<Begasus>
yann64[m], did you try to build code::blocks? (should be do-able if I look at other ports on repology)
<yann64[m]>
Begasus: nope, will try in the coming days
<andreasdr[m]>
I used to use CodeBlocks years again. I'd love to try CodeBlocks on Haiku :)
<Begasus>
andreasdr[m], I think we got most covered in terms of requirements, it's a matter of grabbing the source, patches (you can find through repology) and build yourself ;)
<kquote03[m]>
<x512[m]> "GTK4 with Wayland backend." <- ???? HOW???
<x512[m]>
Serverless Wayland implementation over Haiku API.
<kquote03[m]>
:0
<x512[m]>
Wayland server is loaded with dlopen into application process.
<x512[m]>
No sockets.
<tqh>
great now port XFCE to wayland j/k :)
<kquote03[m]>
Ah yes, just casually PORTS WAYLAND TO HAIKU
<kquote03[m]>
YOURE A MAD MAN.
<tqh>
if we break down packages to seekable filestreams + metadata it might be easy to do a networking version of packagefs...
<PulkoMandy>
you can already run packagefs over a network filesystem of your choice I think
<PulkoMandy>
not sure anyone has tried
<tqh>
but will it it behave like a remote fs or networked block device or read the whole package?
<tqh>
if you have a package that is 2GB, but you want to run the 20KB command, can it just do that already?
<x512[m]>
tqh: XFCE? Do you mean some XFCE programs?
<tqh>
x512[m], no I was joking only do not do it. XFCE wayland port has been in progress for years and years. Someone recently showed some working state though.
<nephele>
kquote03: wayland is a server api, so "serverless" means using one of the application apis instead, so it's not really wayland ported
<x512[m]>
It allows to run applications that use libwayland-client.so.
<nephele>
tqh: networked version of packagefs would be funky, but with remote app_server, and archiving apps you can even do more insane stuff of archiving running apps and sending them to a different machine to run, but keep them displayed here with the remote app_server
<x512[m]>
My implementation is API/ABI compatible with libwayland-client.so, but not with Wayland socket protocol.
<nephele>
x512: yes, one application lib. iirc waddlesplash investigated that but didn't want to do it because there are too many different client libs wayland clients use, but only basically one that is still used for X11
<nephele>
in essence it's the same as xlibe which is kind of cool :)
<nephele>
or well, similar model
<x512[m]>
I still don't know any Wayland applications that don't use libwayland-client.
<nephele>
isn't that the reference implementation with weston and such?
<nephele>
don't enlightenment and kde have their own client stuff?
<x512[m]>
I managed to run some Weston demos. It use libwayland-client.
<tqh>
nephele, that is not what I want, I just want to instantly use packages while they are downloading. Now you download the full package, but if you get the metadata, and start the app it could download files on demand, and backfill rest of package content.
<nephele>
tqh: i think most of the time currently is spend redownloading the list of packages on every invokation of pkgman
<nephele>
it should probably check if it can satisfy your requirements with the local list first, do that and only update if it encounters an error or it was a long time since the last one
<tqh>
that is a different problem, you talk pkgman, I talk packagefs.
<tqh>
today is download, mount, run. I want mount, run, download :)
<nephele>
networked mount kind of?
<nephele>
I also want mDNS discovery of other haiku machines to use their local repos, with your idea you could network mount remote packages of machines in the lan and run them without having to download the damn libreoffice package
<tqh>
yes, that is it
<tqh>
if the repo has filestreams with deduplication, and metadata outside the stream, you can potentially save a lot of space. And if filestreams are seekable, you can mount while downloading..
<tqh>
so if filestream a in package A is same as filestream b in package B we only store it once. No need for name or date to be same as that is part of metadata.
<nephele>
yeah, the packages are compressed but not to each other, especially stuff like haiku packages could be deduplicated a lot
<tqh>
how often does the font files really change...
<nephele>
but this raises the question, atleast for me: why don't we do this on the next iteration of bfs instead? I don't see the specificty on packages for this stuff
<nephele>
zfs for example does this on the filesystem level, deduplicating blocks.. it's more of a server filesystem but it does this nicely for example
<tqh>
bfs deals with low level block storage?
<tqh>
this would be higher level remote file handling?
<tqh>
I like doing one thing and doing it well.
<nephele>
I don't see the point in deduplicating it in the higher level if it would already be deduplicated in a lower level
<tqh>
packges != file storage IMO
<nephele>
they *are* stored on the file storage
<nephele>
on bfs, i mean. the source packages
<tqh>
yes, but so are images, and we don't build image handling into bfs
<nephele>
and i didn't say we should?
<nephele>
I just think deduplicating of file contents is a thing bfs should learn, in the sense that bfs2 should do that
<tqh>
packages present it's own set of problems are what I am saying. Lets not add those problems to the amount of problems to write a FS.
<nephele>
We solve some problems with packages nicely that freebsd for example solves with zfs and boot enviroments
<tqh>
how will that help packages?
<nephele>
some stuff zfs solves we can never do with packages, for example restore a previous version of normal files
<x512[m]>
> xkbcommon: ERROR: failed to add default include path /Haiku/data/packages/wayland/libxkbcommon/build.x86_64/install/share/X11/xkb
<x512[m]>
Where can I get X11 keyboard layout files?
<tqh>
I havn't used FreeBSD when it has ZFS so can't relate.
<nephele>
bfs is very nice for the queryability and file indexing and attributes, but it can be better
<nephele>
well, they basically copy what illumos/openindiana (opensolaris) did with zfs boot enviroments
<tqh>
but downloading and running packages is what I want to talk about
<nephele>
yeah, i am arguing that deduplicating is not directly related and should be done on the fs level instead
<nephele>
x512: i would check what the debian packages lists as source files in such cases
<tqh>
packages are downloaded from a server you need to have the right data + checksumming, I can't see how having proper deduplication in FS can help with that.
<nephele>
how do you intend to deduplicate packages in transit against pavckages stored on the disk without involving the fs? and if the fs already knows you have the block you can simply skip downloading it
<nephele>
I suppose that would work better if packages were a data structure in the fs itself
<nekobot>
[haiku/haiku] c912a3f2307a - xfs: Node, B+Tree based directories and v5 files support
<dpirate>
Hi, is there a full mirror of ftp.be.com somewhere. No torrent on archive.org and only found a partial one on Google.
<tqh>
a stream has unique id, metadata points to unique id(s). To download package you need to get metadata + unique ids. The unique ids can be used by several packages as long as content is same.
<nephele>
isn't that essentially exactly what deduplication based on nodes in a tree is?
<tqh>
Yes but this is for fetching files from a package repository.
<nephele>
which, i would argue, is something the filesystem should track
<nephele>
yeah. but, why should the filesystem not have this info and keep track of the tree?
<tqh>
If you can build a CDN that way...
<nephele>
you can ask the filesystem then directly if this node is already present
<tqh>
how are you going to have repository copies, when unis only give you ftp or http servers?
<tqh>
Ask them to install Haiku with our BFS2 so we can have free hosting of files?
<nephele>
http is a protocol, you can serve whatever you want :P
<tqh>
It's two completly different problems.
<nephele>
no, it isn't. packagefs and the packages within it *already* do chunking and compress those individually
<tqh>
yes, but a filesystem doesn't magically solve file distribution over internet, especially not when you don't control the servers.
<dpirate>
PulkoMandy, any access would work for me
<nephele>
No, it doesn't solve it on it's own, but i am arguing that bfs2 is a part of that solution, and we shouldn't redesign packagefs above it and treat it as a dumb storage
<dpirate>
I want to mirror indeed
<tqh>
nephele, I strongly disagree, reducing duplication between packages is not the same as A package or the filesystem it lands on. But I think there is no point in discussion further.
<nephele>
bfs2 would indeed be some ideas we already did with packagefs implemented on the filesystem level, similar to how zfs btrfs and such do it now
<nephele>
tqh: I don't see why you think that a node in a package would be fundamentally different to a node in the filesystem
<tqh>
I rather look to bcachefs in that case.
<tqh>
I've explained that and you havn't explained how you can do content distribution in a good way.
<nephele>
You didn't ask, and i don't think the question makes any sense
<nephele>
nobody had to redisign https for FreeBSD just because they use zfs
<nephele>
if you want to download objects from a https server that works just fine, now.
<tqh>
I did, see the CDN question above. You just talk about file deduplication only.
<nephele>
" If you can build a CDN that way..." that isn't a question
<nephele>
and I don't see how "we should use the same nodes for both" has anything to do with CDN's at all
<tqh>
Yet, you skip talking about a way to do it, instead argue about what was said.
<nephele>
I don't understand what your question is, you clearly know how to download stuff over https, so I don't understand what problem you feel isn't solved
<nephele>
if you have a filesystem with nodes in a graph/tree, and base the packages on the same set of nodes the server can present to you a package, which is a list of nodes... and then you can compare it against the ones you already have and download the ones you don't
<tqh>
It was the whole start of the conversation above. And now you even ignoring the deduplication between packages. As we can't make progress lets end this discussion.
<nephele>
How am I ignoring deduplication between packages? it was basically the only thing i talked about. But fine
<tqh>
> you clearly know how to download stuff over https, so I don't understand what problem you feel isn't solved
<tqh>
Please, move on.
<nephele>
>But fine
<nephele>
I already agreed to stop the discussion, not sure what quoting stuff helps. And yes, your question does not make any sense to me, so I don't see how you expect a proper answer to it.
<tqh>
then lets move on shall we
x10z has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<rennj>
He was also the founder and chief executive officer of Lighthouse Design, Ltd., a software company focused on the NeXTSTEP platform. Lighthouse was acquired by Sun in 1996.
<rennj>
oh yeah the cube boxes
<rennj>
netwinder boxes or whatever..linux cube servers they paid way to much for
<rennj>
Sun bought MySQL for $1 billion in 2008,
<rennj>
StorageTek for $4.1 billion in 2005
<rennj>
Sun's $2 billion purchase of Cobalt Networks
<rennj>
those boxes..cobalt cubes..programmed them from like lcd interface on box
<PulkoMandy>
I'd recommend setting up the Gerrit commit hook that automatically adds Change-Ids locally in your git clone so all your commits have a change-id
<rennj>
50 languages!!!
<rennj>
Pe support that many?
<Begasus>
Geany already build, but doesn't show any icons
<rennj>
The Wages of Fear (1953) vs Sorcerer (1977)
<rennj>
both good, but i prefer the 53 version
HaikuUser has joined #haiku
HaikuUser has quit []
<andreasdr[m]>
x512: Is this gtk using wayland?
<x512[m]>
Yes, serverless Wayland.
<andreasdr[m]>
Wow. Nice!!!
<andreasdr[m]>
Sir. Is it committed somewhere? When can we start to use it? :DDD Just asking.
<x512[m]>
It is incomplete for now. For example it have no keyboard input yet.
mmu_man has quit [Ping timeout: 480 seconds]
B2IA has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
<andreasdr[m]>
Ah i see. Looking forward to test it later :)
<Begasus>
andreasdr[m], codeblocks still building here :)
gouchi has joined #haiku
<andreasdr[m]>
Nice!!!!
<andreasdr[m]>
Please tell me when finished, would love to try.
<Begasus>
Only in Terminal, so nothing to publish :P
<x512[m]>
Ii is currently started with command like: `WAYLAND_DEBUG=client XDG_RUNTIME_DIR=/tmp LIBRARY_PATH=/Haiku/data/packages/wayland/gtk/build.x86_64/install/lib:/Haiku/data/packages/wayland/libxkbcommon/build.x86_64/install/lib:/Haiku/data/packages/wayland/wayland/build.x86_64/install/lib:/Haiku/data/packages/wayland/wayland-server/build.x86_64:$LIBRARY_PATH gtk4-demo`
<andreasdr[m]>
Looks complicated.
<x512[m]>
Because everything is installed to different build directories.
<andreasdr[m]>
Ok
<Begasus>
And that's only 64bit ;)
<andreasdr[m]>
BTW: Can someone recommend a small Haiku X64 box?
<andreasdr[m]>
:D
<rennj>
steam deck!!! amd! ftw!
<andreasdr[m]>
AMD for sure
<andreasdr[m]>
Small would be nice
<andreasdr[m]>
With a PCIE for my graphics card: :D
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
<yann64[m]>
Anybody having issues doing a "git clone https://review.haiku-os.org/haiku"? Just tried on a new Haiku install and I have an error indicating http.postBuffer being too small
<yann64[m]>
Never had to touch any config parameters on previous installs
<ermo[m]>
x512: You asked me if my i7 2600k w/HD7970 system works with Haiku the other day? I can report that, yes, it does.
<ermo[m]>
Which means I'm ready to test your GCN SI driver work at your leisure.
HaikuUser has joined #haiku
HaikuUser has quit []
<Begasus>
g'night peeps
Begasus has quit [Quit: Leaving]
Begasus_32 has quit [Quit: Vision[]: Gone to the dogs!]