<trungnt2910[m]>
Now HyClone can get official Haiku updates from pkgman.
<trungnt2910[m]>
Extended attributes emulation is also implemented.
_paul0 has quit [Quit: Leaving]
<andreasdr[m]>
Nice
AlienSoldier has quit [Quit: Vision[]: i've been blurred!]
jmairboeck has joined #haiku
skoe has quit [Quit: Vision[]: i've been blurred!]
pvalue has quit [Quit: Vision[]: i've been blurred!]
xet7 has quit [Quit: Leaving]
<Begasus>
nice trungnt2910[m] !
<Begasus>
think I need to install 32bit native on the other laptop, running 64bit, is no use if I can't check 32bit recipes ;)
<andreasdr[m]>
Damn
<Begasus>
I'm 99% sure if it works on 32bit, it will also on 64bit, but not the other way round (and install paths differ due to mixed arch on 32bit) :)
<nephele>
damn waiting for build packages to be activated
<Begasus>
give it a kick (gotta love that one ... NOT) :)
<nephele>
i don't wanna delete the workdir, my patch will leave
<Begasus>
leave the changed files open?
<nephele>
eh...
<Begasus>
hp -c ... hp -b ... (save the open files again) ...?
vdamewood has quit [Remote host closed the connection]
<nephele>
what's hp?
vdamewood has joined #haiku
<Begasus>
hp = haikuporter (if you changed that in your profile (alias))
<Begasus>
not bad, from 25% test passed to 100% passed for librsync ;)
<nekobot>
[haiku/haiku] f6395def5241 - PowerStatus: Use BNumberFormat for percentage value
<Begasus>
If so better to enable static and disable shared?
<Begasus>
x512[m], could you comment on that PR? lets see if more comments follow
<x512[m]>
At least Haiku internally use static AGG.
<Begasus>
yeah, jamfile explicitly mentions it
<x512[m]>
What software use HaikuPorts AGG package?
<Begasus>
from what I see, only gnash
<x512[m]>
Added comment to PR.
<x512[m]>
If Gnash compiles and works fine with static AGG, it will be better to switch to static one.
<Begasus>
thankx x512[m]
<Begasus>
gnash is disabled for 32bit, will run a checkbuild later
<x512[m]>
trungnt2910: Around?
<trungnt2910[m]>
はい。
<x512[m]>
Is it possible to make Haiku->Linux devfs ioctl forwarding?
<x512[m]>
For example to access Linux kernel graphics drivers (DRM) from Haiku applications.
<trungnt2910[m]>
Depending on the code.
<trungnt2910[m]>
By devfs you mean the virtual device mounted at /dev/?
<x512[m]>
Mapping Linux /dev file to Haiku VFS with ioctl support.
<trungnt2910[m]>
It's possible if the code does not collide with any basic Haiku ioctl ops.
<trungnt2910[m]>
monika (the syscall emulator embedded in the modified libroot) forwards all ioctl ops that are larger than B_DEVICE_OP_CODES_END to hyclone_server.
<trungnt2910[m]>
Then the ioctls can be implemented there.
<x512[m]>
Forwarding ioctl on hyclone_server may cause problems because opened FD will belong to server instead of client.
<trungnt2910[m]>
It's also possible to check if the path starts with /dev before doing anything else.
<trungnt2910[m]>
Without forwarding anything to hyclone_server.
<x512[m]>
Are opened FDs belong to server? Do clients gracefully release resources on crash?
<x512[m]>
Also what currently happens if run app_server?
<trungnt2910[m]>
Currently, all fds belong to the emulated process, not the server.
<trungnt2910[m]>
The server only stores paths of the files that these fds point to.
<trungnt2910[m]>
x512[m]: It forks once, runs into a few stubbed `watch_node` syscalls, and then goes into an infinite loop.
<trungnt2910[m]>
The infinite loop takes a suspiciously high amount of CPU.
<x512[m]>
Does it write something to log?
<trungnt2910[m]>
x512[m]: If you mean the kernel log (`_kern_debug_output`) then no, except for the stubbed `watch_node` syscalls.
<trungnt2910[m]>
Also, launch_daemon doesn't seem to launch app_server automatically.
<trungnt2910[m]>
trungnt2910[m]: Actually it doesn't this time. What happens instead is that after launching `app_server`, it is impossible to launch any other HyClone apps. The `hyclone_server` just hangs.
vdamewood has quit [Remote host closed the connection]
vdamewood has joined #haiku
pvalue has joined #haiku
<x512[m]>
hyclone_server freeze sounds like a bug.
vdamewood has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
mmu_man has joined #haiku
gouchi has joined #haiku
MajorBiscuit has joined #haiku
MajorBiscuit has quit []
MajorBiscuit has joined #haiku
pvalue has quit [Quit: Vision[]: i've been blurred!]
pvalue has joined #haiku
nephele has quit [Quit: Vision[]: i've been blurred!]
nephele has joined #haiku
<nephele>
Begasus: I've mostly finished minetest 5.7.0, but I've send a PR upstream to fix a build issue in irrlicht, not sure if i should just patch it out locally or wait for it to be merged and then use it directly :)
Kokito has joined #haiku
<Begasus>
nephele, you probably don't know when it will be merged upstream I gues, so add it localy, and if merged on the next release it can be dropped?
<nephele>
hmm, yeah. but i already need to use a newer git revision to compile it, the last released version tries to use X11 even if sdl2 is configured
<trungnt2910[m]>
<x512[m]> "hyclone_server freeze sounds..." <- Yeah, it seems to be a deadlock.
<Begasus>
issue is the same nephele ? (using a release or a srcGitRev tag) when patching is needed?
<nephele>
Yeah, but i could also use the github "fork" as upstream
<Begasus>
ah, like that :) mostly we want to keep as much upstream SOURCE_URI as possible, if that doesn't work you could mention that in the PR
Kokito has quit [Quit: Vision[]: i've been blurred!]
pvalue has quit [Quit: Vision[]: i've been blurred!]
Blendie has joined #haiku
floof58 has quit [Remote host closed the connection]
<Begasus>
does Haiku support IPV6_TCLASS? (I see some patches disabling it at haikuports and getting an error for it on latest burp)
floof58 has joined #haiku
zard has joined #haiku
<nephele>
Urgh, cmake with it's "this was packaged wrong" is really getting on my nerves
<nephele>
this time it's qtlinquist
<PulkoMandy>
nephele: I don't know if you're still using Renga (you're not in the XMPP chatroom for it anymore), but your issue with bookmarks should be fixed (it was a bug in gloox and is fixed with a patch to gloox 1.0.26 in the current recipe)
<nephele>
PulkoMandy: I wasn't using it, a friend of mine hosted the xmpp server but that went down. I haven't gotten around to getting a new server or account yet
<x512[m]>
Does it change something with running app_server?
<trungnt2910[m]>
<x512[m]> "Does it change something with..." <- Like I said, instead of deadlocking `hyclone_server`, `app_server` now becomes harmless.
<x512[m]>
Do Haiku strace work?
<trungnt2910[m]>
It doesn't because the debugger API is stubbed.
<trungnt2910[m]>
strace is actually on the top of my current todo list, I'll start working on it tomorrow.
<trungnt2910[m]>
- At that time there was no way for `hyclone_server` to interrupt and force another process to spawn the nub thread (now there is, I implemented such a mechanism for `transfer_area`).
<trungnt2910[m]>
- The Haiku debugger protocol is undocumented and a bit complicated.
<trungnt2910[m]>
<x512[m]> "Does it change something with..." <- Some Haiku tools start to recognize the existence of `app_server`. For example, `screenmode` instead of just quitting now hangs and waits for a reply.
<x512[m]>
"nub thread" is a kernel thread on Haiku.
<x512[m]>
Maybe it can be created on hyclone_server side.
<x512[m]>
Haiku kernel allows to create kernel thread inside any process.
<x512[m]>
spawn_kernel_thread_etc
<trungnt2910[m]>
It still seems to belong to the team and reads team data though.
<x512[m]>
I see your hyclone_server can read client process memory.
Skipp_OSX has joined #haiku
floof58 is now known as Guest10520
floof58 has joined #haiku
<trungnt2910[m]>
Yes, we could spawn the thread in hyclone_server.
<trungnt2910[m]>
However, only notifying that the team is being debugged itself requires interrupting the team.
Guest10520 has quit [Ping timeout: 480 seconds]
<trungnt2910[m]>
Either way the debugged HyClone process would have to know the nub port id and report its own syscalls to it.
<trungnt2910[m]>
It is possible to make hyclone_server into some kind of gdb that actually reads the process's symbols and install traps on _kern_*, but that's way too complicated.