<Nephele>
Begasus: strange times. I can go on the mastadon website thingy, and the european commision has one. See a post about right to repair and there is a comment by mmu_man
<Nephele>
small world :D
<Nephele>
Though the mastadon website is really wierd. You can't right click a post and open in new tab, only left click to expand it :(
<Nephele>
don't like it taking my navigation powers away
<Begasus>
heh
<Nephele>
Is there native software to view these?
<Begasus>
I only use Tokodon for that here
<Nephele>
I'm not really convinced yet to get an account. But viewing is kind of interesting somewhat
<Begasus>
started out as a play thing here, but using it to do some Haiku advertising now :)
<Begasus>
too bad no qt6webengine yet :)
<Begasus>
for now tokodon/kdevelop/rkward are borked for the newer versions as they require it
<Nephele>
why does it need webengine? is it just a html view?
<Nephele>
I would consider using it if there was a native api (even if it's yet-another web based one (i.e json over http or something)
* phschafft
waves to nephele.
<Nephele>
but atleast structured *data* instead of rendering instructions and executable code
<Nephele>
hi phschafft
<Begasus>
there is "toot", a cli app :)
<Nephele>
It has a REST api
<Nephele>
Begasus: have any ideas for a "good" UI for a native client ;)
<Nephele>
what would you want to have
<Nephele>
ugh, tokodon needs webengine... and also mpv? wierd.
<Begasus>
if it had functionality like tokodon I would surely check it out
<Begasus>
+ mpvqt
<Nephele>
tokodon doesn
<Nephele>
't work for me
<Nephele>
it's just blank controls with no text
<Begasus>
ah no
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
<Begasus>
in the one from gear 24 (tokodon) needs mpvqt
<Begasus>
but can't connect due to qtwebengine thingy
<Begasus>
think it has to do with authentication
<Nephele>
calligra for qt6 would be nice
<Nephele>
have calligra, vlc, and quaternion on qt5
<Begasus>
shouldn't be too hard I guess (providing it doesn't require qtwebengine) :)
<Begasus>
already have vlc split up into a devel package here
<Begasus>
nephele, do you know what the latest revbump on R1B4 was?
<Begasus>
well, it isn't that hard to grab some color-themes and use them
<Nephele>
I don't want to deal with that honestly. I'd rather it just uses what my system provides
<Begasus>
that's where I fail :)
<Nephele>
Calligra is nice for example, but quite heavy dependency wise. If it would use OS specific Icons more, and the import of images would work it could work for me
<Nephele>
i'm willing to use ported apps if they integrate nicely :)
<Begasus>
had to use a little hack to make latest kdevelop that still does qt5, to use system icons (they made changes there also)
<Begasus>
no go for calligra: devel:libQt5WebKitWidgets$secondaryArchSuffix
<Begasus>
to bring it up to qt6
<Nephele>
That native haiku text editor better be developed xD
<Nephele>
why are ported text editing suites so painfull :(
<Begasus>
cudatext isn't that painfull?
<Nephele>
not sure what that is
<Begasus>
tss :)
<Nephele>
doesn't look like a text editing for me? I mean something like LibreOffice, staroffice, and such. That can do inline images and text reflow on virtual A4 pages and such
<Nephele>
things employers expect you to send by email
DKnoto has joined #haiku
<Begasus>
remember something about employers ... been a while though :D
<Nephele>
hehe :D
<Nephele>
lucky you, who does not have to deal with "make pdf"
<Begasus>
you can do that with TeXstudio too
* Begasus
ducks
<Nephele>
might be a better idea than dealing with libreoffice
<Begasus>
if it wasn't for the texlive dependency ;)
<Nephele>
can it do inline images?
<Begasus>
no idea
<Nephele>
well. You just can't win huh :(
<Begasus>
heh
<Begasus>
Kate?
<Begasus>
hmm ... only see export to html there
<Begasus>
looks weird, I think that should be supported ...
<Begasus>
checking build for okular-24.05.0
<Begasus>
that uses qt6/kf6
<FreeFull>
You can always print to pdf, right?
novaphoenix has quit [Quit: gone fishing]
<Begasus>
sec ...
<Begasus>
okular has export to pdf
novaphoenix has joined #haiku
<Nephele>
Yes indeed, My problem is usually before that. Libreoffice would freeze and use all my 4GB of ram when adding an image, calligra just sais it can't import them. And then, despite both using the same .odt file format can't open it if the other saved it
<jmairboeck>
texinfo uses perl, so it could be related
<waddlesplash>
oh
<Begasus>
anyway, will read up later (or in the morning)
<Begasus>
cu peeps!
<waddlesplash>
I think the regression there was only on nightlies due to new ICU
Begasus has quit [Quit: Vision[]: i've been blurred!]
<waddlesplash>
but this should never happen inside a haikuporter chroot
<waddlesplash>
because we set LC_ALL=C
<jmairboeck>
I haven't looked at it myself yet
<waddlesplash>
dovsienko: someone else reported that same bind-hash KDL
<waddlesplash>
I looked into the code around here a bit more and I can't see how we're freeing sockets too early
<waddlesplash>
but it seems notable the socket itself is in syn-received state. how did we get a socket in that state on shudown?
<waddlesplash>
it looks like it's a socket that got stuck failing to initialize, but that doesn't really answer the questio
<waddlesplash>
question
<dovsienko>
it is difficult to reason about the root cause from my point of view, I just don't know the internals
<waddlesplash>
well, I mean, how does sshd have a socket that failed to initialize here, is what I mean
<dovsienko>
but it certainly looks like a bug in the network stack
<waddlesplash>
I would think that failure would be obvious on the host side
<waddlesplash>
you would try to connect to the VM via SSH and it would just hang
<waddlesplash>
so that's my question, where did this socket come from?
<dovsienko>
the "state" could be just some randomly coincident binary value
<waddlesplash>
highly unlikely
<waddlesplash>
all the other data in that dump looks like a real TCP endpoint state
<waddlesplash>
the flags, lock, etc. all look as expected
<waddlesplash>
the send and receive queues have the expected values
<dovsienko>
does it have the addresses and ports too?
<waddlesplash>
the addresses and ports are in the net_socket
<waddlesplash>
that's the part that got freed
<waddlesplash>
a net_socket object is general across all types of sockets and is the part that's reference counted
<waddlesplash>
the TCPEndpoint object is owned by the net_socket object and should be freed when the net_socket is
<waddlesplash>
but somehow we have gotten into a state where apparently the TCPEndpoint is valid but the net_socket isn't...
<waddlesplash>
hm, maybe the free happened between the first Put and the time we got to the ~TCPEndpoint
_-Caleb-_ has left #haiku [#haiku]
<dovsienko>
given the key event do not occur way too often, it would not generate too many log messages if the stack logged the events
<waddlesplash>
well, knowing what the key event is would be nice
<waddlesplash>
are you specifying any interesting args to SSH on the host?
<dovsienko>
one would be freeing the socket
<dovsienko>
let me check
<waddlesplash>
I just hacked up a build where socket memory is never freed (should catch negative reference counts better) and tried SSH'ing into it but it still didn't trigger this
<dovsienko>
waddlesplash: "Subsystem sftp /boot/system/lib/openssh/sftp-server" and "PermitRootLogin yes", the rest is the defaults
_-Caleb-_ has joined #haiku
<waddlesplash>
ok
<waddlesplash>
and when connecting, any interesting args?
<dovsienko>
it happens only once in a while, could be a race condition
<waddlesplash>
it seems strange that we wind up with a "dead" socket somehow but you don't experience a failure on the host, say
<dovsienko>
at the client: "ForwardAgent yes" and "ServerAliveInterval 120"
<dovsienko>
so when a TCP socket or a net_socket come into existence or get deallocated, that could be a debug message, and then another when one is associated/deassociated with the other, then as soon as there is a timeline of what happened before the crash, you would have a sense of where to look next
<waddlesplash>
sure
<waddlesplash>
but we have a bunch of assertions around this already
<waddlesplash>
and the fact that none fired is very strange
<dovsienko>
could it be that the cause is somewhere else? some other code corrupting random memory with a likelihood of this being the network stack sockets data?
<waddlesplash>
yes, it's possible
<waddlesplash>
but it seems pretty unlikely...
<waddlesplash>
I tried running with a "guarded heap" build, this requires basically a full page for every allocation and sticks guard pages around it, this still didn't crash
<waddlesplash>
but perhaps there is some difference in setup between our systems
<waddlesplash>
if you'd like I can send you that build
<dovsienko>
alternatively, could be a race condition. this VM has 4 cores, and if the condition is that two concurrent processes must be on the same core and in a specific order, that would give a chance of 1/32 (if I am corect), which roughly corresponds with how often this reproduces
<waddlesplash>
I wonder what the race could be, though
<waddlesplash>
the only thing I can think of would be a TCP timer
<waddlesplash>
but the only TCP timer that interferes with reference counts is the TIME_WAIT one
<waddlesplash>
and this socket is not in TIME_WAIT state
<dovsienko>
it could be useful to log TCP state transitions too
<waddlesplash>
hm
<dovsienko>
I have rebooted the VM about 10 times over SSH and ACPI and, of course, it did not reproduce. must be a race condition
<waddlesplash>
or if it's time-wait related then we will only get it after 2 minutes time-wait timeout
<waddlesplash>
let's try reducing the timeout
<dovsienko>
if you make it reproduce reliably, it will be a lot of feedback
<waddlesplash>
yes
<Nephele>
waddlesplash: with the "automatic colors" is it intentional that navigation base may generate as red?
<dovsienko>
so what would be the next step if you see it triggering massively?
<waddlesplash>
nephele: it is if you set "Status bar" to a red color
<waddlesplash>
dovsienko: well, at that point it should be easy to debug and fix
<Nephele>
waddlesplash: it's not. it is pink
<waddlesplash>
ah
<waddlesplash>
yes, that may be expected then
<waddlesplash>
the HSL processing for that color was a bit odd
<Nephele>
if i move the red slider it switches between blue and red
<Nephele>
both kind of suck for navigation colors in dark mode
<waddlesplash>
yeah I think it may only choose a primary color
<waddlesplash>
we should likely improve this logic
<Nephele>
well, since i'm encountering this i'll try to think of a solution to it
juanjo has joined #haiku
axeld has quit [Read error: Connection reset by peer]
higen has quit [Ping timeout: 480 seconds]
nosycat has quit [Quit: Leaving]
mmu_man has quit [Ping timeout: 480 seconds]
zard has joined #haiku
<waddlesplash>
dovsienko: I tried adding a hack to get sockets to stick in syn-recv state (just drop all packets after), wound up with a nice backlog of sockets in the queue but still no KDL
tqh has joined #haiku
Nephele has quit [Quit: Vision[]: i've been blurred!]
erysdren has joined #haiku
mmu_man has joined #haiku
<dovsienko>
waddlesplash: it does not reproduce reliably, so perhaps the best thing to do is to add means to record the sequence of events that leads to the KDL and to wait until it randomly reproduces again
<dovsienko>
is there a way to have conditional debugging that I could keep switched on?
<waddlesplash>
not exactly
<waddlesplash>
yes, maybe so. I think I have some more asserts to add that may catch it at least a little earlier
<dovsienko>
if it is not the network stack code that corrupts that memory, would it make sense to have a periodic integrity check that would verify that the sockets are still in good shape?
<waddlesplash>
I don't think that's the problem
<waddlesplash>
the dump of the socket memory indicates that there is a bunch of 0xcccccc in it
<waddlesplash>
that's what the allocator memsets on allocate on KDEBUG kernels
<waddlesplash>
so it looks like the socket got freed and then something else reused the memory
<waddlesplash>
not very interesting, really. the question is how the socket got freed without anybody noticing for apparently a long time
<dovsienko>
at least this would eliminate other possibilities and prove with some confidence that the invalid data is a product of the network stack
<waddlesplash>
and more importantly, how the network stack still does have a reference to the net_protocol depsite the socket being gone
<waddlesplash>
that's the more interesting part, because all these KDLs are inside netstack code -> TCP
<waddlesplash>
the only thought I have, is that somehow the net_socket->protocol != protocol->socket
<waddlesplash>
I have no idea how that could be the case but it would fit the facts
<dovsienko>
how many lines are there that free a socket?
<waddlesplash>
1
<waddlesplash>
socket_release -> ReleaseReference -> delete'd when refcount drops to 0
<dovsienko>
well, then to me it would make sense to log everything
<waddlesplash>
we unfortunately don't have a BSD-style sysctl interface for turning logging on and off
<waddlesplash>
I can make you a build with more added though
<dovsienko>
if it is the network stack, there must be some odd branch of code somewhere that ends up there one time too many
<waddlesplash>
yes, but I just audited all the acquire_socket/release_socket calls between TCP and the stack itself
<dovsienko>
my VM can run a debug build without a problem
<waddlesplash>
they all look ok. some around time_wait seems suspect
<waddlesplash>
but those in particular never activate for a SYN-RECV socket state
<waddlesplash>
I added some code to assert around them just now anyway and didn't get problems on testing ssh
<dovsienko>
FWIW, it would be a good thing to address after beta5, even if not completely fix it
<waddlesplash>
well, yes
<waddlesplash>
it would be good to address before beta5 honestly
<waddlesplash>
oh, you mean sysctls?
<waddlesplash>
yeah that would be a bigger project
<x512[m]>
Recent TCP code changes cause KDL?
<waddlesplash>
x512[m]: no
<dovsienko>
I mean the KDL bug. it does not reproduce often enough to be a blocker, as far as I see it
<waddlesplash>
this has been around for 3 years at least
<dovsienko>
has been around for a long time, and most users are not even aware this exists
<dovsienko>
must be something particular that makes it more likely
kinkinkijkin has joined #haiku
<dovsienko>
number of CPU cores, RAM size, type of network hardware...
<dovsienko>
back in 15 minutes
<erysdren>
attention haiku gamers: i am bumping QuakeSpasm to the newest revision 0.96.3
<erysdren>
:P
<erysdren>
im also thinking about booting into haiku and mucking around with some software ports
<erysdren>
mostly FPS games of some stripe.
dalme_ has joined #haiku
gouchi has joined #haiku
<erysdren>
quick q for anyone here: what's the best simple music player for haiku? something that just has a playlist view and volume control, basically. on my main desktops i use VLC, but on Haiku i don't like how VLC is stuck in multi-window mode.
zard has quit [Quit: leaving]
<PulkoMandy>
Anything wrong with the built in media player?
<PulkoMandy>
The alternatives are APlayer, and for 32 bit only, SoundPlay and clamp
<erysdren>
hmm, gotta be honest i don't know if i tried the default media player for an extended period of time
<erysdren>
i'll give it a shot next time i boot haiku
_-Caleb-_ has left #haiku [#haiku]
Anarchos has quit [Quit: Vision[]: i've been blurred!]
<botifico>
[haikuports/haikuports] erysdren 88bb8df - Bump QuakeSpasm to 0.96.3 (#10782)
<phschafft>
rocket fuel it is.
<dovsienko>
waddlesplash: possibly the use of SSH server could be the factor
<waddlesplash>
dovsienko: yes, I was testing with SSH server
<waddlesplash>
connecting to Haiku via SSH from a Linux VM
<dovsienko>
that is what I do, and even with that it reproduces only once in a while
<phschafft>
41.44km today.
<dovsienko>
or maybe it is specific to listening sockets rather than SSH
dalme_ has quit [Read error: Connection reset by peer]
dalme_ has joined #haiku
HaikuUser has joined #haiku
juanjo has left #haiku [Disconnected: Hibernating too long]
dalme_ has quit [Ping timeout: 480 seconds]
jmairboeck has quit [Quit: Konversation terminated!]
tqh has quit [Quit: Leaving]
Anarchos has joined #haiku
AlienSoldier has joined #haiku
<Anarchos>
i got this error : SSL certificate problem: certificate is not yet valid
<Anarchos>
with curl in Terminal
SLema has quit [Quit: Vision[]: i've been blurred!]
SLema has joined #haiku
<waddlesplash>
check your system clock
tuaris has quit [Read error: Connection reset by peer]
<dovsienko>
Anarchos: if you run curl with --verbose, you can compare the certificate dates with the host date
tuaris has joined #haiku
<dovsienko>
so it may be not an error, just a statement of the fact
oevl_ has joined #haiku
<AlienSoldier>
waddlesplash i noticied (while having the about crashing) that sysinfo do display the date of the kernel but not the hrev. I think it may be usefull to have there.
nebulabc_ has joined #haiku
DHowett_ has joined #haiku
oevl has quit [Ping timeout: 480 seconds]
nebulabc has quit [Read error: Connection reset by peer]
DHowett has quit [Ping timeout: 480 seconds]
HaikuUser2 has joined #haiku
<HaikuUser2>
hello, what's the build type for haiku, i want to use "./configure --build=" command
<HaikuUser2>
??????
<dovsienko>
as far as I understand, that defaults to what config.guess returns, and on 64-bit Haiku config.guess returns "x86_64-unknown-haiku"
erysdren has quit [Quit: Konversation terminated!]
<HaikuUser2>
oh, tanks dovsienko
<HaikuUser2>
oh, thanks dovsienko
HaikuUser2 has quit [Quit: Vision[]: i've been blurred!]
<Anarchos>
waddlesplash thanks, i didn't think of date . But i suffer a power loss at home, that explains