<nekobot>
• Begasus (50dcbcef): nextcloud_client, fix provides for 32bit (#12680)
<nephele>
what's up you two? :)
<Begasus[m]>
not much here :)
<Begasus[m]>
can't update openjdk on 32bit, on another laptop building Haiku 32bit ...
HaikuUser has joined #haiku
HaikuUser has quit []
<nephele>
I'm wondering if we can try to do some fancy cross compiling in haikuports, then we can use the 64bit image to compile everything
<Begasus[m]>
well, there have been attempts on secondary arch for 64bit, but those are halted afaik
<Begasus[m]>
not really crosscompiling, but could open the door?
BrunoSpr has joined #haiku
<nephele>
Nah, that's something else i think. I mean secondary arch for installing packages for other haiku arches would be very cool, but on the other hand i'm not sure if we want to keep doing the secondary arch aproach
<nephele>
it's somewhat annoying :)
<BrunoSpr>
Hello all, just saw QMMP is 2 times in HaikuDepot! Waddlesplash is sorting the Depot?
<nephele>
I'm a bit amused at qmplay2, it's my goto software to play videos from youtube, but on debian it's not even packaged
<Begasus[m]>
lol
<Begasus[m]>
what's stopping you on building it on Debian? :)
<Begasus[m]>
afk for a bit :)
<nephele>
nothing. They provide an appimage i can use, so that is satisfactory. I'm just used to getting all software i want from a package manager
dodo75 has quit [Quit: Vision[]: i've been blurred!]
<BrunoSpr>
nephele, I love Qmplay2, it is so easy to search, play and download YouTube videos and music
<BrunoSpr>
Begasus[m], do not ask him to work for Debian! We need him here!
dodo75 has joined #haiku
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
<Begasus[m]>
re
<nephele>
welcome back Begasus[m]
<Begasus[m]>
;-)
OscarL has joined #haiku
<OscarL>
Morning, folks.
<Begasus[m]>
Hola OscarL :)
<janking>
Morning
<OscarL>
Begasus[m]: Hey there :-). Looking at repo_consistency.txt report, starting with "requires" issues first...
<Begasus[m]>
idea_community_bin this doesn't work on 32bit due to old openjdk, but it's "any", so we can't have one for 32bit (older version) and the current
<Begasus[m]>
cool OscarL !
<OscarL>
no idea why "requires "lib:libxcb" of package "retext-8.0.2" could not be resolved" :-/
<Begasus[m]>
yeah, there are a few in there
<OscarL>
downloaded the libxcb .hpkg, and both the .so and the provide look OK there :-/
<OscarL>
Mmm, and "regular" report.txt (for x86_64 at least) is now clean. So that retext issue gets even more strange :-D
<Begasus[m]>
right, on 32bit there is only the issue for idea_community_bin
<OscarL>
Begasus[m]: to keep things simple... I'll just try rebuilding libxcb with only change being the switch to Python 3.10 (to avoid possible issues if we were to also update libxcb to latest version).
<OscarL>
Sounds good (for now at least)?
<Begasus[m]>
ah!
<Begasus[m]>
good catch :)
<OscarL>
there are not that many packages that directly depend on libxcb, but I don't have the bandwidth/CPU to test them all.
<OscarL>
fixing that one should clear a bunch of errors, I think.
<Begasus[m]>
I wouldn't update there just yet, one of the requirements for xlibe
<OscarL>
+1
<Begasus[m]>
yeah, didn't look that far :)
<OscarL>
will get into it then, and see what else breaks (or if we improve things a bit at least :-D)
BrunoSpr has joined #haiku
<Begasus[m]>
maybe (as with other recipes) it doesn't even need python :)
<OscarL>
will check.
<Begasus[m]>
+1
<OscarL>
mmm, libxcb has a patch for gcc2, but the recipe is !x86_gcc2 :-D
<Begasus[m]>
probably left-over :P
<OscarL>
yeah, will drop that.
<nephele>
how am i supposed to aprove users on the forum now
<OscarL>
"I am nephele_OK and I approve of this ~message~ user!"
<nephele>
like shit, the forum software tells me it has to be aproved, but the user did not get to post anything i should review
<nephele>
so now what? discriminate based on the username?
* Begasus[m]
is glad to not be in that position ...
<OscarL>
"IamVeryMuchNotABot"... "yeah, that sounds trustworthy!"
<nephele>
<firstname><lastname><numbers>
<nephele>
Could be an AI.... but! some people also use that style
<nephele>
so it's not proof of anything
<OscarL>
"JackBurton76" :-P
<nephele>
This is a bit icky. Now I could aprove the user but then wait 2 hours what they do? but i'd rather not sit infront of the forum for that long
<nephele>
i'd rather the forum tells me the user is allowed in the first place, but then flags his first post for review automatically
<Begasus[m]>
spotted error in idea_community_bin :P
<nephele>
community bin sounds like something where you put your waste
<Begasus[m]>
java waste in this case then :P
<nephele>
waste from the island of java?
<Begasus[m]>
hmmm ... coffee :D
<nephele>
good idea. i made myself coffee... brb
<Begasus[m]>
grr ... download this monster again ...
<nephele>
i started the zig sdl3 thing yesterday, it's a bit hard to get into, luckily they have a template i can use for this exact thing
<Begasus[m]>
maybe I'll make it "all !x86_gcc2" despite it being "any"
<nephele>
in lua i wrote my UI code completely on my own, wondering if i should just use a premade UI library instead
<Begasus[m]>
and then revive the version from last year for 32bit
<Begasus[m]>
if possible, why invent the wheel nephele ?
<nephele>
Layouting code is fun Begasus[m]
<Begasus[m]>
that could be enough of a reason :)
<nephele>
and it sometimes is much more performant to write it in that context, but the more complex the code gets... eventually the library gets better ;)
* Begasus[m]
still gets shivers when reading lua
erysdren has joined #haiku
<Ellenor>
why? :(
<Begasus[m]>
packaging issues Ellenor ;)
<nephele>
lua, redefine everything :D
<nephele>
Begasus[m]: that's one part i don't really get tbh. I wrote lots of lua code, but i never used any lua packages
<nephele>
don't even know how they work
<OscarL>
Begasus[m]: "configure: error: no suitable Python interpreter found" (at least we verified it needs it :-D)
<Begasus[m]>
current state should be enough, at least for me nephele :)
<Begasus[m]>
now make it work with python3.10 OscarL :P
<nephele>
don't worry Begasus[m], if i use lua for my zig stuff now i'll probably just include the lua source directly...
<Begasus[m]>
I hope it does (seeing it still requisted python2.7) :(
<nephele>
vendoring deps makes sense for a game, kind of. makes it predictable
<OscarL>
configure script found python3, but build failed with a complain about a missing xcbgen Python package, lol.
<nephele>
I used to really dislike python code creeping into linux distros.... but now i see haiku also has problems with it
<nephele>
granted, i did not know what a packaging nightmare it was back then
<nephele>
maybe haikuporter should not be in python so we have less dependencies overall...
<nephele>
(for the "core" anyhow)
<erysdren>
snek
<Begasus[m]>
fwiw, I think haikuporter tool is pretty fine, not perfect, but does the job
<nephele>
I don't mind it's commandline interface, except for me not liking the commandline in general... though it's inards could use some cleanup. :P
<nephele>
also we don't have a lot of people who can maintain it, which seems to be a problem aswell
<nephele>
like that "waiting for packages" thing beein unadressed for a really long time but also beeing quite a time drain on the haikuports team
<OscarL>
Python is a *really* good "glue" language (far better than bash at the very least), so it gets used for lots of tooling. Problem (in linux and elsewere) comes when people start installing/using/depending on "system" Python for their own develpment, and then complain when things are missing or their system breaks.
<Begasus[m]>
don't think I'll survive a rewrite for the tool and another way for creating the recipes :P
<OscarL>
(python packaging still a mess)
<Begasus[m]>
+10 there :P
<erysdren>
i use python for things that 10-15 years ago probably would have used a set of bash scripts for
<OscarL>
but as with C++, we only complain about languages that are actually used.
<erysdren>
er, i'm just guessing at that. i haven't even been programming for 10 years ;~;
<nephele>
bash is not that great a language either
<OscarL>
both trailblazer in their own, got a lot right, and a lot wrong along the way :-D
<Begasus[m]>
biab
<OscarL>
(with newer languages learning from other's mistakes)
<erysdren>
i kinda hate bash for anything more complex than running a program with a preset list of arguments
<nephele>
hehe
<nephele>
OscarL: the classic "good enough i guess" in IT always wins
<erysdren>
hence why i use python for tools that i just need to make in 5 minutes
<erysdren>
if i need other people to use the tool, i will write it in C
<nephele>
we kinda need a better scripting language for haiku...
<erysdren>
shrug
<nephele>
something that lets users write scripts to automate little things without accidentally breaking everything
<nephele>
Like figuring out how much to quote arguments in bash ;)
<erysdren>
*a new standard has been added to the bottom of the list*
<nephele>
:D
HaikuUser has joined #haiku
<erysdren>
programmers get that on their HUD as a negative status effect any time they try to invent a new standard rather than using something proven
HaikuUser has quit []
<erysdren>
LOL
<nephele>
Well, if the "proven" thing is something like bash, uhhh....
<erysdren>
:P
<erysdren>
i aint gonna sit here and defend bash, thats for sure :3
<nephele>
I mean, if you look at like the iOS shortcuts app, it is purpose build for that, and it works fine
<nephele>
i would probably go insane if i had to do the same thing with bash or python
<nephele>
i mean yes, it's graphical, but it's still a programming language
<nephele>
(even if it's gui is sometimes really unclear)
<nephele>
We have hey, maybe with the same interface we can write something graphically that is clearer
<erysdren>
if it were me, i'd look around for a good bash alternative (or a general scripting lang) ideally something that is friendly to the Haiku way of life
<erysdren>
rather than try to invent something new immediately
<nephele>
Sure, if something is available that makes sense
<erysdren>
but i'm no stranger to doing it myself if i get frustrated with what's on offer, don't get me wrong
<erysdren>
lua is a fine scripting language but the stdlib is kinda garbage, i've never used it for quick tooling
<nephele>
I don't know a good alternative to "bash", but the problem, as i see it, is not only bash itself but the poxis way of interacting with apps aswell
<erysdren>
again i just use python for that
<erysdren>
ah
<Habbie>
OscarL, +2 fuck yeah
<erysdren>
yeah, good point
<janking>
:)
<nephele>
Like, you can't have good localization this way, the disctiction of short and logn options, some toold using the even other style like dd, windows tools using / flags and that sometimes spilling over etc
<nephele>
I think this needs a more consistent interface, to not have to deal with these details at all... and when you have that interface you can make a graphical and a textual programming/scripting language ontop of that
<nephele>
or maybe adapt existing ones
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
<nephele>
so maybe we can have a command named "copy", with an internal name, but also show it in a localized interface....
<OscarL>
erysdren: there's the (python based) xonsh shell. Tad slow on my slow hardware. Should probably give it another go... but then again... "bash inertia" gets in the way of that being a good investment of my limited attention spans :-D
<nephele>
PulkoMandy: what is the new config for the forum? I see i have to aprove a new user now before they can post
<PulkoMandy>
yes, since no one seems interested in fixing antispam systems, I re-enabled registrations with moderator approval
<PulkoMandy>
at least we don't forget about it as the forum will keep asking us to validate new users...
<PulkoMandy>
I don't know what else to do
<nephele>
The thing is, i don't really have much to go on to aprove or deny the new user
<PulkoMandy>
not really interested in digging into forum antispam configuration at the moment
<nephele>
they have a very generic nickname with <first><last><numbers> but that isn't proof either
<nephele>
I am wondering, if i aprove them now, does their first post get flagged for moderation still?
<PulkoMandy>
I think we can just approve tiem. At least they can't create a hundred accounts nad posts while admins are not around
<nephele>
hmm okay. but that probably means i have to keep a bit of an eye on the forum when i aprove accounts
<nephele>
but it would still stop them from adding 25 at a time to spam indeed
<PulkoMandy>
yes, I don't know what's best
<PulkoMandy>
I hoped there would be more replies to humdinger's mail about it in the mailing list...
<nephele>
well, we could change the rules for the first trust level 0
<PulkoMandy>
I'm not really interested in handling this. I made the emergency measure to stop registrations. Feel free to set up something better or find someone who wants to do it :)
<nephele>
I can't, i don't have administrative acces :)
<nephele>
I don't recall the rules, and not sure how to check, but iirc getting out of trust level 0 to trust level 1 was quite easy... so maybe the trust level 0 post need aproval, but after 2 or 3 posts you'd leave level 0 automatically...
<nephele>
in any case, i will aprove this new user now and then check if they act maliciously...
<waddlesplash>
PulkoMandy: what mail from humdinger are you referring to?
<waddlesplash>
OscarL: we should not have an xcb package
<waddlesplash>
it should be disabled because we don't have a real X11 server
<PulkoMandy>
"Firum gets spam bombed" on haiku-development mailing list, yesterday at 08:25 western europe time
<PulkoMandy>
I replied that I closed registrations to the forum temporarily but I don't know what to do in the long term
<PulkoMandy>
we surely don't want to keep that closed?
<PulkoMandy>
(reopened this morning with moderator approval for all new registrations, but as nephele said, approving registrations based on just username and email is not great)
<OscarL>
waddlesplash: makes sense. Was initially suspicious of it (specially because it seems currently "required" by stuff that already uses Qt), but... what do I know :-D
<OscarL>
Begasus[m]: xcb_proto providing devel: sounds a bit off, and I think it should instead provide a "_python3x" package, I think. (assuming we nead them at all, as waddlesplash points out).
nephele has quit [Ping timeout: 480 seconds]
nephele has joined #haiku
<nephele>
PulkoMandy: i replied with my suggestion on the ML
<nephele>
But as i said, i can't implement it, someone with administrative access (on the forum, not the machine i mean) will have to do that
<OscarL>
Begasus[m]: if you have time... maybe you could try to see if the packages listed via `pkgman search -r libxcb` can be built/used with their libxcb requirement removed?
<nephele>
funny that now the forum times out when i try to go to it...
<OscarL>
Begasus[m]: bothering you with this because libplot, pcmanfm_qt, and retext all require Qt/KDE stuff which I don't have.
<nephele>
does IRC have a way to ack messages? Vision does this pinging stuff to test the lag, and it does reliably tell me that i lost connection.... in the status bar, but i often miss this and as a result loose some messages i send, and am confused why the channel is quiet
<OscarL>
s/libplot/labplot/
<OscarL>
nephele: happened to me a couple of times... Vision just left me talking into the void. Surely to the relief of the channel's regulars :-D
<nephele>
haha :) I don't think any of the regulars here are annoying
<nephele>
maybe i should patch vision so it puts a scary connection problems line in the backlog?
<nephele>
that would be better than nothing
<OscarL>
yeah... a red scary message in the (currently focused) chat window would be a big improvement, IMO.
<Begasus[m]>
re
<Begasus[m]>
let me catch up :)
<nephele>
okay, let me see if i can whip that up, but someone else will have to submit the patch :)
<OscarL>
bet either Begasus[m] or I can push it on your behalf, nephele.
<OscarL>
(merge will need someone with proper rights, of course :-D)
<Begasus[m]>
hmm retext only has it in requires OscarL
<OscarL>
extra odd, yes. did a search for libxcb on retext upstream, got a couple matches, but no idea if it is needed at all for us.
<Begasus[m]>
class _xcb_reply_t(ctypes.Structure):
<Begasus[m]>
in the source it is ... (xsettings.py)
<Begasus[m]>
bah: pyqt6_python310 >= 6.8.0
<Begasus[m]>
need to rebuild this for 6.9.0 I guess :P
<OscarL>
"xcb_library_name = ctypes.util.find_library('xcb')", and raises an exeption if it can't find it :-(
<Begasus[m]>
OscarL: I'll remove/uncomment it in the recipe/push/merge and then check this in qemu here I guess
<Begasus[m]>
eeps
<Begasus[m]>
did you investigate libxcb on newer versions OscarL ?
<OscarL>
no, got side tracked by that "missing xcbgen" python package (that works OK outside of the chroot), but then waddlesplash commented that we should not even have xcb at all, so...
<nephele>
> 'class BMessenger' has no member named 'SendMEssage'; did you mean 'SendMessage'?
<nephele>
maybe i men SeNdMeSsaGe
<nephele>
brb
nephele has quit [Quit: Vision[]: i've been blurred!]
<Begasus[m]>
argh .. WebPositive keeps freezing on gitlab :/
nephele has joined #haiku
<OscarL>
re: retext... will try to see if the XSettingsError exception is fatal, or it can just live without those settings.
<nephele>
hello from new vision, let me just break my connection to see if my new code works :D
<OscarL>
Begasus[m]: seems to me (just by doing a quick read of the code on github), that retext should "survive" just fine if it libxcb is not present.
<nephele>
it doesn't .-.
<OscarL>
Begasus[m]: AFAIK, it should only fail to find an "icon theme" (via xcb, gsettings, or GTK), and that only if it fails to find one from Qt.
<Begasus[m]>
checking libxcb-libxcb-1.14
<OscarL>
Sounds like you should be fine just dropping the xcd.
<Begasus[m]>
bah ... configure: error: Package requirements (xcb-proto >= 1.14) were not met:
<Begasus[m]>
let's drop it then and check in VM
<OscarL>
given the waddlesplash comment, I wouldn't bother (other than do our best to drop xcb as a requirement everywhere)
<nephele>
what do you ned xcb for?
<OscarL>
I don't. some recipes have it as require or build-require for some reason or another.
<nephele>
unless you are working with something that actually uses X11, for example to connect to a remote X11 server, this should not be needed at all
<nephele>
(and that's such a niece usecase i don't think we need to cater to in haikuports, you can probably just use vnc instead...)
<nephele>
nishe? like small not that relevant? i don't know how to spell that
<OscarL>
(probably just because they were mentioned as requirements on readmes from upstream, and packagers didn't knew enough about it to make a judgment call.
<OscarL>
*niche
<OscarL>
(I thinkg :-P)
* OscarL
can't spell/type right anyway
<nephele>
I can't wait for a proper imap inbox app
<nephele>
getting notifications because *I* wrote to a mailing list is a bit annoying
<Begasus[m]>
OscarL: which comment?
<OscarL>
"[08:27] <@waddlesplash> OscarL: we should not have an xcb package"
<Begasus[m]>
ah, missed that one I think :)
<Begasus[m]>
k, let's just drop it and see where it lands (for retext)
<OscarL>
should be fine! (famous last words)
<OscarL>
afaik, it is just used as a "backup" way of trying to figure out which icon theme it should use, if it fails to find a Qt provided one.
<nekobot>
• Begasus (3ae9587e): retext, revbump, drop libxcb dependency (#12681)
<nephele>
it can't suceed, so why bother? :P
<Begasus[m]>
will find out soon enough :)
* OscarL
likes removing "dependencies" (sometimes.... a bit too much :-D)
<Begasus[m]>
heh
<Begasus[m]>
OscarL_Sledge :P
<nephele>
my patch doesn't work OscarL :(
<Begasus[m]>
(hope there isn't one with that nick on matrix) :D
<OscarL>
:-( bad Vision... should call some Avengers to help out making it behave.
<nephele>
no wonder with code with names like "theNick"
<Begasus[m]>
one down in the consistency list OscarL
<OscarL>
\o/
nephele has quit [Quit: Vision[]: i've been blurred!]
nephele has joined #haiku
* nephele
tries a new patch
<nephele>
man, it still doesn't work :(
<OscarL>
can't relate, all my patches work right the first time. (then I wake up and face reality).
<Begasus[m]>
/sources/labplot-2.12.0/build/src/3rdparty/Qt-Advanced-Docking-System/src/qt6advanceddocking_autogen/EWIEGA46WW/../../../../../../../src/3rdparty/Qt-Advanced-Docking-System/src/ads_globals.h:44:10: fatal error: xcb/xcb.h: No such file or directory
<Begasus[m]>
there is something with that advanced-docking ...
mmu_man has joined #haiku
nephele has quit [Quit: Vision[]: i've been blurred!]
nephele has joined #haiku
<nephele>
OscarL: damn, maybe you should have written the patch then ;)
<OscarL>
;-P
<OscarL>
Begasus[m]: will try to take a look at the sources. At least phabricator works way faster than gitlab.
nephele_ has joined #haiku
nephele is now known as Guest23529
nephele_ is now known as nephele
Guest23529 has quit [Remote host closed the connection]
<OscarL>
in that case, xcb_proto should be renamed to xcb_proto_python310 (or even xcbgen_python310), because it only seems to provide that, and a bunch of .xml files under $dataDir.
<OscarL>
in any case, xcb_proto only seems needed by libxcb_devel, so maybe we should just "Drop all the things!".
<nephele>
probably yeah
<OscarL>
issue is... someone will need to patch at least labplot to drop the requirement for xcb at build time
<nephele>
is that problematic?
<Begasus[m]>
maybe ... :)
<nephele>
I mean, i think you said it has a macos codepath, so that should be activateable by setting the right defines
<Begasus[m]>
almost there (I think)
* Begasus[m]
needs focus ... not checking the forum atm ...
<Begasus[m]>
but kglobalaccel isn't tied to dbus iirc nephele
<nephele>
you just said you didn't want to check the forum :P
<Begasus[m]>
poppups appearing :P
<nephele>
just close the webrowser!
<Begasus[m]>
eg forum notifications
<nephele>
:)
<Begasus[m]>
I need it :P
<nephele>
close the tab, then? :P
<Begasus[m]>
°_0
<Begasus[m]>
:P
<Begasus[m]>
WebPositive/Web/Iceweasel open ... which one? :P
<nephele>
just close the discourse tab in webpositive probably
<nephele>
i don't know if the others have notifications
<OscarL>
Begasus[m]: I think you should be able to wrap most (all?) of the code in abs_globals.cpp with "#if defined(Q_OS_UNIX) && !defined(Q_OS_MACOS)", as the .h already only declares those functions conditionally.
<Begasus[m]>
only if I have the forum open on it :)
<Begasus[m]>
already done OscarL and a few AND HAIKU in a cmake file
<OscarL>
+1
<Begasus[m]>
bugger ... ContainerE]+0x38): undefined reference to `ads::CFloatingDockContainer::event(QEvent*)'
<Begasus[m]>
although building with ninja is fine, I always find it hard to track those errors
<OscarL>
heh, I missed the #if defined(Q_OS_UNIX) && !defined(Q_OS_MACOS) in line 55 of ads_globals.cpp. Conditionals should be made easier to spot... just a couple of newlines would help there :-D
<nephele>
I apreciate builds that just put OS specific implementations in seperate files instead of ifdefing everywhere
<OscarL>
need to patch instaweb to use #!python3, or maybe we should make "default" Python provide /bin/python already. Will save some headaches, might cause some with older recipes last built using Python 2.x
<Begasus[m]>
still stuck on this one ... ContainerE]+0x110): undefined reference to `ads::CFloatingDockContainer::resizeEvent(QResizeEvent*)'
<nephele>
i think that is the reason why it's python3 and not python
<nephele>
linux does it similarilly iirc....
<OscarL>
those should be defined in ads_globals.cpp
<OscarL>
^Begasus[m]:
<OscarL>
nephele: yes, but some already "switched" /bin/python to mean 3.x (after a period of having dropped 2.x)
<OscarL>
Begasus[m]: line 53/54, right above (and thus outside) the "#if defined(Q_OS_UNIX) && !defined(Q_OS_MACOS)". maybe you're not getting that file compiled due to other changes in cmake files, or something?
<Begasus[m]>
then I can give a look at the forum what you all have been up to in Playground :P
<Begasus[m]>
OK, had that talk also with KDE dev's
<Begasus[m]>
one thing indeed would be CI/CD backend, but I don't think the guy in charge is really looking forward to it, plenty of issues on their backends already :)
<nephele>
Not sure what the idea here is, we already build kde software automatically for haikuports
<nephele>
do they want one for each commit?
<Begasus[m]>
it's their way of making sure nothing is broken on the ports
<OscarL>
patch looks good to me on a first read. *maybe* some of those " || defined(Q_OS_HAIKU)" are not strictly necessary, but hard to tell just from the patchset.
<Begasus[m]>
but for instance I've seen plenty Fedora CI pipelines failing on branch changes
<Begasus[m]>
yeah, I just added every where I could find that reference OscarL :P
<OscarL>
as in... the may not be related to xcb at all, and maybe Haiku is not exactly behaving as macOS, but... who knows, it might even improve things :-P
<Begasus[m]>
those undefined references weren't all related to xcb also :)
<Begasus[m]>
those files in src/linux also fail to build, so left them out too (as is done for MACOS)
<OscarL>
Begasus[m]: could request a review from 3dEyes.
<Begasus[m]>
could do yes
<Begasus[m]>
grabbing labplot-2.12.0-1-x86_64.hpkg and moving it to /Opslag/haikuports/packages/labplot-2.12.0-1-x86_64.hpkg
<OscarL>
it builds? ship it!
<nephele>
the forum software is so annoying
<Begasus[m]>
it already build (also runs) :P hence I did a clean rebuild
<nephele>
you write an emoji, it corrects it to it's other thing, and then it sends out an email where it sais ":smile:" while the original emoji would have been just fine in the email
<Begasus[m]>
so not sure if a review is really needed
<nephele>
reviews are to double check and increase the quality :)
<OscarL>
even a mention "after the fact", to bring people attention to a particular issue, sounds good to me, if we want to "merge early" (people might not have time to review it anyway :-D)
<Begasus[m]>
PR is up
<OscarL>
+1
<OscarL>
mmm, git's instaweb script needs better patching than what I can currertly do, to make it work with python as webserver. it seems to try to create symlinks in a read only location :-D
<Begasus[m]>
woops :P
<nephele>
linuxes software aproach is fun
<nephele>
-> create symlink -> filesystem is readonly
<nephele>
-> try again with root
mmu_man has quit [Ping timeout: 480 seconds]
<OscarL>
(now has broken symlinks... sigh.)
mmu_man has joined #haiku
<OscarL>
mmm, somehow I have /packages/git-2.48.1-7/.self/data/gitweb/static, but the symlink gitweb generates is for /packages/git-2.48.1-1/.self/data/gitweb/static (resulting in a broken link... notice the -7 vs -1 revision)
<OscarL>
duh. guess I should update my git. have git-2.48.1-1 installed earlier than gitweb I installed today "2.48.1-7". :-(
<nephele>
the revision shouldn't matter then? make it point a relative symlink instead
<OscarL>
errh, the other way around... have git 2.48.1-7, and git_web 2.48.1-1
<OscarL>
nephele: plenty of those edge cases to find out and fix, yes.
<nephele>
yeah but those revisions should be independant of another
<nephele>
it should ideally make a relative symlink based on the structure in /system and not care about /packages
<OscarL>
yes, I had already complained about packages using "versioned" paths in the past, at it tends to brake things in subtle but annoying to debug ways.
<OscarL>
LC_ALL overriting the rest of the LC_* makes sense, but if one calls setlocale() for a specific one, the others should not be affected/touched, that would be my expectation. On Haiku, that only seems to old true if you first call setlocale("LC_ALL") at least once (which is not the case on Linux, in my testing)
<Begasus[m]>
Could NOT find KF6 (missing: Crash) (found suitable version "6.16.0",
<OscarL>
"LC_TIME is the only LC_* envar besides LC_ALL that changes the language in output:" weird bug we have there :-D
<Begasus[m]>
this consistency report is way to smart I think :P
<Begasus[m]>
brake*
mmu_man has quit [Ping timeout: 480 seconds]
nephele_ has joined #haiku
nephele is now known as Guest23537
nephele_ is now known as nephele
Guest23537 has quit []
Nukacal has joined #haiku
Nukacal has quit [Quit: Going offline, see ya! (www.adiirc.com)]
mmu_man has joined #haiku
<OscarL>
Begasus[m]: weird that libfm-qt changelog mentions "Removes XCB from the dependencies" (at least for the 0.11 versions)
<OscarL>
seems that meant "remove it from pkg_check_modules() and try to find it via cmake instead".
<OscarL>
"Done! Took 33037ms, 464002 iterations" this sites using Anubis with dificulty 4 are getting annoying :-D
<OscarL>
s/this/these/
<nephele>
too difficult for your pc?
<nephele>
someone suggested anubis aswell for the forum, but i responded that it wouldn't be a good idea as it is now... :)
<PulkoMandy>
I expected to keep my phone for several more years, but that may end up being what makes it unusable :(
<PulkoMandy>
Or maybe I'll just stop using websites
<nephele>
Maybe we can all switch to gemini ;)
<nephele>
I actually tried to get into it, but the "emulation" of stuff like messageboards in it is so jarring, it's hard to read
<PulkoMandy>
I bet ipv6-only would block the bots just as well, and not be as limiting as gemini
<OscarL>
nephele: my best CPU is Phenom II X4, when it worked... anubis at dificulty 4 was "OK" only if the power settings was "performance", when using power saving (as the CPU eats watts for breakfast)... was as slow as in this example.
<nephele>
PulkoMandy: we could tentatively enable that to only require it for ipv4 and see if that is sufficient
<OscarL>
now using a more efficient, but abot half-slower N4020 CPU :-D
<x512[m]>
I have no IPv6...
<PulkoMandy>
I have ipv6 at home but not at work, I think
<nephele>
PulkoMandy: there is also a way to classify ipv4 adresses for residential vs commercial i think, we could also try that distinction
<nephele>
and then adjust the decision later if it ever becomes a problem
<OscarL>
(no ipv6 down here either.)
<PulkoMandy>
My workplace ip is probably not residential. So that would block me :)
<nephele>
The only place where i don't have ipv6 is my haiku install....
<nephele>
PulkoMandy: no, it'd just make you have to use anubis at work ;)
<nephele>
also we should short circuit this for logged in users, one less time to have to use anubis
<x512[m]>
IPv6 works here only with some proprietary ISP router that I do not want. IPv4 works with standard PPPoE protocol.
<PulkoMandy>
I may need the newer version, it would be easier if I can start from existing recipes and not have to gather all the patches and stuff from the github ticket :)
<Gnter[m]>
<nephele> "PulkoMandy: no, it'd just make..." <- Anubis is such an annoying and useless thing to add. Can be bypassed by changing the user agent lmfao and makes it so real browser visitors have to wait and run their javascript.
<Gnter[m]>
However just doing it for v4 or whatever would be fine I guess lol (who cares, get on the current internet protocol).
<Gnter[m]>
But are bots really that big of an issue that something like that is even necessary?
<waddlesplash>
yes
<waddlesplash>
adding Anubis in front of Trac, traffic *massively* decreased
<waddlesplash>
went from "Trac is crashing due to getting slammed by bots", to "Trac doesn't crash and is faster than ever"
<waddlesplash>
and this after we already had done a number of ANS blocks
<nephele>
Gnter[m]: it does what it's supposed to, so it's objectively not uselsss...
<x512[m]>
I currently see no benefits of using IPv6, especially via VPN.
<nephele>
even on the forums... 10% is views by logged in users 10% logged out ones 50% "known" crawlers and 20% other traffic
<nephele>
If you take a positive estimate we pay 70% bandwidth and performance for robots...
<nephele>
waddlesplash: do we have a uptime status page or something?
<x512[m]>
And IPv6 is not well supported in Haiku yet. As I understand, it currently need to manually specify IPv6 address.
<Gnter[m]>
x512[m]: There is a lot of benefits of using IPv6. But all but one is being ruined by the VPN anyways
<Gnter[m]>
And the remaining benefit is cost of address, which is also pretty useless in a VPN context
<nephele>
waddlesplash: it doesn't load for me. does that track the forums aswell?
<nephele>
it would atleast explain the uptime koma crawler UA
<Gnter[m]>
There would be a benefit for VPN's that allow port forwarding as they can then give you an entire prefix, which is a thing you can buy
<waddlesplash>
nephele: yes
<x512[m]>
No static public IP address is better for privacy.
<nephele>
ipv6 doesn't have static ip adresses either if your ipv4 isn't stable, speaking of residential connections...
<Gnter[m]>
nephele: Usually
<Gnter[m]>
But most places are behind CGNAT for IPv4 at this point unless you pay extra
<Gnter[m]>
And IPv4 is also starting to become a large cost for server providers and people renting servers
<Gnter[m]>
So the Mullvad trick is really only useful for IPv6 only websites
<nephele>
ipv4 is defeinetely a pain nowadays
<x512[m]>
Some ISP provide fixed IPv6 prefix and this behavior can't be changed. This means that every web site you visit, every tracker can uniquely identify you.
<Gnter[m]>
When the website owner doesn't want to pay because your ISP can't adopt the current 30 year old internet protocol
<nephele>
too bad we don't do either ipv6 adress auto method yet...
<Gnter[m]>
nephele: Also internal networking is such a breeze on IPv6
<Gnter[m]>
Just give it a /64
<Gnter[m]>
No conflicts and no having to do annoying subnetting
<nephele>
Gnter[m]: uhh? internal? like LAN?
<Gnter[m]>
nephele: Well for larger networks but yeah
HaikuUser has joined #haiku
<x512[m]>
192.168.xxx.xxx forever.
<nephele>
i don't see why you would care what the internet connection is then
HaikuUser is now known as scantysnax
<nephele>
just use a network local ipv6 prefix
<nephele>
since having multiple adresses per connection is easy there
<Gnter[m]>
For a home connection LAN IPv4 isn't that much of an issue as you just give it a /24 and a 10.10.0.0 or 192.168.0.0 address and increment
<nephele>
per interface, i mean
<Gnter[m]>
nephele: My home lan is dual stack, but most everything is routed over v6 with ULA's
<nephele>
our home LAN has an automatica ipv6 local area prefix, and ips for all devices
<Gnter[m]>
So it won't be a problem going single stack if ISP ever starts providing a NAT64 gateway
<nephele>
and automatic mDNS assigning of names too, and discovery... :)
<scantysnax>
good morning, good afternoon, good evening, and good night :-)
<Gnter[m]>
This Hong Kong research paper from 2012 also shows that even running NAT64 on your home router is more efficient than running NAT44
<Gnter[m]>
Because IPv6 packets can be processed easier by the router
<nephele>
I don't want to run NAT at all tbh In any case, haiku is unfortunately misssing ipv6 adress negotiatin
<nephele>
my home network has ipv6 ever since i fell for an OpenBSD prank
<Gnter[m]>
<nephele> "If you take a positive estimate..." <- Just did a whois on the IP on the website, AWS? 😹 Yeah that's gonna be expensive then
<nephele>
I don't know how it is hosted. though AWS sounds terrible
<scantysnax>
Hi there Begasus[m]! How's it going today?
<Gnter[m]>
nephele: Yeah NAT needs to die already
<x512[m]>
Stateful connection tracking is still needed for security to avoid unintended connections from Internet and attacks.
<Gnter[m]>
But ISP's should start moving from dual stack to just running NAT64 gateways
<Gnter[m]>
So I don't have to have NAT at my home network
<Gnter[m]>
It would also be cheaper for them
<Gnter[m]>
But steam on windows for example won't work with NAT64 unless you setup CLAT
<Gnter[m]>
Because they use IPv4 literals instead of domain names
<x512[m]>
NAT is nice actually. Internet can used in QEMU running on Haiku host because NAY can be implemented on top of regular sockets.
<Begasus[m]>
fine here scantysnax being a bit blurred with that new consistency report, not getting it there OscarL kio still complaining there even though devel:kcrash is added :(
<Gnter[m]>
Just werkz™ on apple products as they have it setup by default
<Gnter[m]>
x512[m]: Well NAT is not going away, it's also fine for stuff like Docker which is how it gets networking
<Gnter[m]>
But the actual host machine shouldn't be behind a NAT or excuse my language double NAT 🤢
<scantysnax>
ah, I see.
<x512[m]>
NAT stacking is an example of NAT flexibility.
<scantysnax>
i haven't spoken with OscarL in a while
<OscarL>
Begasus[m]: for the kio6 related lines on the repo_consistency.txt?
<OscarL>
Hey there scantysnax :-P
<scantysnax>
oh there he is!
<Gnter[m]>
<x512[m]> "Some ISP provide fixed IPv6..." <- Not really, every modern device will use the SLAAC privacy extension
<Begasus[m]>
yeah OscarL
<Gnter[m]>
So privacy will be about the same as IPv4 with NAT except full E2E connectivity
<scantysnax>
OscarL: i found a parallel port cable, going to try your peek/poke driver sometime this week.
<OscarL>
Begasus[m]: those sounds a bit wack, to be honest, but.... what do I know.
<Gnter[m]>
Much less resources required to do the networking and usually lower latency too
<x512[m]>
SLAAC privacy extension have nothing to do with how ISP allocates IPv6 prefix.
<Gnter[m]>
* So privacy will be about the same as IPv4 with NAT with the added benefit of full E2E connectivity
<OscarL>
scantysnax: cool, let me know how that goes!
<scantysnax>
i certainly will. hopefully i
<scantysnax>
will be successful
<Begasus[m]>
kio6_x86-6.16.0-2:
<Begasus[m]>
build-requires "devel:libkf6crash_x86 == 6.16.0" of package "kio6_x86-6.16.0" could not be resolved
<Gnter[m]>
x512[m]: No, but the IPv6 prefix is the same as the single IPv4 address you usually get allocated. When they change it is up to the ISP
<Gnter[m]>
My router says the IPv6 prefix I received has a lease time of 4 hours
<Gnter[m]>
But the prefix really shouldn't change
<nephele>
ipv6 is bad because some isps can be stupid is kind of a wild take....
<OscarL>
scantysnax: I'm slowly writting a Python "wrapper" around /dev/misc/poke. mostly to experiment with my silly EC/SuperIO/thermal reading drivers.
<Gnter[m]>
If you want to switch your address for privacy reasons beyond hiding which device is doing it then a VPN is the solution anyways
<scantysnax>
OscarL: cool. i don't even know one word of pythong
<nephele>
Gnter[m]: the ipv6 prefix changes for me when the router resets it. I also know some other routers have an explicit setting to keep the adress, change only the ipv6 prefix, or the ipv6 prefix AND ipv4 adress
<scantysnax>
python*
<x512[m]>
For IPv4 ISP is have to dynamically allocate IP addresses, so privacy is guaranteed. For IPv6 ISP is no more technically have to dynamically allocate IPv6 prefix that is scary.
<Gnter[m]>
x512[m]: They don't have to, only in a few jurisdictions, but that is also incredibly annoying if hosting anything from your home that you need access to externally like a VPN to your home servers
<Gnter[m]>
But also there is no way for random websites to know if your prefix changed and what prefix size you have anyways
<Gnter[m]>
Non-issue, main problem will be being logged into google or whatever anyways
<scantysnax>
OscarL: your driver can peek/poke the PCI bus, right?
<x512[m]>
They have to for technical reasons (limit of IPv4 addresses). Address shortage is a feature here, not bug.
<Gnter[m]>
If you want to change your IP use a VPN
<OscarL>
scantysnax: can be pretty easy. If I get my modules the way I like it..., I'll post them somewhere. Might end up "re-writing" /bin/poke in Python in the end, instead of re-writing it in C++ :-D
<Gnter[m]>
It's not that hard on any router worth it's salt to just route everything through a VPN if that's what you want
<nephele>
No, they don't. The number of adresses stays the same no matter if you rotate them or not
<x512[m]>
> main problem will be being logged into google or whatever anyways
<scantysnax>
interesting. i wish i knew some python
<x512[m]>
I never use login to Google etc.
<x512[m]>
It is just absurd.
HaikuUser has joined #haiku
<Gnter[m]>
x512[m]: I just have my gmail for legacy reasons, but that's routed through a proxy
HaikuUser has quit []
<x512[m]>
Gmail and Google search can be separated so it will not share cookies and login information.
<scantysnax>
i hate google, they are trying to take over everything,
<Gnter[m]>
Also really fingerprinting is a better way to identify users than IP addresses anyways
<OscarL>
scantysnax: yes.
<Gnter[m]>
And pretty sure whatever fingerprinting is happening is per IP and not per prefix unless really advanced
<x512[m]>
IPv6 prefux is perfect fingerprint.
<Begasus[m]>
here we go again ....
<Gnter[m]>
Entering schizo territory where again a VPN would be the solution
<nekobot>
[haiku/haiku] 1486d5790807 - BView: Check if the region is in-bounds before invalidating.
<Gnter[m]>
jk 192.168.0.0 for the win
<nekobot>
[haiku/haiku] 1483db6e7167 - Revert "BColumnListView: Don't invalidate if the row is below the visible rect."
<Begasus[m]>
no place like 127.0.0.1
<nephele>
that's not funny Gnter[m], that's my ip adress
<scantysnax>
so apparently, this laptop has two cores, with two threads per core, so it's like having four cpus
<nephele>
that's what intel wants you to believe :)
<x512[m]>
Also I wonder why IPv6 have /1 localhost address. For IPv4 I can assign 127.0.0.1, 127.0.0.2, 127.0.0.3 to my internal servers.
<scantysnax>
nephele: ;-)
<Gnter[m]>
<x512[m]> "Stateful connection tracking..." <- A stateful firewall is a must in any network. But still much more efficient than NAT44 (which should also be running with a stateful firewall!). Also you also don't have to use DHCP or anything like that in v6 as you have SLAAC. But of course you can if you want to, just not for android.
<nephele>
x512[m] you can assign as many localhost ipv6 adresses to your interfaces as you want...
<Gnter[m]>
Begasus[m]: No place like ::1
<x512[m]>
> you can assign as many localhost ipv6 adresses to your interfaces as you want...
<Gnter[m]>
x512[m]: It was a crazy decision allocating an entire /8 for that 😹
<x512[m]>
It need manual configuration, but 127.0.0.2 just works and perfectly secure.
<Gnter[m]>
You don't need more than 1 localhost address
<Gnter[m]>
But you can assign to as many public or ULA addresses you want :)
<nephele>
What are you even talking about? using a second localhost adress on ipv6 and ipv4 takes exactly the same ammount of "manual" configuartion
<Gnter[m]>
x512[m]: Doesn't need manual configuration on my system, it just works ... THE ENTIRE /8 😹
<x512[m]>
I need multiple local addresses to build network of servers that run on single PC and must be never accessed from outside.
<Gnter[m]>
nephele: Pretty sure only ::1 is localhost per the spec
<Gnter[m]>
But of course nothing stopping you from setting up ::2 or whatever manually
<Gnter[m]>
x512[m]: Just give them an ULA prefix
<x512[m]>
That need configuration.
<nephele>
x512[m]: those aren't "local" then, from a network topology standpoint those are severall virtual servers/machines then
<x512[m]>
127.0.0.2 just works. And perfectly portable.
<waddlesplash>
you could just copy the commit hash...
<nephele>
no, vision will open the link if you try to copy from it, at which point it will open github in webpositive, which will freeze or slow down webpositive on this machine...
<waddlesplash>
sounds like a problem in Vision then
<nephele>
sure, vision has many problems :)
<Gnter[m]>
<x512[m]> "Gmail and Google search can be..." <- Google search isn't even that good these days. But I also don't use web mail, I use actual mail clients for my mail unless it's a one off throwaway account or something like that
<nephele>
waddlesplash: do you have a screenshot of the problem?
<waddlesplash>
just open Appearance settings and look at Antialiasing tab for instance
<scantysnax>
vision has several problems. i don't think they are going to be fixed though, since nobody i know is working on them.
<waddlesplash>
or the Volume slider in Apperance preferences
<nephele>
i'm on beta5 on this machine
<waddlesplash>
happens with the default colorscheme
<Gnter[m]>
<x512[m]> "It is known that ISP give IPv6 /..." <- > It is known that ISP give IPv6 /64 subnet
<Gnter[m]>
No, only really bad ISP's give out a /64 and they should be avoided. They should according to the spec provide a /56 but /48 are also common.
<Gnter[m]>
If they give anything less than a /56 they are trash and anything below a /60 is unusable
<nephele>
what do you mean with "..."? I have my irc client open but that doesn't mean i have my dev machine running.... if you want me to investigate it myself you can open a ticket .-.
<waddlesplash>
well I posted a screenshot
<nephele>
hmm okay, from a cursory glance it looks like the high color was never set properly for the text
<nephele>
I'll make a patch later
<Gnter[m]>
Another advantage of infinity addresses where no one knows the exact prefix size is also breaking a lot of badly designed IP rate limiting tools.
<Gnter[m]>
Keep layer 3 on layer 3 and layer 7 on layer 7
<x512[m]>
Statistics of common ISP prefix sizes are known.
<Gnter[m]>
x512[m]: Probably. But who exactly is gonna develop a tool with every single ISP in the world and their prefixes?
<Gnter[m]>
Not all ISP's publish what size they delegate to customers
<Gnter[m]>
And customers can often request a larger one if necessary
<x512[m]>
Tracker and web ads developers will.
<Gnter[m]>
But yeah you could do /64 blocking, then /56 -> /48 and then ASN I guess
<nephele>
I wouldn't put it past shitty advertisers
<x512[m]>
I am quite confident that Google already do it.
<nephele>
But then, this is missing an important point, if you are on a home network, if you have a ipv4 or a prefix, this is in theory trackable more easily.... but if you have severall computers this will coaless together on the same ipv4 adress.... but you have severalll ipv6 adresses, which could make tracking easier
<nephele>
there is literally no point in reverse engineering what the prefi is
<Gnter[m]>
nephele: True
<nephele>
if you try to guess what the prefix is you just managed to group severalll unrelated computers into one, which doesn't make tracking easier but harder
<Gnter[m]>
Unless you make the assumption that one prefix is one user
<Gnter[m]>
But most homes have many different people
<nephele>
you could use the prefix to make an educated guess that the computers are in proximity or have something to do with each other, while in ipv4 you get that info for free
<Gnter[m]>
But it'll be a /64 per VLAN anyways
<Gnter[m]>
So not that hard to track that prefix
<Begasus[m]>
closing down here
<Begasus[m]>
cu peeps!
Begasus has quit [Quit: Vision[]: i've been blurred!]
<scantysnax>
seeya!
<x512[m]>
People in one home (one family etc.) that make a contract to the same ISP are likely related. So it makes sense to track them as the same entity.
<Gnter[m]>
x512[m]: Not really. Most tracking is to serve relevant ads
<nephele>
it does not x512[m]
<nephele>
the entire point of targeted advertising is *targeting* it
<Gnter[m]>
And any federal tracking would be about the same
<Gnter[m]>
* the same as on IPv4
<Gnter[m]>
As they would just contact the ISP or whatever
<Gnter[m]>
However many ISP's are forced to do session logging when doing CGNAT, while not being required or often not even allowed to log IPv6 traffic
<Gnter[m]>
Well without a court order anyways (which to be fair is piss easy for police to get)
<x512[m]>
Tracking on IPv4 will not work because addresses are dynamically allocated from large pool. Every next day you will have different IPv4 address that someone else used yesterday 100 km away of you.
<Gnter[m]>
But if your threat actor is the government a VPN is the solution anyways
<x512[m]>
Threat actor are Google, Facebook etc..
<Gnter[m]>
x512[m]: Okay? ISP's still log who has what IP when
<Gnter[m]>
And also not a given
<Gnter[m]>
Many places still have static IPv4 addresses
<x512[m]>
Google have no access to ISP logs.
<nephele>
x512[m]: the pools are quite shallow
<nephele>
if you have any cookie they can also trace that back to your activity from yesterday...
<x512[m]>
Depends on ISP and region.
<Gnter[m]>
And? Unless I login authenticating myself anyways google have no idea how large my prefix is or how often it's rotated
<Gnter[m]>
And their advertising bullshit is blocked anyways on both a client and network level
<nephele>
If you are worried about fb or google the only real solution is to either block them completely or make your UA as un-fingerprintable as possible
<x512[m]>
Cookies is a thing that I can control. I can reset cookies for web sites that I do not trust.
<nephele>
they get their vast majority of fingerptinable data from the browser, and not the networking layer
<Gnter[m]>
Only browser that has even somewhat good fingerprint protection is the Tor and Mullvad browsers
<Gnter[m]>
But everyone uses chrome anyway
<nephele>
some of that data should probably be faked
<nephele>
I don't like that this website knows my actual screen size without me asking it to be fullscreen
<Gnter[m]>
nephele: Faking it still gives you a unique fingerprint unless you fake it to the same as a lot of other people
<Gnter[m]>
Like Mullvad Browser and Tor does
<nephele>
yes, i know
<Gnter[m]>
nephele: Yeah I think letterboxing is the only solution to that
<nephele>
I wonder if you could get away with omitting the UA string completely
<nephele>
my solution is to block known trackers and malicious code as much as possible, since the other option of no js and letterboxing and no customization does not seam feesible to me
<nephele>
among other things because i *rely* on the css dark mode
<Gnter[m]>
There really is a lot that could be improved
<Gnter[m]>
I don't see why websites need to have my timezone for example
<x512[m]>
Omit UA string is a bad idea. You need to pretend you are Google Chrome.
<nephele>
No, I don't
<x512[m]>
At least for privacy.
<nephele>
Gnter[m]: huh, that one is actually not reported properly on that site, it sais ETC/Unknown
<Gnter[m]>
Wait actually just realized, why the fuck can any website see how much storage space I have?
<Gnter[m]>
Memory is wrong however
<nephele>
in case you want to download something?
<nephele>
I think i should probably hammer in a more privacy mode in webpositive that ommits a lot of this shit
<nephele>
Yes, you aren't fingerprintable against the *whole* net, but that's not the point. a tracker can figure out you are using haikuwebkit very easily, so given that it would be great to remove a lot of the other ways to fingerprint haikuwebkit browsers against each other
<Gnter[m]>
Gnter[m]: Anyways a fingerprint like this is a much better way to fingerprint people than by IP
<nephele>
Gnter[m]: my prior UI of "links <verion> on FreeBSD <version>" was 100% unique according to some UA investigating website :P
<x512[m]>
Haiku in UA contributes a lot in fingerprinting.
<nephele>
yes, let's just delete it entirely
<x512[m]>
Because there are not so many Haiku users (yet?).
<nephele>
good websites do feature negotiation
<x512[m]>
But it can be used for various statistics sites to count Haiku users.
<x512[m]>
Or tell that there are 3% of Linux users.
<nephele>
I don't care about that usecase :)
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
<nephele>
and if you omit the UA or just send "Webrowser" or something that is still sufficient for that... untill other browsers do the same
<x512[m]>
It is a bit fun that many Linux users like privacy that makes hard to correctly estimate world count of Linux users.
<jezek2>
I just grab whatever is the latest firefox's UA string on windows from time to time
<nephele>
linux has a majority market share, having most of the smartphone market....
<jezek2>
as some websites (and cloudflare too I've learned recently) checks for using versions that are recent
<jezek2>
for no reason, it works on older versions just fine
ThinkPadSL510Fox has joined #haiku
<nephele>
Well, i am amused by websites telling me to upgrade my safari...
<ThinkPadSL510Fox>
Good evening, everyone
<ThinkPadSL510Fox>
How's everyone doing this evening?
<nephele>
regardless, webkit has a mechanism to deal with bad websites, if we make the UA much less descriptive we can still have a more descriptive one for bad behaving websites
<jezek2>
cloudflare was funny, was debugging why I get nasty gatekeeping stuff when in firefox it didn't do anything and loaded the website directly :D
<jezek2>
thought about some precise tracking of headers or whatever
<jezek2>
or TLS fingerprinting :D
<jezek2>
nope, just UA header
<jezek2>
I also defeated their captcha in fully automatic way
<jezek2>
was pretty easy in fact
<jezek2>
I think they're just snake oil in their checks :)
ThinkPadSL510Fox has quit []
<nephele>
yes... probably just big bandwidth
<scantysnax>
any UI people here, i have a question about saving the state of child windows.
* nephele
disapears behind a curtain
<matthewcroughan1>
Can Haiku be built on macOS?
<nephele>
scantysnax: just ask your question, don't ask to ask... it's bad netiquette, because if i now say "yes" or something, that implicitly obligates me to answer your real question, even if i have no clue
<matthewcroughan1>
Full bootstrap, no binaries precompiled
<Habbie>
matthewcroughan1, for Linux, the answer is yes, at least :)
<nephele>
Uhhh.... full-bootstrap is very difficult.... in that case you'd have to try it yourself, in general it can be build on macos with caveats...
<waddlesplash>
it's not as difficult as it once was
<nephele>
those are documented in the build requirements page
<waddlesplash>
I did a successful bootstrap a few months ago
<scantysnax>
okay, so there are two ways i can save the state of a window. Hide() it on QuitRequested() or should I Archive the whole window?
<nephele>
waddlesplash: sure, but on macos?
<matthewcroughan1>
I ask because there's a funny problem where the Linux kernel can't be compiled on Darwin because of some non portable glibc stuff
<waddlesplash>
yeah, that will be more difficult
<matthewcroughan1>
nephele: you've told me many times the full bootstrap is "not easy", I appreciate your concern, but I'm a big boy and I know how a computer works
<waddlesplash>
building on macOS is annoying in general because we need a case-sensitive FS
<scantysnax>
oh, and return false on QuitRequested
<nephele>
scantysnax: depends what you are trying to do
<nephele>
matthewcroughan1: you asked for a definitive answer, there isn't one. Regardless of you beeing a "big boy" or not. The answer still stands, you will have to try it for yourself
<scantysnax>
just trying to save the state of child windows,.
<waddlesplash>
we don't officially support any platforms for bootstrap builds besides Linux
<nephele>
scantysnax: what state? Hide() only makes sense if your team keeps running, which isn't really the case if QuitRequested() will be called
<waddlesplash>
even Haiku doesn't work right, it needs some additional work in
<matthewcroughan1>
nephele: Sure, I fully intend to. I'm asking the question in a room full of knowledgeable people to get answers to questions, not to be told "don't do that it's scary hard"
<waddlesplash>
HaikuPorter for that to work properly
<matthewcroughan1>
if you don't know anything at all, maybe you don't need to tell me anything :D
<waddlesplash>
matthewcroughan1: well, we answered: it's difficult and nobody's done it.
<matthewcroughan1>
You do this every time I ask anything about bootstrapping, just don't say anything?
<waddlesplash>
yes, because bootstrapping is hard!
<waddlesplash>
it's hard even on Linux! it's going to be much harder elsewhere!
<scantysnax>
well, like i said i could Hide() it and return false in QuitRequested, then the window will be hidden, but it's state remains the same
<matthewcroughan1>
Of course, but nobody needs to say it's hard, you could just wait for someone to answer, I'm not expecting a quick response
<waddlesplash>
and your questions very obviously imply you have no tried it on Linux
<matthewcroughan1>
waddlesplash: I have, I've made quite a bit of progress Nixifying it actually
<waddlesplash>
so, "can I do this thing the documentation doesn't say anything about": well, maybe, but did you try the thing the documentation does support?
<matthewcroughan1>
And I'm aware of some of the challenges
<nephele>
matthewcroughan1: you can decide for your self if you want to do it. I don't keep score on who i told that it is diffcult or not. You are doing this on a platform we have never tried this on. what answers do you expecte?
<matthewcroughan1>
Anyway, I was just asking if anyone has bootstrapped Haiku from a Darwin system or not, if it has been done I would be interested
<matthewcroughan1>
Because it makes it possible to use Haiku as a tool to compile the Linux kernel on Darwin, in a fully reproducible and pure way
<nephele>
scantysnax: yes, but what state are we talking about? is this a find window you want to bring up again? or is this about an app shutting down and getting the same state when starting up again
<matthewcroughan1>
Same for any other thing that can be compiled natively on Darwin, that can then be used to compile Linux, because Darwin itself is insufficient
<nephele>
that wasn't your question matthewcroughan1, and the answer is still "no, nobody has done it"
<scantysnax>
i want to hide a window, whilst save the sate of its controls
<scantysnax>
so next time it is brought up, it will be in the same place, and sate that it was..
<nephele>
yes, scantysnax, but does the app quit?
<scantysnax>
yes
<nephele>
if the app does not quit you can just Hide() and Show() the window
<nephele>
if it does quit you need to use the archiving facility
<nephele>
or alternatively save this in a settings file and restore it based on that manually
<scantysnax>
OK, so I will look into it, including archiving
<scantysnax>
i can just archive the window, and then unarchive it
<scantysnax>
when it needs to be used
<nephele>
I'm not sure how the archiving stuff works exactly, i've not tried that yet, someone else may have a better answer for you for those :)
HaikuUser has joined #haiku
<nephele>
Yes, that can work i think, aslong as you don't have any state depending on the current time or network connectivity or something
<scantysnax>
nephele: that's fine, i appreciate your comments
HaikuUser has quit []
<nephele>
(though even then you can probably deal with this in the unarchiving method)
<scantysnax>
i will look further into it and find a solution,
<scantysnax>
so i can archive/unarchive, or just use Hide() and return false in the window's QuitRequested()
<scantysnax>
does that seem reasonable?
<nephele>
no, Hide() will not acomplish what you want
<scantysnax>
ok, then i will look into archiving
<scantysnax>
which i have no idea how to do yet, but i'm reading the Be Book now,
<scantysnax>
nephele: thank you for your suggestionsl
<scantysnax>
suggestions*
janking has joined #haiku
janking has quit [Quit: Vision[]: i've been blurred!]
janking has joined #haiku
mmu_man is now known as Guest23548
mmu_man has joined #haiku
Aedil has quit [Quit: leaving]
Guest23548 has quit [Ping timeout: 480 seconds]
<scantysnax>
hmm archiving is a litte confusing.
<nephele>
i think the basic gist is that archiving is supposed to create a representation of the entire class that can be saved to disk or memory as an obaque object
<nephele>
and when you unarchive the class has a seperate constructor to the normal one that just takes the archive as single parameter
<scantysnax>
ah
<scantysnax>
maybe not so bad then, i'll play with it later this evening,
<nephele>
scantysnax: check the activitymonitor/ActivityView.cpp::651
<nephele>
It basically just asks BView to archive itself (so that is easy), and then adds some more stuff relevant to the class
<nephele>
and this all into a BMessage
<scantysnax>
ok, cool. thanks, i'll take a look at it later
<scantysnax>
anyways, done for the day. bbl.
scantysnax has quit [Quit: Vision[]: i've been blurred!]
janking has quit [Quit: Vision[]: i've been blurred!]
janking has joined #haiku
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
jmairboeck has joined #haiku
janking has quit [Quit: Vision[]: i've been blurred!]
jmairboeck has quit [Read error: Connection reset by peer]
janking has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
<nekobot>
• kallisti5 (6c4817bd): buildmaster/builderctl: use nproc for cpu count…
nephele has quit [Quit: Vision[]: i've been blurred!]
vdamewood has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
* OscarL
is tempted to just merge the PR for libxcb/xcb_proto. It either cleanups a bunch of those repo_consistency.txt issues (<happy Begasus[m] face>) or breaks a bunch of kde/qt stuff (<murderous Begasus[m] face>).