<SArpnt[m]>
i want to change my input settings (keyboard repeat, mouse sensitivity) to exact values to match my other computers, but the Input application just has sliders with no shown numbers
<SArpnt[m]>
is there a file i can edit somewhere for this
<SArpnt[m]>
* is there a file i can edit somewhere for this
deneel has quit [Remote host closed the connection]
<augiedoggie_>
SArpnt[m]: looks like the format of the settings file isn't the easiest to edit, you can cheat a bit and send messages to the Input app while it's open though
<SArpnt[m]>
how do i send messages to it to set what i want
<augiedoggie_>
while the Input app is running, you can run a command like: hey Input let View 0 of View 1 of Window 0 do 'SLrr' with "be:value"='int32(200)'
<augiedoggie_>
the slider won't update but the buttons will, you can close the app and reopen and it should show the new value
<augiedoggie_>
you have to be viewing the keyboard info in the app when you run the command
<SArpnt[m]>
not familiar with the `hey` command at all
<SArpnt[m]>
how do i know what views or windows or other settings to use
<augiedoggie_>
sometimes you can refer to views by name, in this case i had to "scan" the hierarchy manually
<augiedoggie_>
hey is one of the most powerful tools on BeOS/Haiku
<augiedoggie_>
that command could be changed to say "View InputKeyboard" instead of the first "View 0"
<SArpnt[m]>
the command you sent doesn't seem to do anything
<SArpnt[m]>
key repeat feels the same (not correct in delay or rate)
<SArpnt[m]>
i'm not really sure how you're scanning anything
<augiedoggie_>
you have the Input app open and viewing the keyboard info?
<SArpnt[m]>
yes
<augiedoggie_>
what does this say? hey Input get InternalName of View 0 of View 1 of Window 0
<SArpnt[m]>
result InputKeyboard error 0
<SArpnt[m]>
* result InputKeyboard error 0
<augiedoggie_>
ok, so it's finding the right view, the command should work but as i said, the sliders won't show a change unless you reopen the app
<augiedoggie_>
i can feel an immediate change while typing in Terminal right after changing it
<augiedoggie_>
right after sending the hey command*
<SArpnt[m]>
still not working for me
<augiedoggie_>
dunno then
<SArpnt[m]>
`hey Input let View 0 of View 1 of Window 0 do 'SLrr' with "be:value`='int32(200)'`
<SArpnt[m]>
i do have the command correct right
<SArpnt[m]>
s/`=/"=/
<SArpnt[m]>
s/`=/"=/
<augiedoggie_>
looks ok
<SArpnt[m]>
is there supposed to be a rely
<SArpnt[m]>
s/rely/reply/
<SArpnt[m]>
i get BMessage B_NO_REPLY
<augiedoggie_>
yeah, that's fine
<augiedoggie_>
hey is generally used to access the "scripting api" but in this case we're just sending a raw message so there isn't a reply
AlaskanEmily has quit [Remote host closed the connection]
janking has joined #haiku
mmu_man has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
deneel has joined #haiku
nekobot has quit [Quit: Exiting on terminated]
nekobot has joined #haiku
arjen_ has joined #haiku
arjen__ has joined #haiku
jmairboeck has joined #haiku
Samanth0r has joined #haiku
Samanth0r has quit []
HaikuUser has joined #haiku
diver has quit [Read error: Connection reset by peer]
HaikuUser has quit []
mittwerk has joined #haiku
diver has joined #haiku
mittwerk has quit []
BrunoSpr has joined #haiku
diver has quit [Read error: Connection reset by peer]
diver has joined #haiku
BrunoSpr has quit []
dovsienko has quit [Quit: Leaving]
BrunoSpr has joined #haiku
<BrunoSpr>
Begasus[m]... mesa_swrast could not find it! Mesaswpipe I tried to install. But no!
diver has quit [Read error: Connection reset by peer]
TMM has joined #haiku
diver has joined #haiku
diver has quit [Read error: Connection reset by peer]
BrunoSpr has quit []
diver has joined #haiku
BrunoSpr has joined #haiku
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
diver has quit [Read error: Connection reset by peer]
diver has joined #haiku
BrunoSpr has joined #haiku
diver has quit [Read error: Connection reset by peer]
BrunoSpr has quit []
jmairboeck has quit [Ping timeout: 480 seconds]
OscarL has joined #haiku
BrunoSpr has joined #haiku
diver has joined #haiku
diver has quit [Read error: Connection reset by peer]
<OscarL>
Used nightly .iso to install on an USB pendrive... install took just a few seconds. Now reboot is slow AF, as it actually has to write the data to the slow USB drive, and you get no indication of the progress :-(
<OscarL>
Guess there's a write-cache that is just too large for this use case? (I remember having to tweak Linux to avoid this bad behaviour).
diver has joined #haiku
<OscarL>
I also wonder how many boot failures we get from people facing a similar situation, and just unpluging the USB once the Installer says it has finished.
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
<Monni>
OscarL: A lot I guess... It's a lot easier to install operating system to a pendrive than to install a new internal drive or repartition the existing internal drive...
diver has quit [Read error: Connection reset by peer]
diver has joined #haiku
gouchi has joined #haiku
<OscarL>
right.
diver has quit [Read error: Connection reset by peer]
<OscarL>
As Installer just uses regular file copy operations (AFAIK), I guess solving this issue won't be quite as easy as saying "don't use write-cache for this".
<OscarL>
(unless there's a separate write-cache for removable media that could be tuned?)
diver has joined #haiku
<Monni>
OscarL: Write-cache is pretty much mandatory with current solid-state drives as writing is insanely slow compared to old spinning hard disks... People only get close the theoretical maximum write speed on brand new SSD or pendrive...
<OscarL>
yes, but having users waiting for 10/15 minutes till data is properly written to a slow media, without any indication of what's going on, it is hardly a good situation.
<OscarL>
Windows handles this far more gracefully, for example.
<OscarL>
You see fast writes at first, but it quickly adjusts progress speed of the copy operations for USB pendrives.
<OscarL>
On Linux I remember managing to get something similar by tuning some "dirty_ratio / dirty_size" values somewhere.
<Monni>
OscarL: 10/15 minutes is still pretty fast... The best I measured myself was 58 minutes... There was a story about EA having to slow down their game updates, because those use so much disk space and write so much that the drives overheat and cause premature failure...
diver has quit [Read error: Connection reset by peer]
<OscarL>
without tha affecting the speed of HDD/SSD.
* phschafft
blinks.
diver has joined #haiku
<OscarL>
Monni: "fast" in this case because it was just Installer from nightly image (563 MB of actual data on the 650 MB .iso).
<Monni>
OscarL: I've never seen modern drives use DMA transfers (which used to be fast), basically when my PC writes to SSD, CPU usage spikes...
diver has quit [Read error: Connection reset by peer]
<phschafft>
do I miss something here?
<phschafft>
I mean, storage became faster all the time over the last like 40 years.
<phschafft>
not counting a few software bugs here and there.
<OscarL>
last time I had an HDD not doing DMA was in a poorly setup Win95,
<Monni>
phschafft: Reading speed got faster, but not constant writing speed...
<phschafft>
also writing speed. if you have a problem with write speed you have a broken driver or something.
<OscarL>
(Or in BeOS before the ide_replacement drivers, I guess :-D)
<phschafft>
naturally write speed is slower on some types of storage than read speed. but it is still getting faster and faster.
<OscarL>
phschafft: nah... this are cheap pendrives that can't do more than 4 MB/s in a good day.
diver has joined #haiku
<OscarL>
read speed is 12/15 MB/s.
<phschafft>
OscarL: yes. but for the same money you got one that was like 1MB/s 5 years ago and one that was like 0.5MB/s 10 years ago. ;)
<Monni>
My latest pendrive is 64 GB USB 3.1 ... However not all pendrives are USB 3.1 yet...
<OscarL>
I see what you mean, but the timescale is off. This 4 GB pendrive is ancient :-P
diver has quit [Read error: Connection reset by peer]
<phschafft>
OscarL: basically your point is about writing dirty pages. going for /direct/ write (disable the cache) is what spikes the CPU.
<phschafft>
so using relativly speaking a good bit of RAM for dirty pages does make it faster overall.
<phschafft>
but specifically in a unmount situation it is clearly a good idea to tell the user when it's safe to unplug.
<phschafft>
OscarL: then shift it around a few years to your taste. my personal timescale is clearly off as I'm a bit more used to pro hardware when it comes to storage.
<Monni>
My old pendrive had LED that did tell when it was safe to unplug... Haven't checked the new pendrive as my current PC is under the table...
<OscarL>
I still think it is better to have a slower copy, than leave it to the moment of unmount.
diver has joined #haiku
<phschafft>
OscarL: so basiclly what you ask for is like a write prograss bar as my old cam had: it showed on the side how full the cache is.
mmu_man has joined #haiku
<phschafft>
OscarL: my point is not slower vs. faster. but buffered vs. direct.
<OscarL>
I understand the difference between buffered and direct, but how that helps the use case I mentioned at the beginning. it is just a bad user experience.
<phschafft>
also the question here is if you want, in an installer, show the file writes or the block writes.
<OscarL>
In this case, I mentioned installer, but same thing would apply if copying a large file to any slow media.
<phschafft>
yes. but if you go for e.g. random access or read-then-write patterns it might be the other way around.
<OscarL>
I'm saying... don't read all the big file into memory, tell me that it the copy is done... only to then leave me waiting for who knows how long before data gets actually synced.
<phschafft>
it's not easy to do this right for all cases.
<OscarL>
Again, Windows seems to handle this faaar better.
<OscarL>
and I can at least tweak it on Linux.
_-Caleb-_ has joined #haiku
<phschafft>
what is the problem with the installer just asking the kernel if it's done with the file?
<OscarL>
IIRC, in linux you get by default a percentage of RAM for those "dirty pages". In my experience, using a small fixed size (48/64 MB, IIRC), gave a far better experience that it taking GBs.
<phschafft>
like you copy your files but you only show that a file is complete if the kernel tells you so?
<phschafft>
OscarL: and for all my usecases going for some GB is *way* more performant and gives a better experience.
<phschafft>
the point is, only the application can know what it needs. the kernel must guess.
<OscarL>
your usecases in a Desktop oriented OS?
<phschafft>
so if the application is telling the kernel and showing the result to the user you can have the best of all worlds.
<OscarL>
(Guess that adding some sync() calls in the Installer copy loop might help that case at least?)
<phschafft>
e.g. you write files in one thread (which is hopefully not your GUI thread ;).
<phschafft>
once a file is done, transfer that to a second thread.
<phschafft>
that second thread then calls fsync() on the file.
<phschafft>
once fsync compeltes it shows the user that this file is done.
nephele has joined #haiku
<phschafft>
that way you can use a very large buffer, still have per-file accurate GUI display.
<Monni>
Linux likes doing file I/O in GUI thread, Windows doesn't... gotta yield once in a while or the GUI gets seriously unresponsive ;)
<nephele>
never yield, never surrender
<phschafft>
nephele++
<Monni>
I've seen a few PC catch on fire, when they thought yielding was bad...
<nephele>
sounds like a hardware design problem...
<nephele>
temperature sensor + emergency power off is not that expensive to build
<OscarL>
phschafft: now how to fix the case of copying a large file (say 4 GB movie) to a pendrive that does at best 14 MB/s (with slows downs into the 1-4 MB/s range) ?
<Monni>
Normal BIOS should shutdown if CPU temperature goes critical, but at least Windows can override that...
<nephele>
defective BIOSD
<phschafft>
OscarL: if you really, really want to go with slow copy, call fdatasync() every few MB.
<nephele>
OscarL: have a filesystem that can checksum blocks :D then we can resume copying later if interupted
<phschafft>
OscarL: other than that, use some API as provided by your system for that. it depends a bit on the system.
<nephele>
phschafft: i was thinking tuesday 16:00
<Monni>
My dad had 127 degrees Centigrade/Celsius as highest displayed temperature on his PC... I haven't seen anything higher than 72...
<OscarL>
phschafft: would be nice if Tracker used that then when dealing with slow removable media :-)
<phschafft>
127... think about that number for a second....
<phschafft>
;)
<phschafft>
OscarL: it's also a matter on large file and slow media.
<phschafft>
what are the limits for that? how do you know?
<Monni>
Well above boiling temperature of water, which is about 96~100 degreed Centigrade/Celsius...
<OscarL>
Use case I mention is pretty normal stuff... write a movie to a pendrive, unplug, give drive to friend, whatever.
<phschafft>
he doesn't get it? does he?
<nephele>
phschafft means that 127 is exactly 2^7 -1
<nephele>
which is very suspicious if a pc turns this up in an unexpected case
<OscarL>
having people unplug the driver because the copy dialog say it was done doesn't sounds so much as user error as bad UX.
<nephele>
yes, sounds like system fault to me OscarL
<OscarL>
and if you know what's going on... having to wait without any indication of progrees is also terrible.
<Monni>
nephele: I know it was maxed out when the PC was sending smoke signals instead of ICMP packets...
<phschafft>
OscarL: my point is 'what counts as slow' and 'how do we messure that'. plus 'what counts as big' (I hope that messuring the file size is not that much of a problem ;)
<nephele>
If you "know" you can run sync on the CLI explicitly
<OscarL>
(when you unmount/sync/eject)
<OscarL>
phschafft: no idea... but clearly there are better ways that current behavior.
<nephele>
Monni: It is doubtfull that a temperature sensor designed for PC's can even properly measure 127°C
<OscarL>
again even crappy windows handles this better.
<nephele>
OscarL: you are correct, we definetely should do better
<phschafft>
OscarL: I'm not saying the current one is good. I'm just saying be careful to make a change that improves your situation but not at the cost of other situations.
<nephele>
but also beta label :D
<phschafft>
and I also made a few suggestions to explore.
<Monni>
nephele: Haven't checked what is the critical temperature of an IDE cable.... But it melted just fine...
<nephele>
depends on the cable
<phschafft>
like having an indication for the diry page count (which might be fun just form a visualisation perspective alone!). or calling f(data)sync().
<nephele>
phschafft: 16:00 tuesday for the next stream i ment. Would that work for you? Not sure if i have to be "Let in" for such a stream
<OscarL>
I'll be back soon (testing latest nightly on a cursed netbook).
* nephele
curses OscarL's other laptop aswell
<Monni>
Still on hrev57937 here...
<nephele>
hmm, i could update
<Monni>
I'm too lazy to RTFM how to create bare ISO from source tree...
<nephele>
what is a "bare" ISO?
<nephele>
doesn't the combined image work for you? @nightly-anyboot ?
<Monni>
That's still building... have to copy it from Linux to Haiku when it finishes...
<nephele>
If you just want to update I am wondering why you would build a full iso though
<Monni>
nephele: Because I get KDL if I try to download anything large...
nephele has quit [Quit: Vision[]: i've been blurred!]
nephele has joined #haiku
<nephele>
How will an iso help there?
<phschafft>
OscarL: one extra thing that any copying software should do (if possible) is to tell the kernel ahead about the size of the file. that will improve speed on all types of media, plus reduce wear.
<Monni>
nephele: I mount it and copy just the hpkg files...
<phschafft>
e.g. posix_fallocate().
<nephele>
... so just build the packages and put them on a thumb drive, and skip the iso
<phschafft>
as for mmap based transfers there is a lot that can be done with the memory based API as well.
<Monni>
nephele: That's the "bare" part...
<nephele>
That's not an iso then
<nephele>
You can just do it like this
<nephele>
jam -j4 @nightly-anyboot haiku.hpkg haiku_datatranslators.hpkg haiku_devel.hpkg
<Monni>
nephele: But mounting a ISO image is easier than installing USB controller on live VM...
<phschafft>
nephele: hm. we can try.
<nephele>
phschafft: I'd write on the forum then about that. Maybe keep the link hidden till the stream or so
* phschafft
nods.
MonniTheKitten has quit [Quit: Vision[]: i've been blurred!]
<Begasus[m]>
'lo peeps
<Monni>
Nightly throws instantly to KDL... !page->IsMapped() ... lol..
<Begasus[m]>
still the same here ... no valid boot volume
<OscarL>
nephele: at least the LM35DZ sensor that can be found on PSU and on some (old) motherboards is "Rated for full −55˚ to +150˚C range".
<OscarL>
3 legged T0-92 little thing, makes for a nice sensor when toying with arduino/esp8266.
<OscarL>
phschafft: your suggestions sounds good to me... too bad I'm too dumb to actually implement any of that :-)
<OscarL>
anywah.... I CREATED A MONSTER!
<OscarL>
both pendrive *AND* HDD are initialized as MBR... (even with BootManager installed), but has a little ESP partition at the end.
<OscarL>
The cursed netbook seems (now) happy to boot in either BIOS or EFI mode. LOL.
MonniTheKitten has joined #haiku
<OscarL>
pendrive has both 64 and 32 bits Haiku. Used it to install 64 bits version (monitoring that one via SSH + htop now)
gouchi has quit [Remote host closed the connection]
<OscarL>
but... if I boot in EFI mode from the pendrive, the Haiku boot loader only sees the partition on the HDD, not the one from the pendrive.
<Begasus>
Hi OscarL :)
<OscarL>
so to actually boot Haiku from pendrive... I need to use BIOS/Legacy. From HDD, I can now use either BootManager, or rEFInd, all from a non-GPT disk. cool!
<Begasus>
cool if you're able to :)
<OscarL>
Begasus: yeah... that part of the pendrive only seeing the partiion on HDD though... reminded me of your case.
<Begasus>
yeah, that should be the first it should look at
<OscarL>
but you are not using such a crazy pendrive like I so... no idea.
<nephele>
gonna play some video games now (gal civ III), laters :)
nephele has quit [Quit: Vision[]: i've been blurred!]
<Begasus>
oh, have don't several attempts with variable setups on the pendrives OscarL :)
<OscarL>
I can imagine, yes :-)
<OscarL>
what I mean... plain dd'ed anyboot image shouldn't have so much issues finding the volume on USB... but here we are :-(
<OscarL>
BIOS can be crazy...
<Begasus>
let's blame it all on M$ :P
<OscarL>
the one from the cursed netbook, allows you to disable EFI *and* legacy boot at the same time. Then of course it complains it cannot boot :-D
<Begasus>
heh
<OscarL>
didn't used the cursed netbook since august... somehow the date was set to sometime in 2075 ?!?
<Begasus>
high enerrgy battery?? ;)
<OscarL>
I've seen RTCs getting wrong dates, but never in the future :-)
arjen__ has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
Begasus: heh, doubt it... only had a "slightly" used CR2032 to put in there to replace the dead one it had :-)
<OscarL>
would be nice if "launch_roster stop" actually stopped the service/job you give it, instead of just disabling it.
<Begasus>
I don't even have a clue what's going on :)
<OscarL>
I just want to disable media_server/midi_server/power_daemon, as I have no use for them on this netbook (and not running media_addon_server might actually be the key to have this thing be stable :-D)
arjen__ has joined #haiku
<Anarchos>
OscarL you can black list them ?
<OscarL>
Anarchos: not for now, I'm just testing stuff at the moment.
<OscarL>
this machine has given me a lot of problems in the past... then it worked almost perfectly for one day... then stopped working :-/.
<OscarL>
giving it another go now with hrev58228.
mittwerk has joined #haiku
<OscarL>
`time python3.13 -c ""` >>> real 0.233s. Not that bad considering this is an Atom CPU with HDD.
mittwerk has quit []
<Monni>
hrev58228 didn't even try to boot here... gonna go back few revisions...
<OscarL>
python3.10 is consistently 100ms slower on startup. Maybe I should "back-port" my changes from 3.12/3.13.
diver has quit [Ping timeout: 480 seconds]
<OscarL>
listdev could use a "compact" output mode, more like "lspci -n"
MonniTheKitten has quit [Quit: Vision[]: i've been blurred!]
diver has joined #haiku
usrmittwerk[m]1 has joined #haiku
vezhlys has quit [Remote host closed the connection]
<OscarL>
jmairboeck: I guess "--bootstrap-system-{libuv,jsoncpp,librhash}" means avoid building the embedded copies of those during bootstrap, right?
<jmairboeck>
yes, that's just a small optimisation to make the bootstrap faster
<OscarL>
cool, I wasn't really sure about those, and I wansn't really looking forward to make more repeated builds :-D
<jmairboeck>
we need the libraries anyway for the main build, and the pkg-config and which dependencies shouldn't be a problem
<OscarL>
thanks for working on it :-)
<OscarL>
jmairboeck: did "== $portVersion base" worked for you? I was getting some weird error with that (thus why I ended up with >= 3)
<Begasus>
hi jmairboeck
<Begasus>
haven't checked there yet :)
<jmairboeck>
I successfully installed cmake_gui
<jmairboeck>
before the upgrade to .5 though
<OscarL>
isn't "base" used for packages on the same recipe? (or maybe thats where I was getting the error? can't recall, really).
<waddlesplash>
Monni: in addition to the syslog from the failed boot on that earlier hrev, if you're making custom builds, can you enable the TRACEs (just make them dprintfs) in the method that's in the call stack on the most recent failure, the free_unused_bootloader ranges method?
<waddlesplash>
and then get the syslog from that failure with the extra dprintfs
<jmairboeck>
it is to denote the "base" package, but for the hpkg it shouldn't make a difference if they come from separate recipes or not
MonniTheKitten has quit [Quit: Vision[]: i've been blurred!]
<Begasus[m]>
no build checked, but I guess it this case it makes sense (had something simular here but without bootstrap) to require base
<OscarL>
jmairboeck: ok. still reads a bit weird to me, but... oh well :-).
<Begasus[m]>
jmairboeck did you check if the gui could work with another cmake "version"?
<jmairboeck>
I thought it would be best to keep the versions in sync
erysdren has quit [Quit: Konversation terminated!]
<OscarL>
I might have tried to install via pkgman install -H, and that might have caused my problems, I guess?
<jmairboeck>
yes, it seems that is exactly what "base" disallows then
<jmairboeck>
mixing system and user installations
<OscarL>
given that cmake_gui is pretty standalone...
<jmairboeck>
ok, I can drop the base part
<OscarL>
and only needs cmakes data... I might be tempted to keep it >= 3, but I won't fight it :-)
<OscarL>
not really planing to touching cmake if I can avoid it :-P
<jmairboeck>
I'm not sure if it works with a different version, I think I'll keep "== $portVersion"
<jmairboeck>
for _devel packages the "base" definitely makes sense, as the lib*.so in develop/lib is just a symlink, so that would break otherwise
<OscarL>
Right. Glad to have you around so I can better understand how base actually works :-)
<jmairboeck>
I didn't know exactly either. I knew it was documented somewhere in the devel docs, but I had to look it up too :)
<Begasus>
+1 from here too
<Begasus>
lazy sunday here, so I'll be sure to check it out later/tomorrow if it hasn't been merged yet :)
<Begasus>
checking something from KDE that isn't in the depot :)
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
bbjimmy has joined #haiku
diver has joined #haiku
<OscarL>
2 and a half hour of uptime on the "cursed netbook". /me keeps his fingers crossed.
<OscarL>
(no wonder my typing sucks)
* Halian
sends OscarL his energy
<OscarL>
thanks Halian :-)
<Halian>
Your'e welcome :3
<Halian>
*You're
<jmairboeck>
OscarL: I updated the test report (didn't measure the runtime though and only tested 64 bit)
<OscarL>
+1
<jmairboeck>
most of the differences are the test numbers it seems. And I had fewer tests (because of the missing GUI?)
<Begasus>
waddlesplash, jmairboeck, checked, this works: pkgman uninstall *_devel-6.7.0 :)
<OscarL>
jmairboeck: yeah, most likely missing tests specific to the gui.
<jmairboeck>
but I don't think I can add tests to the cmake_gui recipe, because that now builds only the GUI and its dependencies
<Begasus>
and this works for installing :) pkgman install *_devel-6.7.0* -y
* OscarL
runs "python3 -m ensurepip --user --default-pip". Worked pretty well (other that pip "complaining" that there's a new version available :-D).
<jmairboeck>
my colleague at work always wondered why would one use add_subdirectory in a cmake build and use multiple CMakeLists.txt files (one per subdirectory), but in this case it is helpful :)
<Begasus>
they probably didn't do much cmake builds out of the office? ;)
<jmairboeck>
I tried also running cmake in that subdirectory but that fails because it is missing some dependencies then and would fail to link. But using the top-level CMakeLists.txt and then invoking just make in a subdirectory works :)
<Begasus>
nice catch!
<jmairboeck>
I'm using -Cbuild-gui/Source/QtDialog for the make calls, so it only builds and installs that
<OscarL>
silly pip upgrade decided to install stuff in system instead of upgrading in the original location (--user) :-/
<OscarL>
forcing users to re-iterate --user for upgrade command is kinda bad UX. /me shakes fist at Python developers
<OscarL>
let's see how broken "pyperfomance run --fast" is :-)
<OscarL>
mmm, it is downloading its own copy of "psutil", so I doubt we will get too far :-(
<jmairboeck>
pip's UX isn't the best anyway. One problem I once had that I needed to use --force-reinstall when upgrading from a git repo where I changed something but the version number didn't. Otherwise it would re-download the changes, notice that "nothing changed" (sic!) and do nothing, without any error.
<OscarL>
:-)
<Begasus>
nice, only 3 pkgman calls in the script to clean and install the full frameworks now :) thanks guys!
<OscarL>
( this thing seems stuck in some install/error loop :-( )
<OscarL>
or maybe not... output is quite confusing :-D
<Begasus>
it's doing a OscarL thingy? ;)
<OscarL>
"platform haiku1 is not supported" there we go. psutil is busted.
<OscarL>
it is trying, but it only keeps failing and failing. Won't work unless upstream pstutil gets Haiku support.
<Begasus>
who broke that! :P
<OscarL>
seems this "pyperformance" benchmark suite makes heavy use of virtual envs, so even if we have patched psutil installed... it is of no use here :-(
<OscarL>
oh well. was worth a try.
<Begasus>
guess a case of "can't win them all"?
HaikuUser has joined #haiku
HaikuUser has quit []
flag_ has joined #haiku
vdamewood has joined #haiku
<jmairboeck>
OscarL: without the "base" I get a policy error when packaging
<OscarL>
I think I got a policy error to, thus why the ">= 3".
<OscarL>
s/to/too/
<jmairboeck>
ok, I get it with "base" too, I'll revert that
<jmairboeck>
maybe it works without an explicit version, i.e. the implicit one added by haikuporter
<OscarL>
I recall looking somewhere for which version cmakes were compatible one with each other... but can recall where I looked :-(
flag has quit [Ping timeout: 480 seconds]
<jmairboeck>
that's weird: the automatic one added by haikuporter is >= 3.30.3 (i.e. the installed one) despite having 3.30.5 active in the chroot
<OscarL>
that recipe certainly didn't help my failing mental health :-D
<Monni>
Hehe... Managed to make Putty talk with Haiku kernel...
<jmairboeck>
I think I'll remove the version requirement altogether from REQUIRES and let's see what the buildmasters say.
<OscarL>
jmairboeck: +1. not like we have some < 3 cmake around anyway :-)
<jmairboeck>
normally it should work when only having a requirement on BUILD_REQUIRES and using the automatic one for REQUIRES
<jmairboeck>
for library deps, that is
<jmairboeck>
so I'm not really sure in this case
<OscarL>
mmm now that I think more about it... for me using "$portVersion" failed, ">= 3.30" failed, and thus why I ended up with ">= 3" (didn't even tried with no version restriction)
<OscarL>
lost track of errors between recipe changes and builds on both 32 and 64 bits. I should stick to easier stuff :-)
<jmairboeck>
I'll merge it now if there are no more objections
<OscarL>
that last screenshot from Ilovehotdog is kinda crazy :-)
<Begasus[m]>
whoot!
<Begasus[m]>
not something to start working in though :D
<OscarL>
at first I thought... must be using a mid-to-large swapfile... but seems his install is only 825 MB in total.
<OscarL>
a nightly is around 560 MB of actual packages, so... not *that* much swap :-)
<OscarL>
Begasus[m]: what? don't dare to build rust on that thing? :-P
<Begasus[m]>
not even zlib OscarL :P
<OscarL>
some recipes are more spicy than expected :-)
<OscarL>
(too bad I failed to mark it as draft before it was too late)
<Begasus[m]>
rebuilding qtwebengine ... after 2 years ....
<Begasus[m]>
it's done, have to deal with it and move on, maybe one day ...
<Begasus[m]>
food bbl
AlienSoldier has joined #haiku
<AlienSoldier>
waddlesplash: i just updated and the kernel icon dont light up and it stay like that. I cant seem to be able to use previous state. Any way to get out of that?
<waddlesplash>
why can't you use previous state?
bbjimmy_64 has quit [Quit: Vision[]: i've been blurred!]
<waddlesplash>
there's no way out of this besides using a prior state
<AlienSoldier>
i was in a kernel before 209, and i think it was 228 when i updated.
<waddlesplash>
are you on 32bit?
<AlienSoldier>
waddlesplash: i can, but my previous one seem to do the same.
gouchi has quit [Remote host closed the connection]
<AlienSoldier>
yes 32 bit
<waddlesplash>
hmm. it is possible I somehow broke the bootloader
<waddlesplash>
ok, it could be failed PAE init too
<waddlesplash>
any chance you can collect a serial log?
<waddlesplash>
do any safe mode options help? e.g. disable SMP
<nephele>
unless AlienSoldier manually updated the bootloader that seems unlikely
<AlienSoldier>
the bootloader work for my other haiku install on the same HD
<waddlesplash>
or: if you're on 32bit does the "previous syslog" option show anything?
<nephele>
more likely that the last state was just another software install/deinstall, and than an even earlier state then works
<waddlesplash>
nephele: no. the 32bit loader is updated by pkgman.
<waddlesplash>
it's only EFI that isn't
<waddlesplash>
so if the bootloader is broken then prior states won't help
<nephele>
The one in the partition? isn't that versioned just the same?
<waddlesplash>
AlienSoldier: I'd check the previous syslog option in bootloader and see if we have any errors in there, after a warm reset from the failed boot
<waddlesplash>
nephele: ?
bbjimmy_64 has joined #haiku
<waddlesplash>
it's versioned sure but we can't load it from an earlier state
<waddlesplash>
the bootloader is the thing displaying the states menu in the first place
<nephele>
well the one in the mbr doesn't get updated, i'd had assumed it was able to load previous ones from the packages
<nephele>
oh well, guess that is one thing we can improve upon with efi boot
<waddlesplash>
the MBR loader is a tiny piece of code to load the real bootloader from the package
<waddlesplash>
that's it
<waddlesplash>
it is pretty dumb
<waddlesplash>
AlienSoldier: interesting that your other install works. Well if it does then you can try downgrading the bootloader package that way by replacing it, and see if it helps
<AlienSoldier>
changing the one in my main partition by the one on this safety partition?
<AlienSoldier>
waddlesplash: how is it named?
<waddlesplash>
haiku_loader-...hpkg
<waddlesplash>
AlienSoldier: but please try and get that log first
<waddlesplash>
it will help a lot in trying to diagnose this
<waddlesplash>
please also try Disable SMP and Disable memory beyond 4G
<waddlesplash>
and see if either of those make things better
Nasina has joined #haiku
Anarchos has quit [Quit: Vision[]: i've been blurred!]
byte has joined #haiku
<byte>
does haiku do blinkin lights?
byte is now known as Guest6226
<Guest6226>
i wish to put something together with some led's and one of my arduino's
<phschafft>
if you do, post pictures. :)
<Guest6226>
does haiku have any mechanisms to trigger blinkin lights?
n_crm has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<scantysnax>
Guest6226, i made a blinkenlights back in the BeOS days. It only did two cpus, and needed a parallel port. iirc i used and avr mcu and a bunch of 75HC595s
<scantysnax>
(and leds, goes without saying)
AlaskanEmily has joined #haiku
<phschafft>
75HC595 seems to be a fun part.
<scantysnax>
yep, you can do a lot with them.
<phschafft>
(I normally have specific requirements on ports OR I have microntroller ports left unused because I need to use a bigger model anyway. so hardly having that problem. still sounds like a useful addition to your 'toolbox')
<scantysnax>
indeed.
<phschafft>
maybe add some to my next order :)
<scantysnax>
they are pretty cheap.
<phschafft>
they also sounds like they could be of good educational value when you show people stuff.
<scantysnax>
and really simple to use.
<phschafft>
where exactly that is important: the parts being cheap ;)
<scantysnax>
i think i did 16 leds per cpu.
<scantysnax>
that would make each cpu need 2 '595s
* phschafft
nods.
<scantysnax>
wish i still had the code.
<scantysnax>
it's lost on a hard drive that i probably have somewhere in storage.
<scantysnax>
probably better to start from scratch at this point.
<phschafft>
I'm a bit of interested in doing similar for my own computer build here. currently I only have a single LED that indicates if the CPU is in run or halt state.
<phschafft>
(it's red/green, so it 'fades' depending on load :)
<scantysnax>
ah interesting
<phschafft>
but I think Guest6226 also wanted to know if Haiku specifically has some way of hooking to the OS state to send data to such a hardware, correct?
<scantysnax>
yes, that is what i presumed as well.
<Skipp_OSX>
Be really missed the boat on multi-processor Intel machines, if they had started a few years earlier they could have taken off when the ABIT BP-6 came out.
<phschafft>
scantysnax: one day I had the idea that in the halt function of the kernel I could just switch that GPIO on and off. and I must say it was like one of the best additions (also given it was only two lines of code ;)
<scantysnax>
nice :)
<phschafft>
but I was thinking about adding a bit more later on. even had a few discussions at work about what is part of the machine and what is part of the terminal (including user session).
<phschafft>
so, happy to learn about other people's projects.
<phschafft>
Guest6226: so, you want to mainly display the CPU load?
<Skipp_OSX>
If Be had Adobe software in 1999 things could have been different, unfortunately Windows NT was good enough, kind of a perennial problem.
arjen__ has quit [Quit: Vision[]: i've been blurred!]
<Skipp_OSX>
There was a missed opportunity to become the multi-processor OS there when the iron was hot, maybe there was. Hindsight is 20/20 as they say.
flag has joined #haiku
<AlienSoldier>
waddlesplash: my hrev was 201, it say in the syslog welcome to the haiku bootloader of that hrev in the syslog, but it never say it for the 228 one. The only reference to 228 is close to the end of the syslog and is related to the upgraded packages.
<scantysnax>
sadly, the bp6, and other boards of that era sufferred the "capacitor electrolyte problem"
janking has quit [Quit: Vision[]: i've been blurred!]
flag_ has quit [Ping timeout: 480 seconds]
<AlienSoldier>
waddlesplash: in other word, i think nothin is wrote in the syslog at all, so i guess i should now atempt te haiku loader transplant?
HaikuUser has joined #haiku
<waddlesplash>
AlienSoldier: did you try Disable memory beyond 4G and Disable SMP?
<AlienSoldier>
not yet
<AlienSoldier>
will try that first
<waddlesplash>
ok, please do
<AlienSoldier>
ok, going to bios
arjen_ has quit [Ping timeout: 480 seconds]
AlienSoldier has quit [Remote host closed the connection]
diver1 has joined #haiku
<bbjimmy>
Skipp_OSX I had a BP6 running BeOS with two processors back in the day.
diver has quit [Read error: No route to host]
<bbjimmy>
Alass, the Mobo died.
<scantysnax>
I had an ECS D6VAA running two P3-700's. best computer ever. had to recap the board in early 2000s
<Skipp_OSX>
good man
<scantysnax>
from then on, always made sure to get the best capacitors for my projects.
<Skipp_OSX>
I missed the boat on that one unfortunately, had a single processor Celeron like a savage.
<scantysnax>
nice one.
<Skipp_OSX>
BP6 was revolutionary despite its capacitors and its flawed ATA controller. Definitely a touchstone in computing I guess you had to be there in 1999 to get it.
<Skipp_OSX>
Nowadays having 2 processors is normal but back then it was a breath of fresh air assuming you were willing to run NT or Linux and NT4 was... well not great.
<scantysnax>
i found the d6vaa on an old beos hardware compatibility list.
<bbjimmy>
it ran like a dream for about 6 years.
<Skipp_OSX>
Actually Linux was not great either recently saw vid on RedHat 6.1 I remember it well.
<scantysnax>
I remember old linux too, mostly slackware.
<scantysnax>
i think at that time, OSS cost money if you wanted sound.
<Skipp_OSX>
yeah I remember accelerated X good times, paying for an X11 server to get graphic acceleration because XFree86 didn't have it.
<scantysnax>
this was 98 or 99 i think.
<scantysnax>
fun times.
<scantysnax>
i think i first got cable internet in 98
<scantysnax>
maybe 99.
<Guest6226>
phschafft, correct
<phschafft>
I mean there is some way to read the CPU usage... but I'm not an expert on that.
<phschafft>
and maybe you can use that and write it down on some port of some sort.
<Skipp_OSX>
RedHat 6.1 released October 1999 Abit BP6 in 1999 sometime...
nephele has quit [Quit: Vision[]: i've been blurred!]
gouchi has joined #haiku
AlienSoldier has joined #haiku
<AlienSoldier>
waddlesplash: well that made it simple, those option dont even exist on my old laptop
<waddlesplash>
AlienSoldier: they do
<waddlesplash>
they're Haiku bootloader options
<waddlesplash>
not BIOS options
<waddlesplash>
they're in the safe mode options menu in the bootloader
<AlienSoldier>
ho, true, ok, back to it!
AlienSoldier has quit [Remote host closed the connection]
<phschafft>
waddlesplash: do you maybe have a very quick hint for Guest6226 on how to read the current CPU usage?
<Guest6226>
a pointer to the correct documentation or Header in the source code be a great help
<Guest6226>
im not a haiku person unfortunately, iv been asked on behalf of a friend to do this
<Guest6226>
he giving me a small financial incentive to do it,
<phschafft>
Guest6226: may I ask where you're located?
<Guest6226>
Thats fine, England
<phschafft>
ah, not too far away.
diver1 has quit [Read error: No route to host]
<Skipp_OSX>
CPU usage is found in Process Controller in the Deskbar. You can use Activity Monitor to get a graph or Process Controller to see each thread in real time.
<phschafft>
Skipp_OSX: yes; but the question is more how do they know?
<Guest6226>
ill git clone the source of haiku when i get home later and have a poke around
AlienSoldier has joined #haiku
<AlienSoldier>
waddlesplash: does nothing kernel still dont even light up.
<AlienSoldier>
waddlesplash: so i guess i need to have the old haiku loader name match the new one i will replace?
<OscarL>
"ERROR: event_base_loop failed: No error" ok then :-D
<AlienSoldier>
waddlesplash: ok, renamed the haiku loader from 201 to 228 in /system/packages and going to try to boot that
AlienSoldier has quit [Remote host closed the connection]
Julia[m] has joined #haiku
<phschafft>
OscarL: thank you. :))
<Begasus>
jmairboeck, not sure if I'm doing something wrong here but ... https://0x0.st/X6wD.png
<Monni>
OscarL: Is that when things are expected to fail, but don't fail, so "No error" is fatal...
<OscarL>
no problem :-) easy to rembember that Pulse was basically the PoorMan's blinken lights :-P
<Begasus>
doesn't find cmake binary? (me thinks)
<Skipp_OSX>
no grey stripe, must be Qt
<OscarL>
Monni: :-D. Not in this case. Seems some spurious error from libevent.
<Begasus>
done for today, will catch up in the morning
<Begasus>
cu peeps!
Begasus has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
Monni: in case you're interested... try "links haiku-os.org", exit once it has loaded. you'll get that error twice on the shell.
<waddlesplash>
Monni: did you see my messages above?
<waddlesplash>
about collecting more logs
<OscarL>
meanwhile... hrev58228 going strong on the cursed netboot. 6 hs uptime.
tuaris has joined #haiku
<jmairboeck>
Begasus[m]: thanks, looks like there is something wrong
<jmairboeck>
Begasus[m]: looks like the cmake data path is configured wrong in the GUI version (I did get redefinition warnings when building it)
<Monni>
waddlesplash: I got it as far as package_daemon complaining about missing userland_fs... Seems like there is some wonky going on with creation of the nightly anyboot iso...
AlienSoldier has joined #haiku
<AlienSoldier>
waddlesplash ok, i am now able to boot in my 210 revision with my previous state, 228 revision still don't light the kernel icon and HD don't show activity.
<AlienSoldier>
waddlesplash oops, i mean 201 revision
MonniTheKitten has joined #haiku
<OscarL>
alright... almost 6:42 hs uptime... this thing is holding better than I am. /me crashes.
<OscarL>
See you around folks.
OscarL has quit [Quit: zzzZZZzzz]
<scantysnax>
are the "idle threads (as seen in ProcessController)" supposed to always be all green, or can they vary like other threads?
<AlienSoldier>
scantysnax they vary here
<AlienSoldier>
in theory they should get shorter as other task grow longer
<scantysnax>
two cpus are shorter from running falkon
<AlienSoldier>
waddlesplash can't see the SMP option in the bootloader, so there is still that i was not able to test.
mmu_man is now known as Guest6242
mmu_man has joined #haiku
<waddlesplash>
AlienSoldier: if the machine has only 1 core then it may not appear
<waddlesplash>
yes, a serial log from the failed/stalled boot will be very helpful
Guest6242 has quit [Ping timeout: 480 seconds]
<waddlesplash>
I hope anyway
nephele has joined #haiku
gatekeeper_ has joined #haiku
gouchi has quit [Remote host closed the connection]
Guest6226 has quit [Read error: Connection reset by peer]
nephele has quit [Quit: Vision[]: i've been blurred!]
nephele has joined #haiku
nephele has quit []
gatekeeper_ has quit [Quit: Leaving]
xet7 has quit [Quit: Leaving]
xet7 has joined #haiku
ermo[m] has joined #haiku
<ermo[m]>
Hi. Haiku supporter here. I was looking at the website for intel_extreme and radeon_hd information. It was mentioned that the intel_extreme driver supports chips as far back as haswell? What does "support" imply in this context? Any acceleration? Or is it just "this is recognised and can do basic 2D drawing and possibly supports power-saving"?
<ermo[m]>
or, to rephrase the question: Suppose I have a couple of CPUs with haswell and skylake iGPUs and that I have various radeon cards. Which of those is more likely to be power efficient, assuming there is no real 3D acceleration in Haiku (because it's hard or something similar)?
<ermo[m]>
The context is that I would like starting using Haiku more regularly, so was wondering which system I should be using for it. =)
<ermo[m]>
* The context is that I would like to start to use Haiku more regularly, so was wondering which system I should be using for it. =)
<waddlesplash>
its just modesetting and sometimes accelerated cursor
<waddlesplash>
no accelerated drawing
<waddlesplash>
the Intel driver may be the better bet atm
<ermo[m]>
(sorry for the edits; I forget that IRC does not support it natively)
<ermo[m]>
Ack. Thanks. =)
<ermo[m]>
How the NVIDIA Vulkan thing going? ISTR that there was a focus on that because it is ... "easier" than attempting to retrofit the linux DRM/KMS stuff onto the Haiku code base?
<ermo[m]>
How *is
_-Caleb-_ has left #haiku [#haiku]
_-Caleb-_ has joined #haiku
flag_ has joined #haiku
<AlienSoldier>
waddlesplash i will be back in town with my pc's in about a month. if the provlem is still there i will connect them then.
flag has quit [Ping timeout: 480 seconds]
nephele has joined #haiku
<nephele>
Begasus[m]: seems minetest changed their name to Luanti, neat.
<FreeFull>
Seems they're currently in the process of changing the name
<nephele>
Well, they published a blog article about it, so i consider this finished :)