<OscarL>
ok, tried it on and old Linux install (but using version 4.6.1 of squashfs-tools) ... there `mksquashfs` also creates files with the first 4 bytes reading "hsqs"... so I guess we're good to go with my patch changing endianness then :-)
<OscarL>
nice... uses all available cores automatically :-)
<OscarL>
now we need squashfs support on Beezer.
ablyss has joined #haiku
<augiedoggie>
PRs accepted
* OscarL
wonders who might help with that :-)
<OscarL>
ouch :-D
<_-Caleb-__Haiku>
thanks guys
v_harkonnen has quit [Ping timeout: 480 seconds]
<_-Caleb-__Haiku>
is strange compress and uncompress in squashfs if not are a operating system, etc.... :-) but i use for backups because are read-only files :-))
_-Caleb-__Haiku has quit [Quit: shutting down the laptop!]
<augiedoggie>
you could open an enhancement ticket and maybe Teknomancer will get motivated to work on it
ablyss has quit [Remote host closed the connection]
<OscarL>
good old Ramshankar... I remember he (back in the days) liking Cradle of Filth, and not knowing who The Beatles were :-)
HaikuUser has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
each succesive IOM window I open (and then close)... starts higher and higher than the previous one :-)
ablyss has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<augiedoggie>
mine usually goes off the screen at the bottom
<augiedoggie>
oh, close too
<OscarL>
heh :-)
<augiedoggie>
i've noticed it's bad about window positioning in several ways
v_harkonnen has joined #haiku
<OscarL>
this "escalation" for me happens opening different files, then closing, opening a new one, etc.
<OscarL>
then we have some older apps that seemingly try to start at 0 height, but end up with part of the title bar off-screen (Pe, tends do do this, for example).
v_harkonnen has quit [Ping timeout: 480 seconds]
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
_-Caleb-_ has left #haiku [#haiku]
_-Caleb-_ has joined #haiku
xet7 has quit [Remote host closed the connection]
<OscarL>
gcc2 is getting on my nerves.
ablyss has quit [Remote host closed the connection]
<x512[m]>
OscarL: gcc2 should be not needed at HaikuPorts except some native add-ons.
xet7 has joined #haiku
<OscarL>
x512[m]: Indeed. Too bad Beezer is still GCC2 (albeit I guess only it's Tracker addon really needs to still be on GCC2?)
vdamewood has joined #haiku
rennj has joined #haiku
vdamewood has quit []
vdamewood has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
<OscarL>
heh, got Beezer to show the contents of squashfs files. NOICE :-)
<OscarL>
on 64 bits, thou. No idea why on 32 it doesn't fails, but it doesn't shows the contents either :-D
<rennj>
hmm squashfs
<rennj>
been using it before it was in the linux kernel
<rennj>
unionfs/aufs/overlayfs and squashfs made ram-os immutable possible
<rennj>
knoppix and cloop compressed loopback mounts was inspiration...
<rennj>
i would prefer rom-os, but immutable ram os..loaded from iso into ram..will do..readonly media..i know its static
<_-Caleb-_>
Begasus i've seen you've updated dillo_plus to the last versions... but the package include the dpis (plugins) but it doesn't work! (Gemini, FTP, Gopher, etc) any idea?? the dpi daemon (dpid) show:
<_-Caleb-_>
- Can't find the directory for dpis.
<_-Caleb-_>
[dpid]:
<_-Caleb-_>
Any idea about that? :-(
<_-Caleb-_>
Good morning :-)
<_-Caleb-_>
I supose is because the config files are in a read--only folder... maybe? i don't know i only ask for things :-D
<Begasus>
no idea _-Caleb-_ (without looking into it)
<Begasus>
_-Caleb-_ maybe you can use a paste site instead of flooding the channel ;)
<_-Caleb-_>
maybe only need ajust the package structure
<_-Caleb-_>
OMG im so sorry!!! 🥺
<Begasus>
;)
<Begasus>
from the PR: Basic webbrowser, this is mostly to show our FLTK port is up to building it. :)
<Begasus>
think I only checked it with some local files
manj-gnome has joined #haiku
manj-gnome is now known as kajedev
<Begasus>
bbl
_leosma has joined #haiku
v_harkonnen has joined #haiku
coolcoder613 has quit [Ping timeout: 480 seconds]
HaikuUser has joined #haiku
HaikuUser has quit []
FreeFull has joined #haiku
rennj has joined #haiku
BrunoSpr has joined #haiku
<kallisti5[m]>
⚠️ ⚠️ NOTICE: If you're running xz_utils-5.6.1-1, you may want to upgrade to the latest xz_utils-5.6.1-2. A backdoor was found in xz_utils-5.6.1. While it doesn't seemingly target Haiku, upgrading to the latest release should remove the obviously malicious code. thanks to jmairboeck for getting a rapid fix in.
kajedev has quit [Ping timeout: 480 seconds]
<kallisti5[m]>
However, be leery of xz_utils in general until researchers audit the codebase more closely
<jmairboeck>
thanks kallisti5[m]
<kallisti5[m]>
i'm going to send an email out to the mailing lists. From what I gather, Haiku shouldn't be impacted, so not making a website post or anything
<jmairboeck>
kallisti5[m]: I just noticed that the riscv64 repository definition also uses xz_utils-5.6.1-1. That should be updated too.
<kallisti5[m]>
oof. yeah, I probably just built that
<jmairboeck>
the other architectures which have it, use older versions, so they should be safe
<waddlesplash>
we're safe, there's no reason to panic, the backdoor only targeted Linux
<waddlesplash>
but the backdoor comes from a maintainer who has been modifying xz for years ... so who knows how deep the rabbit hole goes
<waddlesplash>
nobody seems to have analyzed things yet
<PulkoMandy>
People are reviewing the commits and reverting other changes already
<kallisti5[m]>
ok. I sent an official looking email out about it
<dovsienko>
I have nothing to add to what I said about it already
BrunoSpr has quit [Quit: Vision[]: Ich wurde verwaschen!]
BrunoSpr has joined #haiku
<waddlesplash>
kallisti5[m]: you did? where?
<waddlesplash>
hasn't shown up in my inbox yet at least
<kallisti5[m]>
fighting freelists at the moment 🥴
<waddlesplash>
ouch
<dovsienko>
lemme guess, now spam filters reject all e-mails with the string "xz" in it, to ensure 100% uninterrupted total protection from evil hackers?
<kallisti5[m]>
lol. nah. I subscribed to the haiku-security mailing list, which instantly made me NOPOST (while I had an ACL i could post)
<dovsienko>
+2 months free if you pay a yearly subscription in advance
<mmu_man>
hmm, I think we miss a dependency on mesa somewhere
<mmu_man>
GLTeapot won't run on this box I updated recently from a very old install
<mmu_man>
I have mesa_x86 but not mesa installed
<mmu_man>
and glu
<kallisti5[m]>
mmu_man: pkgman search mesa -D ?
<mmu_man>
and now GLTeapot runs but wants a renderer of course
* Anarchos
c'ant remember the pkgman command to list the file content of a package
<Begasus>
nope
<Begasus>
probably something only available for X11? (or the likes)
<augiedoggie>
it's not pkgman, it's `package list foo.hpkg`
<Anarchos>
augiedoggie oh thanks
<Anarchos>
augiedoggie i can't remember having ever used the 'package' command !!
<nosycat>
Me either.
<augiedoggie>
btw, to your question the other day about closing windows from Terminal, `hey Tracker Tciw`
<win8linux[m]>
<kallisti5[m]> "i'm going to send an email out..." <- Sent this over on the mailing lists already, but also putting this here as well in case people miss it:
<win8linux[m]>
Debian folks are mulling over reverting to 5.4.5, the last xz version not touched by Jia Tan (the rogue maintainer).
<win8linux[m]>
This may require patches due to ABI breakage, though.
<nosycat>
And we just patched xz.
flag has quit [Remote host closed the connection]
<win8linux[m]>
Reverting xz to that version may also require a mass rebuild, yes?
<Anarchos>
augiedoggie i still see 'w>XXX' threads in 'ps Tracker' after the hey command.
<win8linux[m]>
Ik that the Ubuntu folks are also in the process of doing this.
<win8linux[m]>
Ik that the Ubuntu folks are already in the process of doing this.
<B2IA>
(AGMS) Phew, my Haiku VirtualBox hosting server is using Fedora 39, which is at xz 5.4.4 (according to their package manager, can't trust running xz to find the version!).
<augiedoggie>
Anarchos: you will always have a couple of windows open, one is for the Desktop and one is a status window for file copies and stuff
<augiedoggie>
they are hidden but still around
<win8linux[m]>
Wait a sec, are there any packages in HaikuPorts that were built around the compromised xz version?
<augiedoggie>
Anarchos: and it only closes the windows on that workspace, so you may have some open on another
flag has joined #haiku
stux|away has quit [Ping timeout: 480 seconds]
<augiedoggie>
probably ffmpeg at least
<Anarchos>
augiedoggie i am connected through ssh ...
<Anarchos>
hmmm gtk3_devel contains gdkwayland.h but no gdkx.h...
<Anarchos>
maybe the haiku gtk3 port uses only wayland backend ?
<kallisti5[m]>
<win8linux[m]> "Sent this over on the mailing..." <- > <@win8linux:kde.org> Sent this over on the mailing lists already, but also putting this here as well in case people miss it:
<kallisti5[m]>
Yeah.. It's a mess. We can downgrade if it helps match what others are doing
<kallisti5[m]>
Lame. We can Rev bump those to rebuild... But let's definitely wait for a solid xz solution first
<Begasus>
yeah
<x512[m]>
> but I don't trust xz at this point
<x512[m]>
It is not a problem specific to XZ. Any open-source project can be attacked in this way.
<nosycat>
In theory. But certain factors make projects more vulnerable.
<x512[m]>
Also it is a vulnerability of Autotools build system. Autotools should be avoided if possible.
<kallisti5[m]>
I know lol. It's an attack path I've wondered about tbh
<kallisti5[m]>
Way too much soft trust
<kallisti5[m]>
Open source is enough work without bad actors
<x512[m]>
Autotools projects include generated scripts in source distribution that are not present in source repo.
<kallisti5[m]>
4k line autoconfs need to die
<Anarchos>
x512[m] i never understood why people commited generated config/make files .
<nosycat>
I see a lot of projects lately going back to simple makefiles.
<waddlesplash>
dovsienko: here's an interesting question. is bluetooth treated as a network interface on other OSes, or is it "sui generis"?
<x512[m]>
Strict repo snapshots should be used instead of "official" source archives that includes potentially infected "configure" script.
<dovsienko>
waddlesplash: in libpcap network interfaces are a subset of capture devices
<waddlesplash>
obviously it has to be at least a little bit in the network stack if socket(AF_BLUETOOTH) will work
<waddlesplash>
right, makes sense
<waddlesplash>
but I am wondering if in Haiku we should merge it into the network system entirely
<waddlesplash>
there are a number of potential downsides here
<dovsienko>
so long as you do not mix PAN into scope, BT is an example of a non-network capture device
<x512[m]>
Anarchos: It is recommended by official Autotools developers team.
<waddlesplash>
but there's also the upside that we wouldn't have to come up with a separate capturing interface
<waddlesplash>
and other management could be merged too
<x512[m]>
Autotools is terrible in any possible way, including security.
<dovsienko>
although I am not entirely sure about Bluetooth, to be honest. USB is a good example of non-network capture device
<x512[m]>
I hope that this incident will stop dangerous practice of providing source archives that do not match repo revision.
<Anarchos>
x512[m] machines are fast enough now to rebuild config/make files, so i don't feel the need to commit generated files. As at work , we try to never duplicate information...
<dovsienko>
although it has a hub... okay, D-bus and NFLOG are most certainly non-network producers of packets/frames/messages
<waddlesplash>
x512[m]: it does not matter. the actual object code in this case was checked into git in the tests directory. the only thing the release archive has different is the code to add it to the build.
<x512[m]>
Autotools fans claim that providing generated configure script will simplify build process.
<waddlesplash>
x512[m]: a few more levels of subtlety and the release archive / git repo difference wouldn't have existed
<Anarchos>
x512[m] i hope this fan base reduce soon with age :)
<waddlesplash>
dovsienko: in Haiku at present, L2CAP is mostly dealt with inside net_buffers, though they aren't sent/received through the network stack interface. with some rework they could be
* Anarchos
wonder if a haiku dev has troubles too with 'jam unittests' ?
<x512[m]>
waddlesplash: Test archives contents don't matter unless starter script is added to source archive.
<x512[m]>
So Autotools is a root cause.
<waddlesplash>
no? this could also be done through other build systems?
<waddlesplash>
the tests are not in a separate archive
<waddlesplash>
or at least these aren't
<x512[m]>
Other build systems do not need to provide source archive that don't exactly match repo revision.
<waddlesplash>
and neither does autotools
<waddlesplash>
some repos check ./configure into SCM
<x512[m]>
Regular GitHub automatic source archive generation for release tags if perfectly fine for every popular build system except Autotools.
<dovsienko>
that was the way in tcpdump, until it was not (too much generated changes for simple commits, also Autoconf flavours are different)
<waddlesplash>
regenerating ./configure isn't easy though because autoconf versions are generally not forwards OR backwards compatible
<waddlesplash>
you pretty much have to use the exact version they did
<dovsienko>
that said, release tarballs have ./configure so host without the required Autoconf version can build it
<dovsienko>
tcpdump and libpcap do it the old way because they have been doing this well before git (or Subversion) (or maybe even CVS) existed
<x512[m]>
waddlesplash: It is not an excuse at all. Every open source project should document build requirements. How developers are supposed to make contributions if they are unable to chsnge build definitions?
<Anarchos>
waddlesplash didn't know that for the autoconf forwards OR backwards compatible. How rude...
<dovsienko>
"this is the release tarball. this is the signature. you install into production from that. for development, git is over there"
<x512[m]>
dovsienko: Who guarantee that release tarball is not compromised if it includes some random scripts not present in repo that is supposed to be run with root privileges on install?
<x512[m]>
In GitHub it is guaranteed by automatic release archive generation.
<dovsienko>
it is trivially convertible into a to-do list
extrowerk_ has quit [Quit: Bye!]
<nosycat>
F-Droid started using reproducible builds a while ago, but I'm not sure how it helps except with middleman attacks.
GenericConfluent has quit [Ping timeout: 480 seconds]
<dovsienko>
x512[m]: the release engineer, who cryptographically signs the tarball
<x512[m]>
Release engineer of XZ was an attacker.
extrowerk has joined #haiku
<nosycat>
Exactly.
<x512[m]>
More people to be trusted -- more vulnerabilities.
<dovsienko>
a backdoor can be embedded straight into git repository, in this specific case it was supposedly not
<nosycat>
This was a human problem, not a tech problem.
<nosycat>
Also an audit problem, because *everything* is too complicated today.
<x512[m]>
Git repo have a history of commits that can be easily checked.
<dovsienko>
what would prevent an attacker from adding a botched obvious backdoor to the tarball and adding a well hidden one to the "trusted" git repository, which everyone would suddenly rush to start using?
extrowerk has quit []
<nosycat>
Plans within plans, and traps within traps?
<x512[m]>
It is definitely a tech problem, terribly insecure practice of making Autotools-based projects source releases.
GenericConfluent has joined #haiku
<x512[m]>
Source archives should be generated fully automatically without any intervention of release engineers etc.
<dovsienko>
it is not fair to blame tools when the root cause is an ill-intended actor
<x512[m]>
More human intervention -- more vulnerabilities.
<x512[m]>
Tool that need more trust for many people is more vulnerable. Obvious thing.
<dovsienko>
all successful IT attacks were accomplished by people who drink water and without water they would not be able to mount the attack, so let's declare water illegal
<x512[m]>
Autotools also use very old and obscure languages such as M4 that a few people can audit for attacks.
<nosycat>
That's a better point.
<dovsienko>
I find M4 about as intelligible as CMake sometimes
<x512[m]>
M4 was a final launcher of attack.
extrowerk has joined #haiku
<nosycat>
M4 is also antiquated.
<FreeFull>
The attack would have taken a lot longer to find if someone didn't realise sshd is 0.5s too slow
<dovsienko>
it is and I very much disliked it on its technical grounds when I had to use it regularly, but once again you try to ignore the root cause
<dovsienko>
FreeFull: you are just a couple steps from realising how many other backdoors have not been detected yet because they were better engineered (or marketed!)
<nosycat>
I saw links to that older article about xz's inherent downsides.
<nosycat>
Who'd have thought improbably good compression is dubious.
<kallisti5[m]>
barf.. sorry for top posting on the ML
<dovsienko>
waddlesplash: working pcap_inject() would enable a couple other programs to work on Haiku. kernel-mode BPF would make performance better if that's a problem for anyone
<Anarchos>
FreeFull nah i stick to sectorC side :)
<nosycat>
Even with -O2, C code compiles a lot faster than C++.
<FreeFull>
You could use sectorc to compile tcc, and then tcc to compile gcc or something
<nosycat>
And even textbook optimization can go a long way.
<x512[m]>
It depends on what C++ features are used.
<dovsienko>
waddlesplash: also perhaps it could be useful to enable Haiku USB as a capture device, if anyone is into USB driver development or some such
<x512[m]>
C with classes code compiles fast.
<waddlesplash>
dovsienko: got any examples of what the pcap code would look like for that?
<nosycat>
Yeah, but I use modern features quite a lot.
<waddlesplash>
dovsienko: I have worked a lot on our USB stack, would have to see about where we need to insert code to log packets
<Anarchos>
FreeFull right. And if you don't use partition table, i really wonder if you can boot an entire OS only with sectorC...
mattlacey has joined #haiku
<dovsienko>
waddlesplash: see pcap-usb-linux.c and --enable-usb
<waddlesplash>
one tricky thing is going to be adding hooks at the necessary level
<waddlesplash>
right now USB "packetization" happens inside the HCI drivers
<waddlesplash>
by the time you get to the stack things have been un-packet'ed
<dovsienko>
Guy Harris understands best how the USB code works, if you ask on tcpdump-workers@, he may provide a lot of useful starting points
<dovsienko>
I do not remember using USB capture
<dovsienko>
not long ago I did a CAN-bus capture for the first time to confirm a bug report. a couple years ago I tested DVB-capturing code (another good example of non-network capture)
stux|away has joined #haiku
<dovsienko>
Linux netlink and netfilter would make another good example of non-network
iar[m] has quit [Quit: Client limit exceeded: 20000]
<dovsienko>
hopefully libpcap 1.10.5 will be released soon, then you will have the recent lot of easy improvements and bug fixes available and it should be easier to decide what to do next
jmairboeck has quit [Ping timeout: 480 seconds]
Anarchos has quit [Quit: Vision[]: i've been blurred!]
mmu_man has quit [Ping timeout: 480 seconds]
mattlacey has quit [Ping timeout: 480 seconds]
GenericConfluent has quit [Ping timeout: 480 seconds]
<dovsienko>
waddlesplash: to answer your original question, if you configure BT as PAN in Linux, you can pair two dongles and use them as short-range wireless network, then you have an ordinary Ethernet-link network interface and can assign addresses to it
<dovsienko>
in fact, I used such a network in a home lab setup many years ago. that said, PAN is just one mode of Bluetooth and the rest do not expose network interfaces
<dovsienko>
you have to use libbluez or some such to work with headsets, mice etc.
<dovsienko>
you have backups of your own git repositories, correct?
<dovsienko>
as the fine print surely explains it, the git hoster may have a different idea of what is "your own"
<nosycat>
I treat Codeberg as a public mirror.
<nosycat>
Anyway, see you!
nosycat has quit [Quit: Leaving]
gouchi has joined #haiku
<Begasus>
closing down here, cu peeps!
Begasus has quit [Quit: Vision[]: i've been blurred!]
<win8linux[m]>
<kallisti5[m]> "https://github.com/tukaani-..."; <- Yeah, that's real helpful GitHub. Could've made the repo read-only but *noooooo*, take down the whole thing (including the website).
<win8linux[m]>
Also underscores that maybe relying on GitHub long-term may not be such a good idea.
jacereda has joined #haiku
<win8linux[m]>
Wait, autotools doesn't have a maintainer?
<jacereda>
Can I somehow rollback to a previous system version if I can't use pkgman add-repo?
<jacereda>
or to the latest version, I don't care as long as I can recover from there...
_-Caleb-_ has left #haiku [#haiku]
_-Caleb-_ has joined #haiku
vezhlys has joined #haiku
GenericConfluent has joined #haiku
Anarchos has joined #haiku
<GenericConfluent>
Does Haiku happen to have a virtiofs driver? From my poking around in the source I didn't see one.
<Anarchos>
GenericConfluent i am pretty sure there is one ?
<GenericConfluent>
That's great then
<Anarchos>
"pkgman search libvirt" --> An API for virtualization
<Anarchos>
could be that package ?
<GenericConfluent>
I see
<GenericConfluent>
I thought it was just installed by default. My mistake. I was trying to mount something in QEMU and couldn't figure out why my mount tag wasn't appearing.
<GenericConfluent>
Thanks
GenericConfluent has quit [Remote host closed the connection]
misthalu has quit [Quit: Leaving]
OscarL has joined #haiku
<OscarL>
libvirt is for mananing VMs, it doesn't provides virtio drivers of any kind, as far as I can see from its recipe.
<OscarL>
_-Caleb-_: updated squashfs_tools should be installable now on both 32 and 64 bits.
<_-Caleb-_>
Hi! Tried in 64 bits without problems... Tonight i Will try the 32 bits, thanks !!
<OscarL>
Great, thanks for reporting!
stratofax has joined #haiku
<_-Caleb-_>
Oh i can report 10 or 15 things/day ajjaja but that's not the thing ... Atm im working on compile and/or package some "small web" utils, like a gemini protocol browser or a gopher protocol browser
<Anarchos>
OscarL i thought of the link but pkgman was shorter to type, my sorry
<OscarL>
_-Caleb-_: every bit counts! (I tell myself that, otherwise my contributions won't ammount to much :-P)
<_-Caleb-_>
And reporting things in Depot that doesn't work fine like dillo browser today (It cant open gemini or gopher pages)
* Anarchos
is testing his port of gtksourceview3 in order to bring coq with IDE and compcert to Haiku.
<OscarL>
Anarchos: no problem. good luck with your porting efforts!
<OscarL>
_-Caleb-_: one might first start reporting things... then you notice a thing or two you might actually fix yourself... soon you're knee-deep in a sea of .recipes, work-in-progres git branches, and half-assed bug reports :-P
<_-Caleb-_>
Hahahaha well, if with that i can help a little ... 😜
<Anarchos>
OscarL my server is just slow to recompile the whole ocaml environment to get to coq....
<OscarL>
maybe a Haiku VM over a *BSD host might be faster?
* OscarL
hides before Anarchos gets mad at him for suggesting anything but bare-metal Haiku :-P
v_harkonnen has quit [Ping timeout: 480 seconds]
<Anarchos>
OscarL why using a vm will make things better ? You suggest the haiku scheduler is not optimized ?
OscarL_ has joined #haiku
<OscarL_>
Damn hexchat, and its disconnects :-/
<OscarL_>
Anarchos: aren't you still running an ubuntu VM on QEMU from a Haiku host?
<OscarL_>
that's why I semi-jokingly suggested running QEMU over one of the *BSDs instead (as you didn't liked my previous suggestion to do so from a Linux host)
v_harkonnen has joined #haiku
<OscarL_>
if you're just compiling on native Haiku... disregard my comment.
<Anarchos>
OscarL_ i wait for nvmm to land to run ubuntu again on qemu from my haiku server
<Anarchos>
OscarL_ i compile on native haiku :)
ghishadow has quit []
<Anarchos>
and on tuesday i will be given a decent laptop too : mine (which i only use to ssh to my haiku server nowadays) is a toshiba from 2011....
<OscarL_>
albeit Haiku *IS* much slower than linux compiling stuff (I bet in part due to due to BFS vs ext4).
<Anarchos>
OscarL_ yes i could never figured why bfs is so slow though so perfectly designed by Domininc Gianpaolo
<OscarL_>
no design is free of tradeoffs.
OscarL has quit [Ping timeout: 480 seconds]
<stratofax>
It looks like the nightly build server might be offline? "ping: s3.wasabisys.com: Name or service not known"
<Anarchos>
OscarL_ you think it is due to bfs ? i thought more on the scheduler ?
<Anarchos>
stratofax wasabi is not hot anymore (sorry for the joke)
<OscarL_>
Anarchos: try compiling the same code from bfs, vs an ext4 partition vs one from ramfs (/var/log/shared_memory), and share the results? :-)
<Anarchos>
OscarL_ i have no ext4 partition, and i don't know how to use ramfs
<Anarchos>
OscarL_ did you do the test yourself ?
<OscarL_>
Anarchos: a BFS partition without query support will surely be faster, but who can live without queries?
<OscarL_>
Anarchos: ramfs is already there, just copy stuff under /dev/log/shared_memory/
<OscarL_>
Anarchos: I haven't compared BFS vs ext4 BOTH from Haiku, only Haiku/BFS, Linux/ext4.
<Anarchos>
OscarL_ did you test with ramfs ?
<stratofax>
Anarchos I can't reach any of the repo servers either, like eu.hpkg.haiku-os.org
<Anarchos>
stratofax pkgman update works here
<OscarL_>
Yes, and it was faster, but my tests was not exactly apples to apples (because I was running haikuporter over BFS, vs plain make from RAMFS).
<Anarchos>
OscarL_ ok. we should test with bfs without queries too
<Anarchos>
OscarL_ personnaly, i don't user query very much, i have no use case for them :(
<OscarL_>
for HaikuPorts work... I cannot live without queries.
<Anarchos>
OscarL_ what do you put in your queries ?
<OscarL_>
mostly filenames / project names
<OscarL_>
so I can easily jump from one .recipe to another.
<OscarL_>
header files too.
<OscarL_>
say... `qn endian.h` (qn is in my profile/bashrc as "query -a -name $@"), then I just click on the link from Terminal to open it.
HaikuUser has joined #haiku
<OscarL_>
sorry, that's: "query -a -name=$@"
<Anarchos>
OscarL_ in Terminal i use "query myfile" and it works well. why do you add '-a name=' ?
<OscarL_>
Anarchos: name= you might drop... but -a makes it search in all "volumes".
<OscarL_>
say... `query endian.h` only finds stuff under $HOME for me.... while `query -a endian.h` finds the headers under /system too.
HaikuUser has quit []
<Anarchos>
OscarL_ ah yes -a i used it from many days now
<Anarchos>
didn't know about it before...
<Anarchos>
oh hello jessicah, i did'nt see you there from a long time
<OscarL_>
_-Caleb-_: thing I'm toying with... Do you recognize the file I have open there? :-) https://ibb.co/PtZFtnh
<_-Caleb-_>
Hahahaha yessss
<_-Caleb-_>
Ure working on a beezer compstibility???
jacereda has quit [Quit: Vision[]: i've been blurred!]
<OscarL_>
working already (at least just for listing the content :-D)
<OscarL_>
Will surely need some polishing by either Teknomancer and/or augiedoggie, but... all in all... it was much easier than what I expected.
<OscarL_>
adding at least extraction of individual files would be my goal. Others can add the rest of functionality. (specially considering my short attention spans :-P)
<OscarL_>
Beezer has been around a *looong* time already.
<OscarL_>
I'm just toying around by trying to add SquashFS support for it.
n0toose has joined #haiku
n0toose has quit []
<Anarchos>
OscarL_ ok
<Anarchos>
OscarL_ one of my leftover work is an unfinisehd rehearsal of BeTeX. the point which annoys me the more is syntax coloring is really slow on big files
<Anarchos>
don't know why and how i could improve it
<Anarchos>
though i am not the original writer of BeTeX
<OscarL_>
Anarchos: even big projects a Scintilla editor can have problems with syntax-coloring on big files (at least on some edge cases). And even the very best editors tend to disable syntax-coloring for "large-enough" files.
<Anarchos>
OscarL_ my troubles begin at ~1000 lines which is short to me
<OscarL_>
guess different algorithms/methods can have radically different complexities. Maybe betex uses a naive approach that has, say x2 / x^n complexity vs something more complex, but more efficient? (I only have a *very* shallow graps on these things... welp... in most things :-P)
HaikuUser has joined #haiku
gouchi has quit [Remote host closed the connection]
<zdykstra>
Evening all
* OscarL_
tries Genio on a "real" project (Beezer)... Very cool! (even if a tad slow on the VM)
<zdykstra>
How's life, OscarL_?
<OscarL_>
hello zdykstra!
<OscarL_>
Still too hot of a summer, but otherwise, can't complain much :-), you?
OscarL_ is now known as OscarL
<zdykstra>
I start a new job on Monday, so I'm having a weekend of relaxation before I dive into the deep end
<Anarchos>
hello zdykstra
<zdykstra>
Hi Anarchos :)
<Anarchos>
zdykstra are you dutch ?
<OscarL>
cool! Congrats on the new job zdykstra! Hoping it goes very well for you!
<zdykstra>
American, but both sides of my family are Dutch
<zdykstra>
I know of distant relatives in Holland
<zdykstra>
And just like the "real Dutch" I'm tall :P
<Anarchos>
zdykstra your name remembered me of Dijkstra, the famous dutch computer scientist
<Anarchos>
Begasus[m] coqide built with gtksourceview3 :)
<Anarchos>
time to remotedesktop to test it :)
coolcoder613 has joined #haiku
dovsienko has quit [Quit: Leaving]
v_harkonnen has quit [Ping timeout: 480 seconds]
kajedev has quit [Quit: Leaving]
zard has joined #haiku
<Anarchos>
i can't make a recipe put the resulting lib in /system/lib, (but it works for the devel package to put the lib_devel in /system/develop/lib). Can someone help me ?
<OscarL>
post your recipe on bpa.st or 0x0.st, maybe I (or others) can spot something.
<OscarL>
or link to your repo, if you pushed your changes somewhere.
<OscarL>
Anarchos: not the main issue, but on your "REQUIRES_devel" you have: "gtksourceview$secondaryArchSuffix == $portVersion base". I think you need to change that to: "gtksourceview3$secondaryArchSuffix == $portVersion base" (notice the added "3").
v_harkonnen has joined #haiku
<Anarchos>
ok
<Anarchos>
i can try to runConfigure", like "--libdir=$libDir" too ..
<OscarL>
shouldn't be necessary, unless "make install" is placing the libgtksourceview.so in some weird place (if that's the case, it is most likely NOT paying attention to --libdir).
<OscarL>
I guess you could make the recipe stop after "make install", so you can inspect where the files end up under /work-dir/packaging/
<Anarchos>
OscarL how to make the recipe stop after make install ?
<OscarL>
just push something that break the bash script :-D. I use "exit 1", or "error"
<OscarL>
s/push/put/
<OscarL>
as in... right after "make install" add a line containing "STOP!!!". that should also work :-P