<nephele>
in general this means however, that our error handeling is lacking and not specific enough, and maybe you should complain about it :P
<divad27182>
network seems to work
<nephele>
hmm. I think you can run it with strace to see just which system call is the one that would block
cassisian has joined #haiku
<divad27182>
trying...
<phschafft>
find .../docs/ -type f -print0 | xargs -0 sed 's/\<successful return\>/vitory/' -i -- # make reading docs more fun
<nephele>
phschafft: maybe functions should pass "Congratulations! You passed the test!" As sucess instead of a 0
<phschafft>
'0 but true'
<phschafft>
and next we learn about different types of NaN.
cassisian has quit [Ping timeout: 480 seconds]
<nephele>
"The returned value is Not A NaN, a NaNaN"
<divad27182>
My errors are:
<divad27182>
1* Operation not allowed
<divad27182>
8* No such file or directory
<divad27182>
7* Is a directory
<divad27182>
1* Read-only file system
<divad27182>
none of them seem to explain "would block"
<phschafft>
nephele: haha.
<phschafft>
nephele: I mean it could still be something else fun, like minus uncountable infinite ;)
<nephele>
divad27182: still, would block means that B_WOULD_BLOCK was returned, and this normally means that you are supposed to try and operation again later, as the function call otherwise would have blocked... similar to posix EAGAIN (which it is aliased to in Haiku)
<divad27182>
I've always wanted to explore some of the larger infinities...
<nephele>
the question would be in your case where it does thig
<nephele>
but i would suggest just filling a ticket for now, and maybe adding the trace (though not sure if it will help or not, but if you already have it :)
<phschafft>
divad27182++
<divad27182>
right, B_WOULD_BLOCK would be 0x8000000b which does not appear in my strace output.
<divad27182>
part of it: need a better strace
<nephele>
It can also be returned by library routines instead of system calls
<nephele>
that would however not show up in strace
<nephele>
ah, webpositive is wasting cpu cycles with a running curl thread again
<nephele>
that to me seems more annoying than a memory leak. I am curious though, many linuxes do not show CPU utilization in their normal GUI at all. Maybe CPU utilization is aproached differently because of that
<divad27182>
is there an ltrace for haiku?
<nephele>
In Haiku with it always visible it becomes clear very quickly if something is running
<divad27182>
bingo... found an 800000b....
<nephele>
doesn't seem so, but the tool sounds interesting
<divad27182>
OK, I reran with -s -t -T added, and it turns out it is a connect() call in another thread/process/whatever...
<divad27182>
ltrace is a library call tracing tool available on linux.
<divad27182>
Apparently, what broke, I turned on IPv6, with no internet access....
<nephele>
have you managed to get an ipv6 connection before on Haiku?
<nephele>
I tried configuring my interface manually but did not get a connection that way
frkazoid333 has quit [Ping timeout: 480 seconds]
<divad27182>
I've had it pinging sucessfully...
<divad27182>
Well, that is what went wrong with pkgman
<divad27182>
particularly since I was using an FD00::/8 address
erysdren has joined #haiku
xet7 has quit [Remote host closed the connection]
<divad27182>
well, I have just proven I can ssh into and out of haiku using IPv6
<divad27182>
interestingly, the IPv6 entries do not show up in netstat
<divad27182>
no wait... they do, but without addresses
cassisian has joined #haiku
frkazoid333 has joined #haiku
cassisian has quit [Ping timeout: 480 seconds]
<divad27182>
Is there a way to force Haiku to redo DHCP?
<nephele>
quiting (or killing) net_server might acomplish that, but i'm not sure
<AlienSoldier>
divad27182 in network pref you can disable DHCP and re-enable it. Don't know how to do it by code or script that said. I need to do it at my summer house ISP at around 3-4 hours of use.
<divad27182>
I found you need to set it to "Disabled" and the back to "DHCP". Setting it to static doesn't do it.
<AlienSoldier>
yes, back to DHCP
<nephele>
annoyingly the ipv4 and ipv6 settings are somehow linked and do stupid stuff if you try to use both at the same time
ADS_Sr has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
sure_ has joined #haiku
sure has quit [Ping timeout: 480 seconds]
<divad27182>
well, what I'm currently finding is that IPv6 wants to be "Automatic", no matter what I do.
<divad27182>
and it no longer works at all
<nephele>
it's because you have ipv4 in automatic :/
<nephele>
try using ifconfig instead
<divad27182>
a few minutes back, I had an FDxx::/8 address on it.
<divad27182>
issues I have:
<divad27182>
1) there is this renegotiate button... that is disabled
<divad27182>
2) it is normal, perhaps even required, that a host have multiple IPv6 addresses on one interface
<nephele>
this support is simply lackluster currently divad27182
<divad27182>
(There should be an FE80::/16 address)
<divad27182>
yeah, but this is the CURRENT internet.
<nephele>
atleast our loop interface has ::1
cassisian has joined #haiku
<nephele>
I know :/
<nephele>
for one this preference pane for ipv6 needs to work better. There is no meaningfull netmask either
<divad27182>
and it routes to it twice!
<nephele>
So maybe this should simply be a listview with the first column the ipv6 adress, and right of it flags for it (like prefixlen, is auto or not, is scoped)
<divad27182>
well, IPv6 uses a strict bit count mask.
<divad27182>
unlike IPv4, where you could write your netmask as 255.255.15.240 if you wanted to
<nephele>
I've only seen this expressed as a plain integer named prefixlength in other OS, i think we should just write it like that aswell
<divad27182>
Yes, In IPv6, one normally just uses the prefixlength
tetrislife has joined #haiku
<divad27182>
like I said, it isn't valid in IPv6 to not be a prefixlength
<divad27182>
and many implementations will pretty much lock you in at /64
<divad27182>
except for routing purposes...
<nephele>
locking you in seems wrong. But still, we should do it like this too...
<divad27182>
well, it isn't unreasonable for the netmask of your interface.
<divad27182>
particularly if you are going to use SLAAC
<nephele>
Is the gateway here relevant aswell?
<divad27182>
though I will admit, I've seen /64 on point-to-point interfaces, and that does seem wrong. They could have used /126.
<divad27182>
or even /127....
<divad27182>
though that would have violated some of the old berkeley socket rules.
<divad27182>
(that people think are inviolate!)
<nephele>
I think for the dns server and ntp server adresses it would be cool if it said "configured by your network" or something if we obtain those over dhcp :)
<nephele>
How complex would SLAAC be to implement?
cassisian has quit [Ping timeout: 480 seconds]
<divad27182>
Well... SLAAC is done by receiving certain broadcast packets.
<divad27182>
I think you can ask for them as well.
<divad27182>
They are in the ICMPv6 space
<nephele>
well, we should have atleast some icmpv6 support in place considering our ping tool can do ipv6
<divad27182>
yes, it can...
<divad27182>
I noticed that the machine doesn't necessarily receive IPv6 until after it sends at least once.
nephele_ has joined #haiku
nephele is now known as Guest21187
nephele_ is now known as nephele
<nephele>
divad27182: could be a bug
<divad27182>
there are two other bits of ICMPv6 that one must use...
<divad27182>
one is router advertisements. These announce the routers to the other subnets in the area
<divad27182>
one is neighbor solicitation. This is the in-IPv6 equivalent of ARP in IPv4.
<divad27182>
The later is probably there already... as IPv6 can work.
<nephele>
reading the rfc SLAAC seems quite straightforward
<divad27182>
I believe so.
Guest21187 has quit [Ping timeout: 480 seconds]
<divad27182>
I think one needs a router advertisement daemon (radvd) on ones network.
* phschafft
likes to point out that ARP is in fact independent of IPv4 ;)
<nephele>
what would that daemon do?
<divad27182>
phschafft: right
<divad27182>
that advertises the routes and prefixes.
<divad27182>
it sends the stuff that SLAAC receives.
<nephele>
should usually be done by the router in a home network I suppose?
<nephele>
i certainly have ipv6 in my local LAN, because i fell for an OpenBSD april fools joke
<divad27182>
probably, if you have that kind of simple network (I don't), and your router deals with it.
<nephele>
for what purpose would you make it more complex?
<nephele>
I mean even if you have dhcp6 on another server, why would you have the router advertisement on another one than the actual router?
sure_ has quit [Ping timeout: 480 seconds]
<divad27182>
Well, I have one main network in the house, a second for WIFI, one for each XEN DOM0, and several VPN nets.
<divad27182>
and both XEN DOM0s run radvd, as well as the internet/wifi/VPN router
<divad27182>
but I've never gotten DHCPv6 to work....
<nephele>
Okay. if you have that in addition that can make sense, though that to me is another context
<divad27182>
and of course, on top of all that, I have both a 2000::/3 address (the routable IPv6 internet), and an FD00:;/8 address (the IPv6 equivalent of private use networks).
<divad27182>
and some hosts get easy addresses (FDxx:xxxx:xxxx:1::1)
<nephele>
I have not quite gotten around to understand how to do the private ipv6 LAN prefix stuff
<divad27182>
in addition to SLAAC style addresses (FDxx:xxxx:xxxx:1:yyyy:yyFF:FEyy:yyyy)
<nephele>
i know somehow someons should generate a prefix, and then each computer gets one adress with that prefix that is only valid in the LAN
<divad27182>
pretty much...
<nephele>
which would be great for wireguard on a mobile network, because iOS does not like routing ipv4 in that case (maybe because the mobile network has no ipv4 adress?)
<divad27182>
there is a recommendation to create your own 40 bits through a hash of specific things, including time
<nephele>
divad27182: yes but, who distributes that prefix? the dhcpv6 server? the router advertiemensts? :)
<nephele>
that's what i've not yet figured out
<divad27182>
the router advertisements, at least in my case.
<divad27182>
if I went with a DHCP6 server, it could do so as well
<nephele>
in our lan a lot of computers have stable adresses within the lan (for ipv4) but i would like that for ipv6 aswell
sure has joined #haiku
<divad27182>
you can...
<nephele>
hmm, but the RA are done by the router, and that is proprietary firmware with a fancy webinterface, not sure i can set that up there
<divad27182>
with SLAAC, the bottom 64 bits are set based on the MAC address, with the local bit flipped, and 16 bits added in the middle.
<nephele>
though openBSD got 5 ipv6 adresses on it's interface directly, two 2010: adresses, but two also f80 or something, so maybe it already does this
<divad27182>
FE80 --- that is the link local address
<divad27182>
basically, the mac address of the machine turned into an address
<divad27182>
used to certain operations
<divad27182>
I think when you neighbor solicit, you get one of those back
<divad27182>
and with those, you don't need to neighbot solicit
<nephele>
i have one fe80, and two fdf8
<divad27182>
the 2010: address is part of the global internet.
<divad27182>
ah. fdf8 is your local network.
<divad27182>
your private use area.
<nephele>
so, that's already the local prefix stuff?
<divad27182>
its a /7 block for which the second have is used for that.
<divad27182>
yes.
<nephele>
why does it assign two adresses, one marked temporary?
<divad27182>
ah yes....
<divad27182>
IPv6 addresses assigned with SLAAC include your MAC address. This allows your machine to be tracked as it moves around the internet.
<nephele>
I don't get my dns through wireguard... but thinking more about this, that doesn't make sense... i should be able to get it, it's not multicast traffic....
<divad27182>
Therefore, they created a system of assigning 8 random bytes (or mostly random?) of address for the bottom half.
<divad27182>
and renewing those every so often.
<nephele>
i don't quite understand why you don't randomize the entire part after the prefix, check if there are neighbours, and if there are try again...
<divad27182>
the stack must keep any given address up as long as it is in use, but typically it makes new ones every few hours, and with each new subnet it is part of.
<nephele>
reading the rfc precisely it basically tells you to fail after one attempt and have the system admin correct this
<divad27182>
this feature is optional...
<nephele>
but somehow i don't think the average haiku user should be asked to resolve this....
<divad27182>
but appropriate for laptops and cell phones wandering wifi space.
<nephele>
so, the local prefix, can this change? or can i trust my router to keep this?
<divad27182>
This is something that should be turned on by default, and that the user should have the option of turning off.
<nephele>
if it's stable can i assign them in dhcpv6 to give specific computers specific adresses aswell? (basically to match the ipv4 stuff, easy to remember :D)
<divad27182>
the fdxx prefix should not in general change. If your router assiged it, you might look at seeing if it has an option to SET it so when you replace the router, you can set it.
<divad27182>
of course, you could have any machine on the net advertise this prefix.
<divad27182>
not just the router.
<nephele>
well, if the router doesn't stop, wouldn't that interfere then?
<divad27182>
well, machines could have multiple private use prefixes...
<nephele>
ah. good point
<divad27182>
just as they have that and a 2010:... prefix
<divad27182>
(they don't call it private use, by the way... I don't remember the proper name.)
<nephele>
hmmm, now i would have to figure out what the ipv6 adress of the other computer is
<divad27182>
and these address are allowed on the internet... but don't expect them to route back successfully.
<divad27182>
that what DNS, and various forms of dynamic DNS are for.
<divad27182>
your router may be providing that too....
<nephele>
yeah. But, this computer is 2 meters away from me... :) I would have to check there
<nephele>
in this network DHCP and DNS is done by another computer, not the router
<nephele>
a couple of dell optiplex machines i got for free
<divad27182>
well, if a computer registers with a dynamic DNS server, then others will find it...
<divad27182>
free computers. :-)
<divad27182>
alternatively, you can run a manually configured DNS. I do.
<nephele>
I got like 20 free computers from one company... completely working, but headed to a landfill. Presumeably because windows on them sucked
<nephele>
Yes, our DNS is manual
<divad27182>
and you can put IPv6 address in it too.
<divad27182>
record type AAAA
tetrislife has left #haiku [Error from remote client]
<divad27182>
(There was an A6 type record, but it's effective use required support from ISPs. You know how that worked out.)
<nephele>
and i have access, i'm told. And i do have acess.... if i knew where my ssh key was, so anyway, i will make a new ssh key and hand that over xD
<divad27182>
:-)
<nephele>
divad27182: yes, but before i did not know how to assign the local prefix, since this is already done, now i only need to figure out if computers have gotten the local ipv6 adresses predictably... and if not maybe make the dhcp server hand out those adresses instead
<divad27182>
I use SSH for talking between all my machines. I know where my keys are, and the passwords, except for some old keys from the company I left 12 years ago.
<nephele>
we have a mapping of mac to ipv4 adresses in the config, but not for ipv6
<divad27182>
well, SLAAC is very predictable.
<nephele>
I know some people like to use one ssh key for everything
<nephele>
but i always make one for each purpose
<nephele>
hence i sometimes misplace them, or have them only on one computer
<divad27182>
actually, if you have a MAC/IPv4/Name table, you can build the IPv6 addresses from that
<nephele>
well, i can build a *new* ipv6 adresses from that?
<divad27182>
yes.
<nephele>
not neccesarily the same one slaac would have assigned
<divad27182>
yes!
<nephele>
but i am guessing that it's not really reelvant if there is another adress on the interface :)
<divad27182>
if your MAC is aa:bb:cc:dd:ee:ff
<divad27182>
then your latter part of your SLAAC address is a8bb:ccFF:FEdd:eeff
<divad27182>
well, mac address, IPv4 addresses, and IPv6 addresses are all interface specific
<divad27182>
there's one other weird class of addresses... IPv6 around IPv4 addresses.
<divad27182>
these let you use the IPv6 stack of a dual-stack machine to talk to an IPv4 machine
<nephele>
I'm actually curious about that case. Perhaps tunneling ipv4 traffic through ipv6 would be worth it in a LAN to not have to deal with NAT ;)
<divad27182>
well, that case is more about running a server using only an AF_INET6 port.
<divad27182>
outside your machine, it is IPv4...
<divad27182>
but the software can just listen on IPv6 and get both.
<nephele>
My new mail server is ipv6 only, or should be
<divad27182>
at least, if the OS supports it (Linux does)
<nephele>
hence my interest in getting this damn ipv6 support working in Haiku
<divad27182>
Well, I'm not sure I'm interested in learning Haiku internals, but I can talk IPv6, and maybe review and test.
<divad27182>
and I would like to see IPv6 support (and NFS support) in Haiku.
<nephele>
that's cool. I am very interested in Haiku internals, and about people telling me how to do ipv6 better ;)
<nephele>
well, we have nfs support. but no discovery
<divad27182>
not fully...
<divad27182>
when I do NFS writes as non-user-0, Haiku crashes.
<divad27182>
When I read from NFS, change the file remotely, and read again, Haiku doesn't get the changes
<divad27182>
(but it knows it is changed!)
<divad27182>
These two bugs are pending.
<nephele>
one other thing might be teaching haiku to use some higher source ports for it
<divad27182>
haiku trying to write through a read-only NFS mount is fixed
<divad27182>
actually, NFS is supposed to use <1024 source ports....
<divad27182>
not that that really gets you much security...
<nephele>
wasn't it the other way around? I saw some nfs server had to supply an option called "insecure" if you tried to use those source ports
<divad27182>
but NFS clients are supposed to be "secure"
cassisian has joined #haiku
<divad27182>
no, insecure allows ports >=1024
<divad27182>
on Linux, at least
<nephele>
ah. I see
<divad27182>
quoting:
<divad27182>
secure This option requires that requests originate on an Internet port less than IPPORT_RESERVED (1024). This
<divad27182>
option is on by default. To turn it off, specify insecure.
<nephele>
well, that one i had trouble with in the past
<divad27182>
do you know anything about Haiku's native file sharing?
<divad27182>
and if anything else implements it?
<nephele>
are you talking about netfs?
<divad27182>
I think that was the name
<nephele>
I only know off Haiku that implements it
<nephele>
i've not tried it myself unfortunately, so don't know how well it performs
<divad27182>
I've only got the one Haiku box.... so same here
<nephele>
though, if i had to quess, i'd assume this somehow implements querries...
<nephele>
oh, i have severalll.... just never had the idea to :)
<nephele>
among other things, this has a text config setting... and no gui settings preflet, makes stuff go very unoticed in a "gui first" OS...
<divad27182>
well, personally, I think Haiku is a bit too focused on the GUI....
<divad27182>
when you crash a command line program (run over SSH) and a GUI window pops up, it is a pain.
cassisian has quit [Ping timeout: 480 seconds]
<nephele>
tell it not to do that if you don't want that....
<nephele>
the normal use case is definetely not people sshing into your machine and crashing your software ;)
<nephele>
but the debug server has settings where you can specify that some, or all, programms if they crash are treated as automatic report saved, terminated, or gui debugged
<divad27182>
interesting... where to I get these preferences?
<PulkoMandy>
Begasus[m]: Probably fine for Haiku dependencies, if they are broken by the changes we can patch them
cassisian has quit [Remote host closed the connection]
cassisian has joined #haiku
cassisian has quit [Remote host closed the connection]
cassisian has joined #haiku
r6fej has quit [Quit: Leaving]
mmu_man has quit [Ping timeout: 480 seconds]
r6fej has joined #haiku
cassisian has quit [Ping timeout: 480 seconds]
<Begasus[m]>
finished 4.0.3 earlier, can run a build on Haiku to check PulkoMandy
CRThaze_ has quit [Ping timeout: 480 seconds]
vdamewood has joined #haiku
arjen has joined #haiku
cassisian has joined #haiku
mmu_man has joined #haiku
cassisian has quit [Ping timeout: 480 seconds]
CRThaze has joined #haiku
cassisian has joined #haiku
Nasina has quit [Read error: Connection reset by peer]
Nasina has joined #haiku
Nasina has quit [Read error: Connection reset by peer]
Nasina has joined #haiku
Nasina has quit [Read error: Connection reset by peer]
Nasina has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
Nasina has quit [Ping timeout: 480 seconds]
cassisian has quit [Remote host closed the connection]
divad27182 has quit []
divad27182 has joined #haiku
cassisian has joined #haiku
cassisian has quit [Ping timeout: 480 seconds]
angrystar17043 has quit [Ping timeout: 480 seconds]
angrystar17043 has joined #haiku
cassisian has joined #haiku
vdamewood has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
cassisian has quit [Ping timeout: 480 seconds]
<divad27182>
Is there a "utmp" file somewhere on Haiku?
<divad27182>
it has utmpx.h
cassisian has joined #haiku
<divad27182>
what's the repository for the *_source packages in HaikuPorts ?
<kallisti5[m]>
<Begasus[m]> "kallisti5 (@kallisti5:matrix.org..." <- Nope. Still offline. tbh, if they don't get repowered soon, I may start working on running my own builders. I have a few Intel-nuc machines laying around... maybe I can reclaim them for builders. (though, i'm running out of space in the short term lol)
<Begasus[m]>
kallisti5 (@kallisti5:matrix.org) nothing major in the pipeline for now*, doing some checks on cmake 3.0.4 (need to run the tests for that), checked builds for x86_64 and riscv64 earlier, that went fine (also got the newer zstd installed, no issues so far
<KENZ>
I built haiku_loader.efi with this patch, then I placed it in /efiboot/EFI/BOOT/.
<Begasus[m]>
placed the "patch" in .... ? ;)
<KENZ>
I tried to boot with that efi image from rEFInd, but it just shows empty white screen forever. How can I use this customized efi image? Or something wrong with my patch?
<KENZ>
placed custom "haiku_loader.efi" I meant.
<waddlesplash>
KENZ: did you test your custom loader in a VM? does it work there?
mmu_man has joined #haiku
<KENZ>
waddlesplash: No. Not yet. I'll try that.
Nasina has quit [Ping timeout: 480 seconds]
KENZ has quit [Quit: Vision[]: i've been blurred!]
Skipp_OSX has joined #haiku
dodo75 has joined #haiku
dodo75 has quit []
dodo75 has joined #haiku
Skipp_OSX has quit [Ping timeout: 480 seconds]
<divad27182>
Does anybody know what POSIX specifies for attempting to exec() (or execve()) a directory? Linux gives EPERM, while Haiku gives EISDIR.
<divad27182>
Linux docs say you get EISDIR if the interpreter is a directory, but that seems to give EPERM as well...
<divad27182>
fexecve for instance can do the things it says it cant.
<Ellenor>
>unironically linking X, without copy-pasting the enclosed text
<Ellenor>
what is this, 2021?
* waddlesplash
shrugs
<waddlesplash>
I searched for the joke, and it was the first result
<Ellenor>
meh
<waddlesplash>
the post is from 2018 anyway
<Ellenor>
I hven't had restorative sleep in a full decade, and I'm only 25
<waddlesplash>
??
<Ellenor>
the whole world is like fucking nonsense to me now
<Ellenor>
nothing makes sense
<Ellenor>
you continue your talk, I'll just watch
<divad27182>
Well... My reason for asking this one: I'm working on porting tcsh to Haiku.
<divad27182>
It works!
<divad27182>
but a few of the tests don't
<divad27182>
mostly either from blindly assuming you can hard link.
HaikuUser has joined #haiku
<divad27182>
or from assuming they can write "getent passwd $(id -un) " to get the configured home directory. (getent not found)
<divad27182>
I think that leaves two, one of which was the error message on executing a directory.
HaikuUser has quit []
<waddlesplash>
hardlinks are only not supported on BFS
<waddlesplash>
if you mount an EXT4 partition, hardlinks will work there I believe
<divad27182>
the other is somehow tcsh starts out in a non-standard configuration (echo_style=sysv instead of echo_style=both)
<divad27182>
worth knowing...
<divad27182>
I would have to give it more disk space.... but I can do that.
AndrevS has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
<divad27182>
the weirdest part was dealing with the library free()ing something that didn't come from malloc().
<divad27182>
tcsh provides its own malloc(), calloc(), realloc(), and free().
<divad27182>
so.... If I get this fully working, can it be contributed somehow?
<divad27182>
waddlesplash: can haiku edit the disk partitioning without destroying and recreating the partition?
<divad27182>
As in, can a peal a gig off the end and reallocate it as ext4?
<waddlesplash>
not yet
<waddlesplash>
there are experimental patches for BFS resizing but they weren't ever finished and merged
<waddlesplash>
other filesystem drivers don't support it iether
<waddlesplash>
either
<waddlesplash>
one of these days I should likely see if we can do it for NTFS at least, because we use NTFS3G's code
<divad27182>
I'm not sure that does it either.... I have an NTFS disk I mount that still has its original size... I think it just marked the later stuff as used or bad or some such.
<divad27182>
OTOH, I think Window's own code can actually do it.
<divad27182>
not sure if that was NTFS3G or one of the others...
<divad27182>
can haiku mount a raw disk or does it require partitioning?
<waddlesplash>
raw disks should work
<divad27182>
and I suppose it doesn't notice new disks being attached...
<waddlesplash>
on what?
<waddlesplash>
USB it will definitely detect hotplug
<waddlesplash>
SATA, maybe not
<divad27182>
well... I fear this might be IDE.... SATA should do hotplug, as should the moden bus whose name escapes me at the moment.
<divad27182>
PCIe...
Skipp_OSX has joined #haiku
<divad27182>
OK, how do I set the name?
<divad27182>
more properly, how do I set where it mounts??
<waddlesplash>
why would you want to change that?
<waddlesplash>
it mounts wherever the name is
<divad27182>
it mounted at "/1 GiB Linux Extended File System 2-3-4 volume"
<waddlesplash>
that likely means the partition has no name
<waddlesplash>
anyway you should be able to rename the mountpoint I think
<divad27182>
raw disk...
<divad27182>
ok, how?
<waddlesplash>
well, the filesystem might name a name
<waddlesplash>
*have a name
<waddlesplash>
divad27182: the normal way?
<waddlesplash>
it's just a directory
<divad27182>
well, I shutdown, labeled the drive with "e2label", restarted haiku, and now it uses the label...
* phschafft
offers waddlesplash a cookie for the good work.
<divad27182>
and I appear to have crashed Haiku. It didn't even do KDL.
<phschafft>
divad27182: ref. the exec question: keep in mind that interpreter generally means something different in this case than what people might think.
<phschafft>
the interpreter in this case is *not* what is used to run scripts and/or is given in the shebang.
<divad27182>
well, in this case it means a helper program to run the main program. There are very few things on Linux that do not use one.
<divad27182>
I think the only one I have is busybox.
<phschafft>
the interpreter is what loads the binary image, e.g. the ELF.
<divad27182>
yes...
<phschafft>
it's the job of the interpreter to read the ELF and do things according to that.
<divad27182>
and it gets worse... that may require an interpreter...
<divad27182>
and then a third...
<divad27182>
I do those often enough
<divad27182>
Linux allows 4 layers.
<divad27182>
Which if I built perl for another architecture, I could hit.
<waddlesplash>
Skipp_OSX: I am looking into the menu bug more deeply. Any idea why _Install and _Uninstall are invoked manually? Why aren't they always invoked by the Attach hooks?
<waddlesplash>
and Detach
<Skipp_OSX>
I mean, no idea
<Skipp_OSX>
That's how it's always been.
<divad27182>
19662? :-)
<Skipp_OSX>
If the B_ESCAPE handler calls _Hide() and deletes the menu window the second _Hide() should have no effect.
<Skipp_OSX>
The one at the end of _StartTrack()
<waddlesplash>
the _Uninstall has an effect tho
<waddlesplash>
and that's what does the UAF because nothing uninstalled
<Skipp_OSX>
also I can't verify this but BObjectList has no virtual members other than the destructor and no private members so should be same size as BList
<waddlesplash>
so, I want to try rewriting this to use the hooks
<waddlesplash>
Skipp_OSX: it inherits PointerList which has a bool member
<Skipp_OSX>
I see
<Skipp_OSX>
I'd have to verify if class size changes but idk how to do that.
<waddlesplash>
it definitely changes
<Skipp_OSX>
hmmm ok
<waddlesplash>
and that fix only masks the problem because we still have a use after free anyway
<waddlesplash>
so let's fix this the right way by using hooks
<waddlesplash>
quite possible there is some unimplemented feature in our ext4 driver which makes this not work
<divad27182>
could be something about directory indexing...
<Skipp_OSX>
single directory not full path length?
<divad27182>
in particular, the conversion to index.
<divad27182>
yes, lots of files in one directory.
<Skipp_OSX>
oh number of files in a directory, got it
<divad27182>
that seq/touch I pasted would get ~100 files on ext4
<divad27182>
full 1000 on ext2 or bfs
<Skipp_OSX>
ext[234] filesystems have a fixed maximum number of inodes; every file or directory requires one inode. You can see the current count and limits with df -i
<Skipp_OSX>
But it's unlikely you're hitting that limit especially not with 1000 files
<Skipp_OSX>
bin dir usually has more than that
<divad27182>
well, it isn't just the number of files, but the length of the filename... and the filesystem.
<phschafft>
strange problem.
<phschafft>
maybe the data block size is limited by the driver for directories?
<phschafft>
as in that it can only use one block or something.
<phschafft>
if it uses 1 block at 8KB max that would give you around 100 files with the above specs.
<divad27182>
I think that ext4 switch to an "indexed" mode when the directory gets large enough. I think thats what I'm triggering. In any case, it is now ticket #19665.
* phschafft
nods.
<phschafft>
no matter what it is, it is a good catch :)
<divad27182>
I'm trying to quantify my unexpected crashes.
<divad27182>
:-)
<divad27182>
I think I've submitted 5 bugs in the last 3 days... and now waddlesplash has fixed 2 of them. :-)
<phschafft>
:)
<Skipp_OSX>
usually easy to fix if there are steps to reproduce
* phschafft
should write some random tickets, but also fix them himself...
<Anarchos>
fancy2209[m] the G3 is a known limitation, there was some lenghty discussion in the OpenBeos mailiing discussion some decades ago (around 2003 maybe)
<Anarchos>
fancy2209[m] long story short : G3 is too much a closed platform for open beos team to be able to port the kernel on it
<Skipp_OSX>
idk get what the love for PPC is about
<Skipp_OSX>
last PPC Macs shipped in 2005 it's a dead platform.
<gordonjcp>
I liked MIPS :-/
<Skipp_OSX>
At least with Arm and RiscV I see how the future holds promise.
<gordonjcp>
I wonder if there's an open reference DEC Alpha platform
<divad27182>
One interesting advantage to MIPS: the instruction set was simple enough to write easy emulators for it. The pure Java SQLite implementation is by compile it to MIPS and emulating it in Java. (though that was years ago...)
<divad27182>
(I think that was MIPS...)
<Skipp_OSX>
no explanation forthcoming
<divad27182>
Strange problem. I have a bunch of files ~18 Meg each. I try to copy them from one directory to another, at the command line, and Haiku just STOPs. total shutdown.
<Skipp_OSX>
makes no sense for Haiku to support PPC at this point
<Skipp_OSX>
I guess only for the historical reasons that BeOS did
<nekobot>
[haiku/haiku] 27e99e3a64a7 - su: Don't ask root for passwords.
<nekobot>
[haiku/haiku] e6de99eb01bf - multiuser: Change all groups in setup_environment.
Nasina has joined #haiku
<fancy2209[m]>
<Anarchos> "fancy2209 the G3 is a known..." <- G3Beige is the only one that goes this far
angrystar17043 has quit [Remote host closed the connection]
<fancy2209[m]>
* G3Beige on QEMU is the, * this far though
angrystar17043 has joined #haiku
<fancy2209[m]>
<Skipp_OSX> "last PPC Macs shipped in 2005 it..." <- Xbox 360, Wii, PS3 and WiiU are mainly why people like it
<fancy2209[m]>
<Skipp_OSX> "At least with Arm and RiscV I..." <- I mean POWERPc too, POWER11 is coming out soon
<fancy2209[m]>
I wanted to get haiku working on the Wii
<fancy2209[m]>
That's why I started working on this
<Skipp_OSX>
I see
<fancy2209[m]>
@waddlesplash I hadn't found the ticket, apologies for opening a new one
<waddlesplash>
that's fine, it happens
<fancy2209[m]>
I had been told to open a ticket a few months ago and that's why I did so
<fancy2209[m]>
Will try to grab a serial log from QEMU asap
<fancy2209[m]>
Should I move to that thread?
<fancy2209[m]>
Also I mainly used the unfinished patches for working linker scripts
<waddlesplash>
the unfinished patches change a lot of memory management stuff
<waddlesplash>
it's possible that with them things get futher
<fancy2209[m]>
Or less
<fancy2209[m]>
But I see, I'll rebuild just with the linker scripts then
<waddlesplash>
but really you are not going to get very far here unless you start reading low-level PPC docs and playing with the source code
<waddlesplash>
fancy2209[m]: I have seen logs from yn0ga that get further
<waddlesplash>
on the forums
<waddlesplash>
with what hardware, I don't know
<fancy2209[m]>
Oh I see
<fancy2209[m]>
waddlesplash: I was trying to but wasn't having much success unfortunately. I really need to learn more in the osdev department
<fancy2209[m]>
I'll rebuild haiku with just the linker scripts since those are required for building as far as I remember and post the logs, sorry for not attaching them when making the issue
<waddlesplash>
now that the double fault handlers finally work properly, triple faults are a lot rarer
<divad27182>
problem is the available build of qemu was done by incompetent people at debian: No gtk, No vnc, and spice requiring a client newer than what is available.
<phschafft>
divad27182: which version on Debian are you on?
<divad27182>
Devuan Chimaera
<phschafft>
generally speaking it works within virt-manager.
<phschafft>
that is within Debian.
<divad27182>
right, but it says it needs 8.0, and the available version is 7.0
<phschafft>
of?
<phschafft>
it?
<divad27182>
virt-viewer, actually.
<phschafft>
I said virt-manager
<divad27182>
I don't use GUI wrappers...
* phschafft
has no idea how that relates to the problem at hand.
<divad27182>
"The virt-manager application is a desktop user interface for managing
<divad27182>
virtual machines through libvirt."
<Habbie>
divad27182, i'm confused. you call debian people incompetent but you are not running debian.
<phschafft>
I'm aware of what that software is and does.
<Habbie>
divad27182, surely your complaint is misdirected. it is also rude.
<divad27182>
Devuan is a debian fork that mostly draws from debian.
<Habbie>
it is not debian
<Habbie>
please direct your complaints accordingly
<phschafft>
All I can confirm that qemu on Debian worked (actually running Haiku!) at least all the way from 4 to 12.
<divad27182>
the qemu and virt-viewer packages ARE debian
<phschafft>
divad27182: and on Debian they work.
DKnoto_W has joined #haiku
<divad27182>
well... apparently they did a backport that required other software they didn't both to backport....
<divad27182>
I may have to roll-back my qemu installs
<divad27182>
annoying...
<phschafft>
so you are not only on a different OS but also using backports?
<phschafft>
I'm currently on Debian 12 with a Haiku guest. if you have any questions I'm happy to answer. Given we keep it technical.
erysdren has joined #haiku
Aedil has quit [Quit: leaving]
DKnoto has quit [Ping timeout: 480 seconds]
<phschafft>
mau erysdren.
<phschafft>
hope you are fine, was waiting for you already all day ;)
<erysdren>
hi
illwieckz__ has quit []
<phschafft>
:)
<phschafft>
all good?
Nasina has quit [Read error: Connection reset by peer]
Nasina has joined #haiku
<erysdren>
yeh
<phschafft>
:)
<divad27182>
well... I managed to get qemu and the spice viewer running..... and the damn thing worked. Drat. that means it is probably the Xen that is at fault.
<phschafft>
not sure about that, but it is a hint.
Nasina has quit [Ping timeout: 480 seconds]
DKnoto_W2 has joined #haiku
<divad27182>
I've got issues with the silly GUI. When I try to close a window, move the cursor up into the title area, Haiku decides to FRONT the application behind the window I want to kill. How do I stop this?????.
<x512[m]>
Is it related to Pe?
<x512[m]>
There is a bug in Pe (or Haiku app_server) that cause bringing Pe window to front when tooltip message is displayed.
DKnoto_W has quit [Ping timeout: 480 seconds]
<waddlesplash>
divad27182: I don't think I've ever had that happen lol
<divad27182>
how do you close a window?
illwieckz has joined #haiku
Nasina has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<waddlesplash>
I move it to the title are and click the button
<divad27182>
thats what I try....
<divad27182>
and the window behind the title jumps forward
<bjorkintosh>
divad27182: is this through vnc or something?
<bjorkintosh>
or haiku on bare metal?
<divad27182>
well, VNC, and now spice
<bjorkintosh>
are you sure those might not be the actual problem?
<bjorkintosh>
I can try to play along. which window are you opening and closing?
<divad27182>
seems unlikely, save that I have to move slowly and carefully.
<divad27182>
well, I would, for instance, click on the disk icon, getting a list of files/directories.
<divad27182>
click on a directory there, get a second window.
<divad27182>
click again, third window.
<divad27182>
these stagger down the screen.
<divad27182>
then I want to close the second.
<divad27182>
I click to focus, and move for the button.
<divad27182>
and the first window pops forward
<bjorkintosh>
does Alt-w not work for closing the window?
<divad27182>
but I think it also happens with multiple apps involved.
<divad27182>
alt-w... hmmm.
<divad27182>
would that there were a nice simple cheat sheet for these...