... "pip3.10 install --user pyserial" is still broken (installs to /boot/home/.local/{bin,lib}), but I think I can fix that far easier than the previous issue korli has taclked.
<OscarL>
Habbie: attempting to understand what was going on with my serial_mouse driver.
<Habbie>
ah
<OscarL>
managed to find and fix at least two bugs (on on the serial driver, one on the tty layer)... but there are still problems with the serial driver that prevents me to progress further on fixing that darn mouse driver :-D
<Habbie>
like, an rs232 mouse?
<OscarL>
yes.
<Habbie>
nice
* OscarL
is still using a PS/2 ball mouse, that's able to do serial via a simple adaptor.
<Habbie>
:)
<OscarL>
I lost track of how many times I've changed its microswitch.
<Habbie>
single?
<OscarL>
*microswitchs
<OscarL>
:-D
<OscarL>
hey! it is old, but has 3 buttons. (I alternate with another that also has TWO wheels :-D)
<Habbie>
:)
<zdykstra>
I'm on an Elecom HUGE these days. The trackball space is suprisingly limited compared to regular mice.
<OscarL>
Had to make that to use both wheels on BeOS :-D
<zdykstra>
huh, neat
<OscarL>
Good thing the API was already there with "x vs y" wheels!
<zdykstra>
I went to college with the guy that created the original BeBits. I saw a guy wearing a BeOS shirt during finals, and it surprised me so much I went to talk to him.
<OscarL>
oh, nice! :-D
<Habbie>
OscarL, ohh, do you still have the source for that?
<Habbie>
because i asked about wheel direction here and somebody suggested i could build a filter
<OscarL>
Habbie: heh.
<Habbie>
assuming that i shouldn't be using some newer APIs for that
<OscarL>
Yeah... it happens to be on one of the few backups CDs I ever made (and the ones I've mentioned earlier to zdykstra :-P)
<OscarL>
Will have to dig it up, and upload it as-is to github, I guess.
<Habbie>
:)
* Habbie
uses freshly built pkgdiff to check changes in pkgdiff port