<AlienSoldier>
coolcoder613 got it, thank you. used yt-dlp --list-formats and saw that you don't enter a format but an id to the wanted listed format. So in my case it was 360p and the only one with audio was 18, so yt-dlp --format 18 "my url".
OrangeBomb has quit [Read error: Connection reset by peer]
OrangeBomb has joined #haiku
OrangeBomb has quit [Read error: Connection reset by peer]
<botifico>
[haiku/infrastructure] kallisti5 fa04665 - playground/toolchain: Finish test container-based builder of toolchains
scantysnax has joined #haiku
* OscarL
waves scantysnax
<scantysnax>
hi OscarL!
<OscarL>
Is "Single navigation mode" broken for /boot/system now? (hrev57924). Double clicking a folder there opens a separate window :-(
<waddlesplash>
really?
<waddlesplash>
OscarL: single window still works here for /boot/system
<OscarL>
yeah, no idea what is going on.... I was testing things to report an issue with tracker disabling "New..." incorrectly for /non-packaged/, and hit this weird thing.
<OscarL>
will update to latest hrev, and see if that changes anything before finishing the new ticket.
<OscarL>
waddlesplash: seems OPT is gentting stuck or something (couldn't type on Terminal, LOL). Now it got "unstuck" and things went back to normal :-/
<OscarL>
"cd / && mkdir RAM && mount -t ramfs /RAM" <<< shouldn't that RAM volume be visible under "Disks" ?
<OscarL>
(/me also wishes ramfs used the mount point as "label" instead of always "RAM FS").
<OscarL>
nuff nightlies, /me goes back to beta4 for the Python 3.12.5 update.
<waddlesplash>
OscarL: no need
<waddlesplash>
package builders now based on beta5
<OscarL>
yay! good to know! (my main dev install still on beta4 thou :-D)
<OscarL>
guess its time for the repos switcheroo :-)
* OscarL
adds to the ToDo/Wishlist list: "pkgman update --only-from-repo(s) <repo(s) name(s)>".
<waddlesplash>
you can just update individual packages
<waddlesplash>
"pkgman update haiku" etc.
mmu_man has quit [Ping timeout: 480 seconds]
<OscarL>
Indeed, but it is less convenient if I want to update only the HaikuPorts packages, for example.
linuxmaster has joined #haiku
<OscarL>
I will be happy if we can reliably use "make $jobArgs" again for the optimized Python builds. Worked for 3.13 on nightlies. Not sure if due to changes on Python 3.13, or changes nightlies/beta5 :-D
<Skipp_OSX>
waddlesplash commented on #18975 it is fixed by Shortcuts refactor thankfully, I even identified the line that fixes it!
<OscarL>
> OscarL: "cd / && mkdir RAM && mount -t ramfs /RAM" <<< shouldn't that RAM volume be visible under "Disks" ?
<OscarL>
Hi Skipp_OSX! does the above works with some of your pending Tracker patches?
<Skipp_OSX>
I think we're hiding those under Disks for reasons....
<Skipp_OSX>
I doubt lemme check
<OscarL>
(if its just to hide shared_memory... sounds like the wrong reason :-P)
<Skipp_OSX>
no does not fix
<OscarL>
alright, thanks! (just wanted to avoid opening a ticket for that if a fix was comming)
<Skipp_OSX>
I can see the line of code... PoseView.cpp line 1599
<Skipp_OSX>
if (!volume.IsPersistent()) { continue; }
<Skipp_OSX>
well, ok my line 1599 maybe not yours... around there somewhere in AddRootPoses()
<OscarL>
not ideal (didn't *I* made a change on that regard? /me double checks)
<Skipp_OSX>
no that's not it... it shouldn't be persistent it's just we're not showing those.
<OscarL>
welp, flag makes sense there, I think. :-)
<Skipp_OSX>
somebody made a deliberate choice not to show those
<Skipp_OSX>
(apparently before Tracker was even imported into the Haiku tree)
<Skipp_OSX>
no, that's not it darn
<OscarL>
(yeah... I just half-remembered my change). It might make sense to hid /system/var/shared_memory (appeared as a "RAM FS" volume on tracker) which can be confusing, but I don't think hiding user mounted volumes is better :-D
itaipu has quit [Ping timeout: 480 seconds]
tuaris has quit [Quit: Leaving.]
Begasus has joined #haiku
<Begasus>
g'morning peeps
<OscarL>
'lo Begasus!
_justin_kelly9 has joined #haiku
<Begasus>
Hi OscarL!
<Begasus>
problem 1: nothing provides haiku>=r1~beta5_hrev57937_5-1 needed by gdb-15.1-1 ... tss :P
<Skipp_OSX>
we made a couple of ram fs volumes we're using for packagefs and we hide them in Tracker
<Skipp_OSX>
but if you make your own... well we could hide just ours by name and allow you to see the rest.
AlienSoldier has quit [Quit: Vision[]: i've been blurred!]
Chai-T-Rex has quit [Remote host closed the connection]
<Begasus>
'k reading the log it makes (a bit) of sense now :)
ChaiTRex has joined #haiku
<OscarL>
Skipp_OSX: sounds better that way to me at least.
<OscarL>
(from a end-user perspective)
<OscarL>
Haiku can't see ext4 partitions created by VoidLinux. At first I thought it had to do with my use of extlinux bootloader, but same happens on a netbook where I just use rEFInd.
<OscarL>
(there, Haiku still sees an older ext4 partition just fine.)
<OscarL>
using "sudo file -sL /dev/sdaN" on both... the newer shows "(needs journal recovery)" than the older (the one that works in Haiku) doesnt.
<OscarL>
(/me saves that info, plus "tune2fs -l" for a future ticket)
<PulkoMandy>
The systlog may have some complaints from the ext4 driver (maybe some extension flag we don't support yet)
<OscarL>
I'll take a look at the syslog, thanks.
<OscarL>
"ext2: incompatible features not supported: 2000 (extents 40)". only suspicious line (besides "ext2: Invalid superblock!"). would be nice if the correspoding /dev/disk/xxx was mentioned on or near those lines.
<OscarL>
k, "make $jobArgs" still no good for 3.12 with --enable-optimizations (with me still on beta4). Will test that again after I switch to nightly/beta5.
vdamewood has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<Begasus>
or does it actually pull "that" version?
<Begasus>
600 objects to go ... (latest qt-creator)
systwi has quit [Ping timeout: 480 seconds]
<OscarL>
(no clue about what those docker commands actually entail)
<Begasus>
I guess kallisti5[m] will know :)
<OscarL>
grrr. automatically hit enter on a Tracker crash dialog (instead of saving report), thinking it was the "confirm deletion" dialog I was just expecting to popup :-/
<Begasus>
reproduce? :P
<OscarL>
was removing files that now are gone :-D
<Begasus>
heh
c0m47053[m] has joined #haiku
<OscarL>
"Warning: Error: found provides for "lib:libssl >= 3", but none matching the version requirement" guess I *do* need to update...
<OscarL>
on the other hand... shouldn't that " >= 3" be on the devel:libssl line and not on lib:libssl?
<Begasus>
yeah
<Begasus>
lib:libssl should be defined by the devel:libssl in use
<Begasus>
Not going over all those again, did my chair amount of checking on that PR :P
<Begasus>
chair/fair*
<OscarL>
ahh, "fair share"... was wondering what the chair did :-)
<OscarL>
so... 3.10/3.11/3.12 need fixing.
* OscarL
double checks the rest of waddlesplash commits... kids these days! :-P
<Begasus>
2 days in a row this touchpad is annoying me ... :P
<Begasus>
heh
<OscarL>
k, seems only the python 3.10-3.12 recipes were affected. (/me restarts the 40+ minutes build)
<OscarL>
"problem 1: nothing provides haiku>=r1~beta5_hrev57937_5-1 needed by openssl3-3.0.14-1" much better :-P
<Begasus>
lol
<Begasus>
not hitting that so far :)
<OscarL>
updating to r1~beta5_hrev57943-1, /me crosses fingers
<Begasus>
update arived?
<OscarL>
I switched repos to master.
<Begasus>
ah, waiting for that untill we hit the full r1beta5 repo :)
<Begasus>
last time switching beta -> nightly -> beta went bogus on 32bit
<OscarL>
ok, update/reboot for packages from Haiku repo went fine. Doing "full-sync" with HaikuPorts enabled now (even if those python3.1x packages will need to be rebuild).
<OscarL>
I split the update (Haiku vs HaikuPorts repos) to be able to use the improved download speeds :-D
DKnoto has quit [Ping timeout: 480 seconds]
DKnoto has joined #haiku
<OscarL>
"failed to commit transaction: Failed to change the package activation in packagefs: Name in use" :-/
<OscarL>
which one, ffs!
<Begasus>
fix it!! :P
<Begasus>
maybe it's not "package" but "packages" ;)
<OscarL>
would try to fix it, but packagefs code is not fun to navigate around :-/
<OscarL>
good thing I got a "system_installed_packages_built_by_me" bash function :-P
<Begasus>
most of them not even at haikuports I think :)
<OscarL>
success! (now to cleanup the mess of transaction-n folders)
<OscarL>
Begasus: quite the list :-)
<Begasus>
yeah :D
<Begasus>
hence not looking out for a "full-sync" :)
HaikuUser has joined #haiku
HaikuUser has quit []
<Begasus>
khelpcenter is working nice, too bad KIO isn't :P
<Anarchos>
hello Begasus
<Anarchos>
Begasus: i could flash my bios with another HDD containing a Win7 partition
<Begasus>
ah, problem fixed now?
<Anarchos>
Begasus: not totally : the wake on lan works with Win7. But Haiku seems to erase the NIC internal config for the WOL, so it shutdowns the NIC also when the PC shutdowns
<Begasus>
whoops
<Begasus>
still a wip then Anarchos :)
<Anarchos>
Begasus: as for all my projects... I am tired to never finish something
<Begasus>
I'm only doing the easy things Anarchos, it's guys like you that keep the low level development going on :)
<Begasus>
low level as in, things that need to be done that most users aren't even aware of
<Anarchos>
:)
* OscarL
needs a FlatControlLook that does *not* hides the scrollbar knob when the window looses focus.
<Nephele>
OscarL: honestly that sounds more like a bug than a design decision
<OscarL>
no idea, but it is pretty annoying either way :-D
<Nephele>
patch it, add me to reviewers? :P
<Begasus>
+1 Anarchos
<Begasus>
hi nephele! :)
<OscarL>
I might give it a try tomorrow nephele :-)
<Nephele>
Hello Begasus
* OscarL
also needs a proper dark theme for Pe... and for Pe to suppport themes on the first place :-P
* OscarL
looks at FlatControlLook.h.... notices a ".ar" email address. Darn Argentines... they ruin everything!
<Nephele>
pfft, be happy about it, if you use it :P
<Begasus>
OscarL, adjusted colors in Pe when in dark theme here, then saved, works for me :)
<Begasus>
just Vision doesn't work and I have to manualy switch settings file for dark mode
<Nephele>
I would like if apps would use the system theme, or derived theme per standard
<Nephele>
AboutSystem has it's own colors, but it follows the system theme :)
<OscarL>
Begasus: yeah... but I am TERRIBLE at selecting colors that look good. need some pre-made ones.
<Nephele>
OscarL: heh, that's what the generation code should do for you xD
<Anarchos>
speaking of dark theme, a Vision dark theme would be nice to chat in bed :)
<Nephele>
i'm using vision in dark mode
<Nephele>
but you have to manually set most colors *shrug*
<Nephele>
hmm, settings file has my channels and nickserv otherwise i'd just have sent it to you :)
<OscarL>
DrawScrollBarThumb() as a big-ass "if (isEnabled) { /*lots of stuff */ } else ]
<OscarL>
" just DrawScrollBarBackground() :-/
<Nephele>
OscarL: I won't object to you making that a if (!isEnabled) return;
<Nephele>
or something :)
<Nephele>
Anarchos: Renga uses the system theme though :P
<Nephele>
and we support irc colors now
<OscarL>
yeah, will need to remove that "isEnabled" on all that function it seems (for me at least). Will do that tommorow and test. /me gets back to Python for a bit.
<Nephele>
isEnabled shouldn't be the same as no focus though?
<Anarchos>
nephele: that's what i mean by a theme: not be forced to change the million colors of the Vision window :)
<Nephele>
anyhow, isEnabled seems wierd, shouldn't that be fEnabled or enabled or something? :D
<Nephele>
Anarchos: I just want it to derive from OS colors per default, leave customizing specific colors to people who really want that
<OscarL>
lol, I'm getting some weird visual glitches (I blame FlatDecorator)... Seems to leave some traces behind. Will upload screenie.
<Nephele>
the decorator api may be killed altogether, it's very huge for bascially just beeing "render these two buttons and a shine". Maybe it can be moved into the controllook too
<Nephele>
then we can name it system style or something
<OscarL>
Web+ just crashed :-/
<Nephele>
Web+ crashed? Uhh, gotta go
<OscarL>
heh
<OscarL>
general protection fault... pthread_rwlock_wrlock (on a curlThread apparently).
<Nephele>
i gotta leave for real, but i'll be back later in the day :)
<OscarL>
take care.
Nephele has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
x509_store_add -> CRYPTO_THREAD_write_lock... /me "hmm"s, pretending he understands any of it.
<OscarL>
Seems extremely reproducible. it tries to open the previous URL... https://github.com/notifications (guess I need to kill Web+ settings now)
* Anarchos
dreams of a kdl with source files like in gdb...
<OscarL>
removing "Session" was enough to make it open. but crashes on every url I try.
* Begasus
uses 2 settings files for Vision for colors
<Begasus>
OscarL, from checking that PR for openssl3 I remember having issues with Web+ also (crashing)
<Begasus>
iirc it was related to curl
<OscarL>
seems like it yeah.
* Begasus
sticks to R1B4 for a while :P
<OscarL>
yeah :-)
<Begasus>
gonna launch a build for Haiku here (nightly) and check :)
<Begasus>
6 new commits since yesterday :P
* OscarL
tries not to freeze. Darn, it is getting cold!
<Begasus>
22°C atm, pretty ok for me
<OscarL>
-2°C if I can trust my thermometer.
<Begasus>
jikes
<OscarL>
will be hitting 20+ today according to the weather report at least :-) /me crosses cold fingers :-P
<Begasus>
k, default install is still openssl
<Begasus>
heh
<Begasus>
webpositive no crashes
<Begasus>
running full-sync now :)
<Begasus>
lol ... no space left on device ...
<Begasus>
k 3GB ... should be enough :)
<Begasus>
and crash with webpositive ...
<OscarL>
mmm, even with default decorator I'm seeing redrawing problems :-/ (vertical lines "left behind" from a window's borders when they were on top of other window)
<OscarL>
doesn't happens with Be decorator apparently. 8-/
<OscarL>
"problem 1: package openssl3_devel-3.0.14-1 conflicts with openssl_devel provided by openssl_devel-1.1.1w-2", so glad to be able to see that error from hp now :-D
novaphoenix has quit [Quit: gone fishing]
novaphoenix has joined #haiku
* phschafft
joins a little late today, but brings some frish cookies.
<Begasus>
heh, OscarL, clash between Argentina and Belgium atm :)
tuaris has joined #haiku
<Anarchos>
Begasus: don't mess #haiku with the olympics please :)
<OscarL>
I assume sports related? (haven't paid attention at all to the Olympics this time)
<Begasus>
Womens Hocky ... for bronze medal :)
<Begasus>
Anarchos :)
<Begasus>
Neither am I Anarchos, but sometimes there is a trigger here :D
<OscarL>
Seems we *can* use "make $jobArgs" for Python 3.12 with --enable-optimizations (on beta5).
<OscarL>
always caused problems on beta4
<OscarL>
seems upstream Python doesn't test their build options combinations too much, because using --disable-test-modules interferes with --enable-optimizations thou! :-D
<OscarL>
some of the tests run as part of the profile-guided-optimization requires modules that are not present if using --disable-test-modules :-)
<OscarL>
build took 25m22.713s, will re-do without --disable-test-modules, and also to verify make $jobArgs working wasn't a fluke.
<Begasus>
not that bad
<Anarchos>
OscarL: edge cases should not be 'edge' : they are cases in their own full right. It is just the symptom of a lazy programmer...
<OscarL>
welp, each new option adds (at least) a x2 to the posible combinations, so being able to testing each and every single combo is not a given (when not outright impossible).
<OscarL>
but yeah... this one "disabling test makes no sense if using optimizations" shouldn't have happened without at least a warnning :-D
<Anarchos>
OscarL: moreover, they know optimizations are drived by the test results...
<OscarL>
yup, that should have triggered some alarm bells. But I guess they never feel the need to use --disable-tests-modules on the first place. I only tried trying to reduce build times a bit more (as I might make them 2.5x larger if I enable --with-lto :-D)
<OscarL>
bah... getting "libgcov" profilling errors again. Oh well, /me nukes the build and disables make $jobArgs for optimized build again.
zard has joined #haiku
<Anarchos>
i must ask jmairboeck how to use ghostscript...
<OscarL>
for viewing purposes only, you can use it via BGhostView.
<Begasus>
congrats OscarL :)
<OscarL>
Thanks. I assume our ladies won the Bronze medal?
<Begasus>
yep :)
<OscarL>
Wee! (albeit I haven't paid much attention to them since our legendary Luciana Aymar retired from that team).
<Begasus>
lol, you know atleast a team member :)
<OscarL>
her last international game was in... gasp! 2014 :-/
<Begasus>
heh
<Anarchos>
OscarL: for BeTex, i must be able to launch gs in command line
<Anarchos>
OscarL: the trouble is that the old BeOS port had a 'bealpha4' (and others) device, which Haiku port has not. So i don't know how to render in command line
<OscarL>
yeah, those devices I half-remembering using in the past (along with -dNOPAUSE). Now... sDEVICE=display doesn't seems to be doing much (when it doesn't outrigth crashes for me :-D)
Nephele has joined #haiku
<OscarL>
maybe you can render to one of the "bit*" devices? like bitrgb (or bmp/png ones) and output/insert that?
<Nephele>
hi
<zard>
Woot! gdb finally loaded the symbols for WebProcess
<zard>
Hello nephele :). Trying out gdb for MiniBrowser
Oscar-L has joined #haiku
<zard>
By now, with the amount of debug symbols I have, I doubt Debugger could handle it without running out of memory
<zard>
Let's hope gdb works :)
<Nephele>
why would it run out of memory? :(
<Nephele>
zard: I send a question yesterday about a webkit build server, did you see it?
<Oscar-L>
Begasus: 41m15.934s for the "make -j 1" build version :-(
<zard>
500 Mb of debug symbols seems to turn to 12 GB in Debugger
<Anarchos>
Oscar-L: i can't' test gs right now.
<zard>
The one about building WebKit with ssl?
<Nephele>
no
AlienSoldier has joined #haiku
<zard>
No, I doubt it, especially if only zardshard was around
cocobean has joined #haiku
* zardshard
only hears when zardshard is explicitly in the message
<cocobean>
hrev57943 - downloads at 200-400KB/s performance versus 4-6MB/s.
<Oscar-L>
yup, helps a lot with pkgman here :-)
<cocobean>
:D
OscarL has quit [Ping timeout: 480 seconds]
linuxmaster has quit [Ping timeout: 480 seconds]
Oscar-L is now known as OscarL
cocobean has quit []
<OscarL>
Seems a tiny-bit slow to ramp up. But for large enough packages, it does wonders.
<AlienSoldier>
i noticed webpositive initiate openning bookmark with a black background at the tab creation for a few milisecond.
itaipu has joined #haiku
<OscarL>
nephele: https://ibb.co/pZkPHsS <<< I'm getting those "vertical lines" glitches with both Flat and Default decorators. Using the "Be" one for now :-D
<Nephele>
only happens over Terminal?
<OscarL>
only noticed there, till now, yes.
<Nephele>
hmm. maybe it is a terminal glitch, then.
<Nephele>
did you do anythin special to trigger it?
<Nephele>
not getting it here with HaikuControlLook
<OscarL>
I played a bit with Appearence after updating, changing colors and stuff. But nothing crazy, I would think.
<Nephele>
I mean i can get redraw artifacts related to the title bar on the desktop easily, and the part below it. (old bugs i should probably report...) but nothing on terminal
<OscarL>
this VM session might be borked moving around Keymap preflet, and changing front/back with Appearence on top of it is crazy slow :-)
<Nephele>
oh. vm.
<OscarL>
* missing ":" after borked.
<OscarL>
will try rebooting it after I finish a round of tests.
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
bbjimmy has joined #haiku
<bbjimmy>
Web+ crashes on launch state: Exception (Segment violation) Haiku revision: hrev57943 Aug 9 2024 06:02:38 (x86_64)
<OscarL>
bbjimmy: curl / ssl related it seems, right?
Begasus has quit [Quit: Vision[]: i've been blurred!]
bjorkintosh has joined #haiku
<TmTFx>
I think I found a bug or some kind of problem in Haiku with python
<TmTFx>
but I ask to more experienced people
* OscarL
runs the other way.
nipos has joined #haiku
<zard>
:P
<TmTFx>
is there a memory limit with io.BytesIO()?
* zard
checks the docs for clues
<TmTFx>
because if I exceeds more or less 2GB of ram I get MemoryError
<waddlesplash>
why do you need that much memory anyway?
<TmTFx>
in a 64bit enviroment
<TmTFx>
this shouldn't happen
<waddlesplash>
the point of such IO operations is to work on smaller parts of data at once
<TmTFx>
handle file
<OscarL>
was about to ask if that was on 32 bits.
<waddlesplash>
work on the file as a stream, not all at once, then you can work with files of any size :)
<TmTFx>
I'm creating a tar file
<waddlesplash>
yes... so, stream processing :)
<TmTFx>
it reads from disk and crerates an archive in memory with io.BytesIO(), the same happens when I decompress a bz2 file in ram (a io.BytesIO()) and this exceeds 2GB
<PulkoMandy>
Well don't do it in memory, output to a file and let the filesystem cache manage memory for you
<TmTFx>
I did a workaround using /boot/system/var/shared_memory/
<TmTFx>
I was asking if it was normal that behavior
<zard>
It's a bit weird... are you using Haiku-PyAPI?
<TmTFx>
the interface yes... but the functions are outside the interface in a separate thread
<zard>
Ah, BytesIO is a Python function. No wonder I couldn't find documentation on it
<zard>
What function do you use to compress the file?
<zard>
Ok, let's see if I understand. You essentially use "bz2.BZ2File(input_file, 'rb')" to decompress the file?
<TmTFx>
yes extract the tar file from bz2
<zard>
Ok, and then you extract with essentially tarfile.open(fileobj=io.BytesIO(tar_data), mode="r").extractall(path=output_dir)
<TmTFx>
it should but at file.read() I get the memory error so that part is not even touched
<zard>
Ah, ok
<zard>
Now, how would you decompress without reading the whole file at one time...
<TmTFx>
on compression side I get the Memory error on "tar.add(input_path, arcname=relative_path)"
<TmTFx>
File "/boot/system/lib/python3.10/tarfile.py", line 255, in copyfileobj
<TmTFx>
dst.write(buf)
<zard>
Well, I want to focus on decompression for now
xet7 has joined #haiku
<TmTFx>
I'm workarounding extracting the file in /var/shared_memory (RAM_FS) as a simple file (easy peasy)
<TmTFx>
If it is an intentiona limitation for memory usage, it's ok for me, I didn't know... but if it isn't it should be fixed
<zard>
Well, I think it's a problem that should be fixed. Though that you use that much memory is also a problem...
<TmTFx>
well... if the file is big...
<zard>
Yeah, you want to do some kind of incremental decompression
<TmTFx>
I have the option to select which method use... extraction in file and in ram
<zard>
In essence: Unzip, write to disk, unzip some more, write to disk, etc.
<zard>
(This is essentially what waddlesplash meant by stream processing)
<TmTFx>
well in my case I'm not sure I can do it... the best is use the decompression on disk as pulkomandy said... but I wanted to add the in ram option to save SSDs usage
<Skipp_OSX>
... I will try to get to showing non-persistant volumes in Tracker later on the todo list
<Skipp_OSX>
have to hide the packagefs related ones but rest can show but they don't right now
<zard>
Looks like you can do something like `tar_file = TarFile.open(fileobj=bz2.BZ2File(input_file, 'rb'))`
<zard>
Hopefully, that doesn't read the entire file, instead, it just reads what it needs so far
<TmTFx>
It's exactly what's it is doing right now
<Skipp_OSX>
I tried to zip and unzip on 32-bit Haiku got memory errors. Must be hitting some limit but I know 32-bit Winders can still zip large files which means it is still our fault somewhere somehow on 32-bit
<zard>
But you have a `io.BytesIO` in the middle?
<Skipp_OSX>
for example if you try and zip up Haiku source repo folder on 32-bit Haiku will error out before finishing
<Skipp_OSX>
idk what is going on I assume it's hitting the RAM limit or swap limit
<zard>
Actually, what you're currently doing is something like 1. Open bz2file 2. read it 3. create a bytesio so that you can treat it like a file again 4. extract it
<zard>
At least, that's my interpretation
<zard>
I'm suggesting cutting out steps 2 and 3. Open the bz2file. It's already a file object so I think you can just pass it directly into TarFile.open
<zard>
And I suspect it won't have loaded the whole thing in memory. It'll just be ready to extract the parts it needs to when you request it
<waddlesplash>
yes
<zard>
That's it for today. Bye!
zard has quit [Quit: leaving]
<TmTFx>
Mh.... he didn't read the pms T_T
<TmTFx>
anyway it didn't work like he said
<TmTFx>
using this: tar_data = tarfile.open(fileobj=decompress_bz2(input_file), mode="r")
<TmTFx>
used the ram like previous version, and gave me the same error
<TmTFx>
thank you for your suggestions... Hopefully there's shared_memory there which will help me
<OscarL>
KDL again... PANIC: release_spinlock: attempt to release lock 0x[...]801c5248 with interrupts enabled.
<OscarL>
waddlesplash: anything I should try while on KDL to get more info?
<waddlesplash>
what causes it?
<waddlesplash>
if you have an easy reproducer just open a ticket and attach it :)
<waddlesplash>
TmTFx: don't use decompress_bz2, just tell Tarfile to do it for you
<OscarL>
I just hit Ctrl+C to paste some line of text... and boom :-(
<waddlesplash>
TmTFx: mode="r:bz2"
<waddlesplash>
OscarL: screenshot?
<OscarL>
will link one.
<Anarchos>
waddlesplash i made a ticket for the power off of my network card on shutdown. (it should stay alive for wake on lan)
<waddlesplash>
Anarchos: I saw, I don't intend to look at this in the near future
<Anarchos>
waddlesplash no trouble, i flagged it as low . It is not really annoying.
<OscarL>
(hard to do much with CPUs all at 100%, lol :-/)
<Anarchos>
OscarL the infamous weird named Vision's thread taking 100 % ?
<Anarchos>
(right now it is 'exploding cow' here :) )
<OscarL>
Anarchos: no. KDL on a VM.
<waddlesplash>
TmTFx: there isn't really much reason to separate the tasks, doing it that way consumes more resources needlessly
<waddlesplash>
either in memory or in disk space
<waddlesplash>
it makes much more sense to decompress and read the TAR at once
<waddlesplash>
and to let Python do this, much more efficient than doing it in 2 steps
<nekobot>
[haiku/haiku] 3ecbb34240d2 - IORequest: Correct major oversight in finished callback API.
gouchi has quit [Quit: Quitte]
<mmu_man>
oh the joys of recursive debugging…
<mmu_man>
../src/system/kernel/cache/block_cache.cpp:1308:34: error: 'block' was not declared in this scope; did you mean 'flock'?
<mmu_man>
am I using a too old GCC or something…
<mmu_man>
seems I have BLOCK_CACHE_BLOCK_TRACING enabled
<mmu_man>
oh dear
<mmu_man>
../src/system/kernel/debug/tracing.cpp:599:41: error: format '%lu' expects argument of type 'long unsigned int', but argument 3 has type 'int' [-Werror=format=]
nipos has left #haiku [Disconnected: Replaced by new connection]
nipos has joined #haiku
<mmu_man>
did we change the definition of uint32 recently?
<waddlesplash>
no
<waddlesplash>
but this usually happens on x64 vs x32
<waddlesplash>
use B_PRI* macros
<waddlesplash>
B_PRIu32 etc.
<waddlesplash>
mmu_man: what are you trying to debug anyway lol?
<mmu_man>
media_addon_server but I need to make the haiku pkg
<mmu_man>
well the code does use it
<mmu_man>
I guess I should reconfigure for x86
<waddlesplash>
no?
<waddlesplash>
you can just build individual addons and throw them in nonpackaged
<waddlesplash>
that's what I did
<waddlesplash>
mmu_man: but did you retest if your crash is fixed after hrev57930?
<waddlesplash>
the multi_audio change in that hrev may help you
<mmu_man>
hrev47943 and still crashing
<mmu_man>
er, debugger()
<waddlesplash>
which message?
<mmu_man>
same, performance time too large
<waddlesplash>
is it in the same thread?
<waddlesplash>
well anyway you certainly don't need to rebuild everything
<waddlesplash>
just the two addons and libmedia.so suffices
<mmu_man>
of course, debug bins do work :D
<mmu_man>
maybe it's a race
xet7 has quit [Remote host closed the connection]
Nephele has quit [Quit: Vision[]: i've been blurred!]
<waddlesplash>
yes, it probably is
<waddlesplash>
usually this happens when there is an uninitialized time
<waddlesplash>
current_time=(some large value), previous=0, and difference > 24bits large
<waddlesplash>
so if it's just a bit slower on another thread you get a different time that's nonzero and it works
<mmu_man>
although they don't work when I restart media_server
<mmu_man>
like before
<mmu_man>
well, they start, seem to work, but no audio
<mmu_man>
that hasn't changed.
<mmu_man>
but it could be another bug
<waddlesplash>
probably a similar bug
<waddlesplash>
mixer doesn't behave right if its times are wrong
<waddlesplash>
try disconnecting and reconnecting the mixer, see if it starts working
<waddlesplash>
(with Cortex)
<mmu_man>
haven't tried changing mixer addon yet
<mmu_man>
now I need to wait for gcc to build
<mmu_man>
can't seem to configure from Haiku anymore though
<waddlesplash>
why not?
nipos has left #haiku [#haiku]
nipos has joined #haiku
nyochio has joined #haiku
nyochio has quit [Remote host closed the connection]