<Begasus>
JulesEnriquez[m], seems it's looking in the wrong places? https://bpa.st/KVKDG
<Begasus>
Searching file "data/gui.xml"... (not found) (should be /boot/system/data/gui.xml)
<Begasus>
although I would rather see this ending up in a subdirectory
<Begasus>
PS, still morning here :)
<Begasus>
postinstall script is wrong : cp -r /boot/system/data/libresprite/* /boot/home/config/settings/libresprite/data/
<Begasus>
well actually it's what it should be, packaging is wrong
diver has joined #haiku
<Begasus>
not sure why this is done ... ln -sfr $prefix/bin/libresprite $appsDir/libresprite
<Monni>
Begasus: If it's a GUI application, that symlink makes perfectly sense...
nephele_xmpp has joined #haiku
nephele_xmpp has left #haiku [Error from remote client]
nephele_xmpp has joined #haiku
janking has joined #haiku
DKnoto has quit [Ping timeout: 480 seconds]
<Begasus>
no Monni, it doesn't need one in $biinDir and in $appsDir
|cos| has joined #haiku
<Begasus>
it already has a link to Deskbar menu, so no need for the one in $appsDir
<phschafft>
you know you have a bad morning when you try to tab complete passwords and fail.
<Begasus>
lol phschafft
<Monni>
Begasus: Well... the one in $binDir is kind of redundant IMHO... But that happens when you try to hack something not designed for Haiku filesystem layout...
<Begasus>
Monni, it can be controled in Terminal to some extend, so $binDir is perfectly fine there
<|cos|>
I booted up my Haiku-laptop today. \o/ However attempting to update it fails, `pkgman full-sync` gives checksum error. :(
arjen_ has joined #haiku
<Begasus>
saw something about that in the forum |cos|
<|cos|>
Begasus: Ah. I'll have a look there. Thanks!
<Begasus>
which package |cos|?
<Begasus>
if it's a third-party package it shouldn't be too hard to fix
<JulesEnriquez[m]>
Begasus: Nice, what are the changes needed in the recipe?
<Begasus>
will push :)
<Begasus>
ah, no, need to extract the patch first
HaikuUser has joined #haiku
DKnoto has joined #haiku
HaikuUser has quit []
<|cos|>
In other words, I get the exact same bad checksum for as in the forum. I'll add that to the thread.
<Begasus>
checking rebuild
<Monni>
I don't usually use "pkgman full-sync" unless I switch between two clean installations, like switching from beta4 to beta5...
<Begasus>
"pkgman up" here (too many diverging files)
<Monni>
I just copy the hpkg files... recently package_daemon has glitched due to too many dependent packages...
BrunoSpr has joined #haiku
BrunoSpr has quit []
<Monni>
pkgman just asks me to reboot and then it reactivates all packages... Usually there is a few conflicts and then it starts to resolve them before loading desktop...
BrunoSpr has joined #haiku
BrunoSpr has quit []
<Begasus>
people should check configure output (no mather the build system): -- Looking for connect - found
<Begasus>
wasn't found earlier
<Monni>
Things change... configure scripts must adapt to added or moved functions...
BrunoSpr has joined #haiku
<Begasus[m]>
the source for libresprite didn't change, still on the same version as in haikuports, not bumping version for no good reason (now)* :)
<Monni>
sources don't need to change... sometimes headers change outside the source tree...
<Begasus>
that "mostly" only happens at build time, not when configuring the source
<Begasus>
cmake checking for socket/gethostbyname ... nothing to do with source files
<Monni>
cmake is rocket science... you rely on files in so many directories...
<Begasus>
I'm just a simple guy, not seeing rocket science with cmake :)
<Monni>
I see rocket science even in cows farting...
<phschafft>
oh yes.
ebruce613 has left #haiku [Error from remote client]
<Begasus>
anothing issue ... /bin/sh: line 1: xdg-open: command not found
<Monni>
I hardly ever use things like cmake modules or pkg-config... When things don't make sense, it's usually waste of time trying to figure out where the error actually is...
<Begasus>
it starts with looking at the output in Terminal when building, that the first indecation
<Begasus>
then on failure check the logs (if you can find them :D)
<Begasus>
so now I need to see where it's using xdg-open ...
<Monni>
I know how to run cmake in trace mode or how to use grep to find the current location of the build logs... But sometimes the log doesn't have enough information...
<phschafft>
Begasus: hm. I think the application should call 'open' not 'xdg-open'. but it might be a good idea for Haiku to have an alias for that.
<phschafft>
as such an alias is a very low hanging fruit to get closer to standard.
<Begasus>
ret = std::system(("xdg-open \"" + file + "\"").c_str());
<phschafft>
I mean that is bad code to begin with.
<Begasus>
yeah, seen it around phschafft
<Begasus>
better? ret = std::system(("open \"" + file + "\"").c_str());
<phschafft>
still my statement holds. even for that bad code.
<Begasus>
that's up for the big guns phschafft :)
<phschafft>
Begasus: not really. but it makes it work the same wrong way as before. just before it always failed.
<Begasus>
just need it to open an URL in the app menu
<phschafft>
one-argument system-style calls are a huge problem for both performance and security. beside stability and things like hardly debugable for any enduser.
<phschafft>
but that is something upstream needs to fix.
<Begasus[m]>
now that is rocket science to me :D
ebruce613 has joined #haiku
<phschafft>
the problem with things like that is that if 'file' contains anything that looks like code to the shell it will happily execute that.
<Begasus>
just using "open" instaed of "xdg-open" is enough for now to fix it :) (and no files vanished in checking) :)
<phschafft>
but also your boot is gone, rm -rf does not stop because it cannot delete some parts.
<phschafft>
but yes, having as much of the filesystem read only as possible is clearly a good thing.
<Begasus>
biab :)
deneel has quit [Quit: deneel]
<phschafft>
same for memory. the more of your memory is read only the better.
<Monni>
Some strings definitely need to be sanitized before passing to certain functions...
jmairboeck has quit [Ping timeout: 480 seconds]
<phschafft>
yes. but one-argument system-style calls are just wrong. you cannot make them good or safe.
<phschafft>
they are just a mistake made in the past. and that is where they belong.
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
<Monni>
at least the actual command and its parameter list should be separate arguments... But even that doesn't make sure the argument is what the command actually expects...
janoz has joined #haiku
<phschafft>
naturally. however it takes out the shell running random code. it limits the problems significantly. and it also improves performance hugely.
janus has quit [Ping timeout: 480 seconds]
janoz is now known as janus
<Monni>
phschafft: true...
HaikuUser has joined #haiku
<phschafft>
the what-does-the-program-expect problem is actually something nephele_xmpp and I had chats about weeks ago.
HaikuUser has quit []
<Monni>
bbl... almost dinner time...
<phschafft>
and it came down a bit to us waiting for some ELF related stuff that is needed first in Haiku.
<phschafft>
so that is an open topic. just a second one. :)
<phschafft>
have a good dinner. I for myself will now have a little nap.
<Begasus>
enjoy both, taking a short break in a bit too
<Begasus[m]>
not merging yet, I think it shouldn't be needed to copy the data files around, will do another run later to move all in $appsDir/LibResprite or the like
<Begasus[m]>
break* :)
frkzoid has quit [Ping timeout: 480 seconds]
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
frkazoid333 has joined #haiku
diver has quit [Ping timeout: 480 seconds]
deneel has joined #haiku
arjen_ has quit [Quit: Vision[]: i've been blurred!]
arjen_ has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
ebruce613 has left #haiku [Error from remote client]
ebruce613 has joined #haiku
<Begasus>
break * over
diver has joined #haiku
<Begasus[m]>
something fishy with that checksum, the one installed after an update is different then the one being pulled by others now
diver has quit [Ping timeout: 480 seconds]
BrunoSpr has joined #haiku
DKnoto has quit [Read error: Connection reset by peer]
DKnoto has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
yann64 has quit [Quit: Vision[]: i've been blurred!]
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
BrunoSpr has joined #haiku
BrunoSpr has quit []
janking has quit [Quit: Vision[]: i've been blurred!]
BrunoSpr has joined #haiku
<Begasus>
kallisti5[m] did the haiku repo recently switched host?
<Begasus>
if so, maybe that could explain the wrong checksum (although it shouldn't)?
<kallisti5[m]>
Need more details, but yeah. Haiku changed repos
yann64 has joined #haiku
<Begasus>
on the forum, checksum error on haiku-r1b5 package
<Begasus>
I think I updated before the change, current users are getting the error on updating their system
ebruce613 has left #haiku [Error from remote client]
ebruce613 has joined #haiku
mmu_man has joined #haiku
_-Caleb-_ has left #haiku [#haiku]
jmairboeck has joined #haiku
_-Caleb-_ has joined #haiku
OscarL has joined #haiku
* OscarL
waves to the room, not clearly seeing faces (really bad sight day today, for some reason)
<nephele_xmpp>
Hello OscarL
<OscarL>
hello nephele_xmpp! (and Begasus, and kallisti5[m], if I'm reading logs right, LOL :-D)
<Begasus>
Hi OscarL nephele_xmpp :)
<Begasus>
wasn't sure you were in the channel nephele_xmpp :)
<nephele_xmpp>
Begasus: Hmm, sometimes I am not sure either. XMPP rooms (muc) only have room membership when online, but not sure if the bridge does that correctly
<Begasus>
saw your nick, but yeah, doesn't indicate you are around :)
<Begasus>
anyway, had the same question for you as I had for kallisti5[m] on that checksum
<Begasus>
probably something went wrong with the transition
imrahil has joined #haiku
<OscarL>
updated to latest nightly earlier today, no issues. So problems either are limited to beta5... or a regional thing?
<Begasus>
guess it's only the beta package OscarL
<Begasus>
I still got it before the move
<nephele_xmpp>
Begasus: I don't maintain any Haiku servers so can't help you with the why :/
<Begasus>
nah, was just making noise nephele_xmpp :D
<Begasus>
as some background on the forum issue :)
<OscarL>
seems to me like we need more people helping those that do maintain the servers... but no one knows what do to do to help? (at least that's my impression from the sidelines)
<OscarL>
bus factor of 1.5/2 ain't great.
<nephele_xmpp>
I think nielx and kallisti5 mainly do that work. I am kind of interested in helping, but most of it is on github and containers are a bit out of my depth...
arjen_ has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
on other news... the "cursed netbook" seems to not be so cursed anymore (thanks to waddlesplash's work on nightlies, and nephele_xmpp's "reverse-curse" :-P)... too bad it SO. DARN. SLOW. to compile stuff :-D
arjen_ has joined #haiku
<Begasus>
karma! :P
<OscarL>
nephele_xmpp: right. And besides the actuall helping with proper maintenance/development... I have "joked" in the past with giving Begasus "reset button" to at least the buildmasters... might not be a bad idea, I still think. :-)
<OscarL>
*missed an "a" after Begasus there.
<Begasus[m]>
could use one for the buildmasters yeah! (me grins) :D
<nephele_xmpp>
Might not be a bad idea indeed
<OscarL>
I mean... I know it is not ideal... but beats having the buildmasters stuck for days till maintainers have time to react.
<botifico>
[haikuports/haikuports] Begasus 9a65da6 - TuxPain, add RC for new TuxPaint and stamps (#11224)
<BrunoSpr>
Hello
<Begasus[m]>
Hi BrunoSpr
<BrunoSpr>
Hi Begasus[m]
<OscarL>
on some positive news... I got a new (cheapest ever, still quite a lot of money for me) SSD to use with Haiku!
<BrunoSpr>
I just tried to compile the old SNES Beos port. No success! Too difficult for me! Maybe it works only with 32bit, anyway no idea! I just downloaded the source code from BeShare
<OscarL>
Please welcome (and laugh with me at) my new "Hiksemi WAVE 120 GB" SSD :-D
<B2IA>
(BrunoSpr) I searched for snes and downloaded the snes9x-1.41-1_src-incl.zip
<OscarL>
BrunoSpr: do you need that particular program? or some other emulator would work for you?
<OscarL>
alternatives are: BeNES and Pretendo, at least.
<BrunoSpr>
I just got 512 GB Samsung SSD here, and it makes booting much faster...
yann64 has quit [Ping timeout: 480 seconds]
<BrunoSpr>
I remember it was working for me quuit some time ago, maybe too long time ago, and maybe it was 32bit only
<BrunoSpr>
Thanks OscarL, but I tried to use this two SNES emulators as I remember them working very well
<OscarL>
BrunoSpr: alright. Just suggesting some quick alternative, until someone can help you out with your compile issues :-)
<B2IA>
(BrunoSpr) Ah there is BeNES here on BeShare
<B2IA>
(BrunoSpr) Yes, I will try it
<OscarL>
(not much of an emulator user here, at least not in the last 2 decades)
HaikuUser has joined #haiku
<Begasus>
never got that one to build BrunoSpr
HaikuUser has quit []
<OscarL>
old emulators (from BeOS era), will most problably only work on 32 bits, unless someone actively put some love into them...
<OscarL>
as scanty is doing with Pretendo.
* Habbie
looks up
<Habbie>
i will never get used to there being two entirely unrelated Pretendos
<B2IA>
(BrunoSpr) Oskar, yes it seems it is only 32bit, cannot install it on 64bit here
<B2IA>
(BrunoSpr) it is a hpkg
<OscarL>
Habbie: I had to look up what you were referring to... because for me... there's only one Pretendo :-P
<BrunoSpr>
oh yes gcc2 and X86 only
<Begasus>
or maybe it's you pretending there is only one?
<OscarL>
BrunoSpr: that will be a common thing for old BeOS apps with no current maintainer, sadly.
<BrunoSpr>
These work very well on X86 as I have them there
<OscarL>
Begasus: at least I'm not pretending that there was one *original* "Pretendo"... and it wasn't "Wii U" related :-P
<OscarL>
*"that there wasn't an" (can't type, as usual)
<BrunoSpr>
Begasus, there where lots of offset errors, trying to compile SNes9x
<nephele_xmpp>
someone said Wii U?
<BrunoSpr>
Begasus, if you could not got it compile, how can I
<Begasus>
lol BrunoSpr, I'm no expert on the matter :)
<OscarL>
BrunoSpr: porting programs from 32 bits to 64 is not always straight forward. Dificultly can range from, "Ok, I need to focus on this for a few hours"... to "I better re-write this thing from zero".
<BrunoSpr>
It was my intention, to prove, that a BeSNES only emulator is working much faster and better than the bloated and slow libretro ports
<BrunoSpr>
OscarL, oh it is not straight forward... Ok, well time to give up today...
<OscarL>
BrunoSpr: well... doubt that needs much "proof". Same with older BeMAME vs newer MAME.
<BrunoSpr>
Thanks anyway, I hope someone will have a look into this by accident
<Begasus[m]>
k, PR for rizin is up
<BrunoSpr>
yes what a pitty
<Begasus[m]>
it has been on the table and asked for for a long time already BrunoSpr
<OscarL>
BrunoSpr: usually in emulation things go like this....
<OscarL>
do you want a fast emulator? or an accurate one?
<Habbie>
OscarL, for me there's also one pretendo, and it's not the same one :D
<OscarL>
Habbie: I'm just going for the oldest I know :-)
<BrunoSpr>
Ah, Begasus[m], funny I was trying it almost 4 years ago, thank you!
<Begasus[m]>
and even before that one it has been mentioned :)
<OscarL>
BeMAME was AWESOME even on my Athlon 900 MHz... newer MAME... not sure my Phenom can handle it :-P
tuaris has joined #haiku
<Begasus[m]>
new mame building OscarL? :P
<BrunoSpr>
OscarL, yes maybe, but I do not own a single MAME game
<OscarL>
Begasus[m]: build... and just using it! :-)
<BrunoSpr>
OscarL, oh nice
<Begasus>
building that monster is a pain in the *
<OscarL>
BrunoSpr: I'm just giving examples of projects that got "slower with time", because they started to focus more on accuracy of emulation, than on user experience.
OscarL has quit [Ping timeout: 480 seconds]
<Begasus[m]>
k, package for syncterm added in the forum topic (wip) :)
<Begasus[m]>
switched the link to codeberg :P
OscarL has joined #haiku
MisthaLu has joined #haiku
<OscarL>
darn power line glitch
<BrunoSpr>
Begasus[m], does codeberg works better than Github with HAIKU?
MisthaLu has quit []
<Begasus[m]>
it's still git BrunoSpr
<BrunoSpr>
Oh, ok will have a look then
<Begasus[m]>
and mostly use that from Terminal, so user wise I don't see any real difference
<BrunoSpr>
ah I got it thanks
<Begasus[m]>
and some people have a thing for ...
<BrunoSpr>
Github is slow on WebPositive
<OscarL>
BrunoSpr: codeberg is nice... but I had more issues with it being offline than with github, so far.
<Begasus[m]>
not doing much browser related things lately in the VM, so I use the host for that here (same as atm with matrix)
<BrunoSpr>
for browsing the files...
<OscarL>
Begasus[m]: if you mean only the Web UI... I think codeberg should be just fine.
<OscarL>
sorry, s/Begasus[m]/BrunoSpr/
<Begasus[m]>
I don't have any real problem with either OscarL, just anything browser related stuff is pulling resources away when building
<Begasus[m]>
ah :)
<OscarL>
just mentioning that, while I love what codeberg does... I have it failing to take my "git push"es more times than github.
<OscarL>
*I had it
<Begasus[m]>
not using it that much here, so far no issues on that
<nephele_xmpp>
github not work -> microsoft
<nephele_xmpp>
codeberg not work -> microsoft sabotage!
<nephele_xmpp>
/s
<Begasus[m]>
LOL
<OscarL>
Begasus[m]: not a heavy user of codeberg (so far), either, just mentioning that I had more problems pushing to it that I would like.
<Begasus[m]>
logging in at gitlab can be a pain though :P
<OscarL>
*** thus my need to have a local instance of "Forgejo" going on.
<Begasus[m]>
:)
<Begasus[m]>
worsed is SF
* Begasus[m]
thinks if "worsed" is even a word ...
<OscarL>
I find SF better than gitlab... that should speak volumes :-)
<Begasus[m]>
always had troubles creating/managing branches at SF
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
<OscarL>
I haven't even hooked up my new SSD, and yet... hot dam... this thing feels cheap! :-)
<Begasus[m]>
heh
<OscarL>
not only half+ empty...
bbjimmy has joined #haiku
<OscarL>
but also way too heavy compared to a kingston 240 drive (that has a metal body)
<OscarL>
not to mention the clackely-clack sounds it makes when you tap it :-D
<BrunoSpr>
OH, Pretendo is 32bit only too! What?
* OscarL
hopes he just didn't got some 4 x 32 GB pendrivers hooked up into a SSD interface.
<OscarL>
BrunoSpr: Pretendo was originally 32 bits only, yes... as ALL BeOS software.
<nephele_xmpp>
That's ridicilous. It's 16 x 8GB
<Begasus>
OscarL in raid? :P
<OscarL>
BrunoSpr: At least with Pretendo we have a nice fellow (scanty / scantynax) working on porting it.
<OscarL>
nephele_xmpp: don't give me hopes of more parallel I/O, please! :-)
<OscarL>
Sadly, my best PC only has SATA 3.0 GB/s, USB 2.0, so... I'm f*** anyway :-D
<nephele_xmpp>
You are getting high speed! (according to the usb soec)!
<OscarL>
I'm banking on paralleol I/O being more important than raw read/write speed (compared to HDD). Because 3.0 GB/s SATA vs SATA 3 (6 GB/s) speeds.. yeah... quite a cutoff :-(
nephele has joined #haiku
<OscarL>
welp, at last I got a drive I will treat as "disposable". I don't really care what happens to it, so...
<OscarL>
it will be good to see how much of using it for haikuports works trashes the cheapers ever SSD.
<BrunoSpr>
brb
zard has joined #haiku
BrunoSpr has quit [Quit: Vision[]: Ich wurde eingeweicht!]
<OscarL>
drive doesn't even has screws... it just has rivets instead. *that's* when you know you're in bottom line.
<Begasus[m]>
got a few of those here already :)
<OscarL>
Begasus[m]: technology comes late down here :-D
* nephele
looks up what "rivet" means
<Begasus[m]>
heh
<Begasus[m]>
OscarL also got a pile of pendrives now :D
Begasus has quit [Quit: Vision[]: i've been blurred!]
<nephele>
OscarL: the manufacturer just doesn't want you to open it for self-service
<Begasus[m]>
biab
<OscarL>
nephele: I got that "rivets" after looing up what "remaches" looked like in English. I only knew the word for them in Spanish/Castillian (use the latter if you care).
<OscarL>
right, doubt there's much for self-servicing on this...
<Begasus[m]>
you only get ...... one try
<OscarL>
but I would just seal it shut on that case, rather than using either screws or rivets :-)
<OscarL>
if they go for "use the least possible amount of materials possible"... why not just use a sealed plastic unit=
<OscarL>
s/=/?/
<zard>
Trying to block magnetism?
Begasus has joined #haiku
<OscarL>
zard: more like... try to comply with some weight measure, I think :-)
* OscarL
worked long enough with QA workers to know a few tricks around "passing" tests.
<Begasus[m]>
with spell correction?
dovsienko has joined #haiku
<OscarL>
Begasus[m]: correct spelling wasn't part of the specs we needed to match results against, so... no :-P
<OscarL>
only math :-)
mattlacey has joined #haiku
<OscarL>
(welp.. we *tried* to match spelling too, but... higher ups deemed it "not worth of such effort"... till they got hit by a law-suit! :-D
janking has joined #haiku
<Begasus[m]>
lol
<Begasus>
OscarL, and every one was like "WAT"!? :)
<OscarL>
Begasus: worse...
<Begasus>
career end?
<OscarL>
"why we weren't told about this *obvious* issue"?
<OscarL>
we had trails of several mails warning about those potential bugs.
<Begasus>
hmm ... comunication ... where have I seen that before on the workfloor ..
<OscarL>
still, in the end, we (whole QA team) got cut off, so... there's your answer...
<Begasus>
:(
<Begasus>
well, your input here IS appreceated! (no pay though) :)
<OscarL>
also--- EPSON's leadership in my country's market dropped HARD in the last decade (from when we where laid off to today)-
<OscarL>
Begasus: thank you :-D
<Begasus>
they are not as comon here also as they used to be
<Begasus>
libs/filesystem/src/directory.cpp:321:21: error: invalid application of 'sizeof' to incomplete type 'char []'
<Begasus>
so far trying to build boost1.83 with icu74
janking has quit [Ping timeout: 480 seconds]
HaikuUser has joined #haiku
ebruce613 has left #haiku [Error from remote client]
HaikuUser has quit []
<OscarL>
Here EPSON, at some point, had the Point-of-Sale market pretty much as a monopoly... we were that good down here.
mmu_man has quit [Ping timeout: 480 seconds]
<OscarL>
not so much on home-printers market.
<OscarL>
that was always really crappy.
SLema has quit [Ping timeout: 480 seconds]
_-Caleb-_ has left #haiku [#haiku]
_-Caleb-_ has joined #haiku
<OscarL>
heck, a team of less tha 8 people handled POS firmware + QA for more than 8 countries. How such a team gets scrapped..., and "sales team" getting bonuses insted of us... I would never understand.
<Begasus>
it's beyond our minds OscarL, no answers to those questions :(
<OscarL>
but HR somehow expanded 3 fold in the same period. Hmmm.
<Begasus>
I'm glad I don't have to deal with that sh*t anymore :)
<OscarL>
Begasus: yeah...
<OscarL>
after I got laid out in 2014... at first I was really scared..
<OscarL>
then I learned to think: f*** it. I don't need most of this.
ebruce613 has joined #haiku
<dovsienko>
let's say 99% of all businesses that were started sooner or later folded, and quite a lot fold due to internal problems, no matter how famous
<OscarL>
Hello there dovsienko!
<dovsienko>
hi. just nipping to report that AMD triple fault still stands (syslog attached to the bug report). waddlesplash, if you need any additional diagnostics, let me know
<nephele>
dovsienko: I know, i get an email if you comment on a subscibed ticket ;)
<OscarL>
Wanted to share with you the good news that I got an 120 GB SSD (cheapest one ever, mind you, and on a slow SSD 3.0 GB/s interface, but still!!! :-D)
<nephele>
waddlesplash should get one too
<dovsienko>
otherwise it is perfectly fine to run the box with SMP disabled, it is taking too much engineering effort to support this hardware given its age and low spec
<dovsienko>
OscarL: finally some relief. you will like the difference
mmu_man has joined #haiku
<Begasus[m]>
hi there dovsienko
* OscarL
needs to be reminded to do proper benchmarks between same CPU/RAM and HDD vs SSD configurations, at least when building Python 3.13
<dovsienko>
if you want the best performance, keep the entire build in RAMFS if possible
<nephele>
not sure if our RAMFS is really the fastest filesystem :P
<OscarL>
dovsienko: that's one of my goals, yes, but I have tested RAMFS, and not convinced it it that much faster.
<dovsienko>
my impression used to be it is faster than disk, let's benchmark it
<nephele>
that's good. Numbers we can work with
Coldfirex has joined #haiku
<nephele>
if you find any terrible bottlenecks would be nice to report them too
<OscarL>
I/O stalls seems to be localed elsewhere, IME.
<OscarL>
*located
<Coldfirex>
Is the a cli command to get the nic link speed? The GUI is showing auto-select for me
<OscarL>
I can't really type realiably anymore today... I better sign off. See you around folks.
<Begasus>
cu OscarL!
<OscarL>
Have a great day.
<Begasus>
take care
OscarL has quit [Quit: Gone with the wind]
<Coldfirex>
Thanks. ifconfig also returns Auto-select. This is using the newer VMXNET3 driver
* dovsienko
is benchamrking RAMFS...
<Begasus>
getting typos too dovsienko? :)
<dovsienko>
Coldfirex: if that's a virtual device, the notion of link speed either does not apply, or applies differently
<Coldfirex>
Im guessing differently. That driver, for windows at least, will show 10gb, etc where applicable
MonniTheCat has joined #haiku
janking has joined #haiku
<Monni>
I get 100 Mbit and 100BASE-TX in ifconfig with virtual NIC..
<dovsienko>
Coldfirex: there is quite a few engineering layers even in the simplest NIC, so if a driver expects a real/emulated hardware, one of the elements would be the link speed, but there would also be other bits that require proper driving
<nephele>
I was quite amused at the Haiku driver figuring out the link speed of wires for ethernet
<nephele>
I just took some long telephone wire and crimped it to ethernet
<nephele>
sometimes it would be a fastish cable, sometimes slow
<dovsienko>
if, however, the NIC is virtual by design, in this case the intention is to cut as many corners as possible, and the hardware blocks that would produce the link details would be just missing
<nephele>
it failed in the "ethernet over twisted pair" bit, since there was no twicting
<dovsienko>
in which case the driver could still report some hard-coded value to the OS for backward compatibility reasons, but the actual throughput would be decided by the hypervisor
<dovsienko>
so it would be best not to presume that example as the right thing to do (unless it can be reasoned about in engineering terms)
<Coldfirex>
Other virtual NICs report a link speed (at least the 1000 I usually use). I was just curious if I could get it to be 10GB like my other VMs
<dovsienko>
the nominal label can be 10Mb/s or 400Gb/s, but you need to benchmark it to see how it actually performs
<phschafft>
nephele: there are actually modes for cat < 5.
<phschafft>
nephele: I have some cards that support 100Base-T4, which allows for cat 3.
<nephele>
phschafft: hmm, i don't know about that, i only searched for a 8p8c ethernet standard that worked with *only* 4 wires
<dovsienko>
the hardware block that brings the link up according to autoneg/manual settings is either the PHY or the one next to it (my memory isn't a reliable source here)
<phschafft>
I mean different there are standards for for 1, 2, and 4 pairs. but I have also seen some card working fine with 3 pairs.
<dovsienko>
the point it, the driver does not get to decide whether the link is present, instead it tell the PHY or the PMA or the PMD what settings it should use, and those report the result
<dovsienko>
for 100Mb/s Ethernet there are two standards, which are often confused: 100Base-T and 100Base-TX
<dovsienko>
if you don't know, it most likely is 100Base-TX, which uses only 2 pairs
<dovsienko>
a very long time ago I worked as one of the IT guys in an organisation with a campus, and they had Ethernet links between the buildings
<Coldfirex>
not fiber?
<dovsienko>
nope, not at the time
<dovsienko>
so one time I needed to have a look at one of the cables, which was believed to be 200m long, I think, so I assumed there was a repeater or two somewhere in the middle
<Monni>
I remember the times when schools started shifting from Fat Ethernet to Fiber ;)
<Monni>
You only need repeater after about 300 metres ;)
<dovsienko>
the local cable guy eventually took me to a room where one half of the link joined with the other, and it turned out to be two pieces of very old phone cable joined by jumper cables
<dovsienko>
that's when I realised it was already good enough the link mostly worked at 10Mb/s
<Monni>
Pretty much kills the connection only when using computer connected to far end of the cable...
<dovsienko>
it probably helped that the wires were very thick and properly soldered to the contacts, and the cable had a thick screen around it
<dovsienko>
52 | # The default lists apply if the user does not specify any file.
<dovsienko>
then subsequent builds did not encounter it. possibly there is another bug in RAMFS waiting to be identified
<Habbie>
ouch
<Begasus>
always fun ... rust ...
<Begasus>
almost like booting Windows after an update in the late nineties :)
arti has quit [Ping timeout: 480 seconds]
gouchi has joined #haiku
arti has joined #haiku
<Monni>
I still try to avoid compiling rust at all cost... Takes hours even on modern computer...
<Monni>
It's already enough that compiling gcc takes 2.5 hours...
<Monni>
Everything I need to upgrade to new OS release I wait until there is binaries for at least gcc 14.x and rust...
ManBoy has joined #haiku
ebruce613 has left #haiku [Error from remote client]
Skipp_OSX has joined #haiku
<Begasus>
where's the fun in that :P
ebruce613 has joined #haiku
<Monni>
Begasus: Just the fact that clang uses gcc 13.x behind the scenes makes me cough furballs...
<dovsienko>
not using RAMFS, libpcap build matrix took 26m59.800s on a single-core physical machine and 12m59.969s on a 4-core VM
<dovsienko>
hmmmmm, so RAMFS actually slows the process down, which does not look right
<nephele>
huh? i'm confused. how are those comparable?
<Begasus>
Monni, only means it's time to trim the cat :D
<Monni>
compiling is RAM heavy... If RAMFS takes too much memory, it needs to swap...
<dovsienko>
could be an effect of ccache misses during the first round after the source changed after "git pull" , let's run RAMFS one more time
<Monni>
Begasus: I have short hair cat ;)
mattlacey has quit [Ping timeout: 480 seconds]
<Begasus>
:)
<dovsienko>
nephele: it is comparable with the results from 35 minutes ago
diver has joined #haiku
<Monni>
Also, 4 virtual cores is not comparable to 1 physical core... More than 1 virtual core might share same physical core and sometimes not all virtual cores are even fully utilized...
<nephele>
I've not seen you post any other build times
<nephele>
also, using ccache kind of taints it ;)
dby has joined #haiku
janking has quit [Quit: Vision[]: i've been blurred!]
<dovsienko>
I use ccache when I can, on various OSes.
<nephele>
sure. but it does not belong in a benchmark.
<dovsienko>
that's disputable
<nephele>
I don't see how. A perfect cache would result in a build time of 0s, anything else is state that messes up the build comparability
<dovsienko>
if ccache is an element of most of my builds, why not make it a part of every benchmark round?
<nephele>
because it makes the benchmark useless to compare ramfs with
<dovsienko>
the build time is far from 0 even with a 100% hit rate
<Monni>
still would need to ignore result of the first round as that has too many cache misses...
<Monni>
Depends on number of files but just going through all the different source files can take almost 40% of the build time...
janking has joined #haiku
<Monni>
When I build Haiku, it can take anywhere between 1 minutes 50 seconds and 4 minutes 55 seconds...
<dovsienko>
the compiler invocation, whether ccache'd or not, still reads .c and .h files from the project source tree and still writes an .o file, so using (or not) a memory filesystem still makes an obvious difference
tqh has joined #haiku
<dovsienko>
until now it used to be uniformly faster to use a memory filesystem
<Skipp_OSX>
If I switch to an old branch and rebase, even if the changes are small the compile time is great because of all the changed files.
<dovsienko>
yes, the hit rate is not guaranteed to be 100%, it is a function of what you compiled before and what you are compiling now
<dovsienko>
if you compile the same few projects again and again (as I do), the hit rate is high and the time to compile is several times lower
humdinger has joined #haiku
<dovsienko>
then, obviously, when a compiler package updates, you need to discard the cache
<Monni>
I always use same compiler... when new compiler is released I add new entry to build matrix...
<Monni>
Still using gcc 6.5 to do some of my builds...
<dovsienko>
(you don't have to have the same opinion, but if you want to know what it looks like to me, that's how it does)
imrahil has left #haiku [#haiku]
tqh1 has joined #haiku
imrahil has joined #haiku
tqh has quit [Read error: Connection reset by peer]
tqh2 has joined #haiku
tqh1 has quit [Read error: No route to host]
<Monni>
ccache used to be one of the first things I install on a new operating system... But I never tried to benchmark builds with it...
tqh2 has quit [Read error: Connection reset by peer]
<botifico>
[haikuports/haikuports] humdingerb e4e76ec - quicklaunch: update to v1.6
<dovsienko>
using RAMFS (again), libpcap build matrix took 26m43.067s on a single-core physical machine and 13m20.908s on a 4-core VM
<dovsienko>
this time negligibly faster on the low-end hardware, and negligibly slower in a faster VM
nephele_xmpp has left #haiku [Error from remote client]
<dovsienko>
perhaps the correct way to test it would be moving the entire source tree to RAMFS, but let's not split the hair
<nephele>
... uhh how else are you doing it?
<dovsienko>
with the compiler's temporary files (.s etc) and the destination install directory on RAMFS only
janking has quit [Quit: Vision[]: i've been blurred!]
<dovsienko>
that's different from the most effective setup, yes, but this test was using my normal development repository, not a throwaway repository in memory
<dovsienko>
it takes more than one attempt to get every detail right
<nephele>
ah, i see. :)
<dovsienko>
Begasus: stress-ng 0.18.5 is out and it compiles on Haiku with no patching required
<dovsienko>
(if compiled using GCC, that is)
diver has quit [Ping timeout: 480 seconds]
nephele_xmpp has joined #haiku
nephele_xmpp has left #haiku [Error from remote client]
<dovsienko>
alright, let's not spam the channel anymore. waddlesplash, if you need any additional diagnostics or remote access to the box, let me know. it has no serial port though (maybe this could be fixed with an add-on PCIe card, not sure)
dovsienko has quit [Quit: Leaving]
<waddlesplash>
probably won't help, really
Halian has joined #haiku
arti has quit [Ping timeout: 480 seconds]
janking has joined #haiku
stux- has joined #haiku
<Habbie>
i see dovsienko crashed it again
<Habbie>
anything useful i can add this round, waddlesplash?
<waddlesplash>
probably not
<Habbie>
alright
arjen_ has quit [Quit: Vision[]: i've been blurred!]
nephele_xmpp has joined #haiku
stux|away has quit [Ping timeout: 480 seconds]
janking has quit [Quit: Vision[]: i've been blurred!]
janking has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
MonniTheCat has quit [Quit: Vision[]: i've been blurred!]
mmu_man has joined #haiku
janking has quit [Quit: Vision[]: i've been blurred!]
janking has joined #haiku
humdinger has quit [Quit: Vision[]: Oi with the poodles already!!]
<waddlesplash>
Skipp_OSX: re your question yesterday. Yes I think it'd be better without those convenience methods. But removing them should be easy?
<waddlesplash>
delete, then find&replace
<Skipp_OSX>
just for FilePanel and Desktop or all of them?
ebruce613 has joined #haiku
markynose has joined #haiku
markynose has left #haiku [#haiku]
mmlr has quit [Quit: Vision[]: i've been blurred!]
Anarchos has quit [Quit: Vision[]: i've been blurred!]
mmlr has joined #haiku
markynose has joined #haiku
markynose has quit []
HaikuUser has quit [Quit: Vision[]: i've been blurred!]
<nephele>
okay Skipp_OSX
<nephele>
I've updated the review with new values
janking has quit [Quit: Vision[]: i've been blurred!]
<nephele>
should be good now
<Skipp_OSX>
ok that looks good now, I am testing with 0.10 on dark mode.
<nephele>
0.10 in my testing was too dark to make out, 0.15 seemed fine. 0.2 may be a bit better though
<nephele>
but i think we can leave it as is and adjust based on feedback
<Skipp_OSX>
yeah ok whatever works, it was def not enough contrast at 0.05
<nephele>
anyway, hope that commit shows what i ment with using the hsl api ;)
<Skipp_OSX>
I may have to implement hsl color in color pickers...
<nephele>
Well, anything is better than the 3 sliders in pref/appearence xD
Monni has joined #haiku
<Skipp_OSX>
https://0x0.st/XI7X.png <= 0.05 and 0.95 on left, 0.1 and 0.9 on right for light and dark
<nephele>
Yeah, i think light mode is better on the left, and dark mode better on the right
<Skipp_OSX>
I like the 0.95 on light mode, and then dark mode needs the extra bit of contrast so at least 0.1 it is hard to tell it is dark red, well, I guess not impossible to see but not as easy to see as on light mode.
<nephele>
indeed
<nephele>
from humdingers screenshot 0.95 light looks good
<nephele>
I tested against another app though, the app_server test. so it looks a bit different for me. but in your example 0.10 would also look okay
<nephele>
let me know if you want me to change that still now
<nephele>
Then i can commit that, otherwise might call it a day soon :)
<nekobot>
[haiku/haiku] 48c3df2246d1 - kernel/vm: Use the proper definition for HAIKU_PHYSICAL_64_BIT.
HaikuUser has joined #haiku
HaikuUser has quit []
<Habbie>
waddlesplash, i am (sometimes) reading along to your commits - but please do let me (and dovsienko) know if you see any chance that a commit improved the AMD situation :)
HaikuUser has joined #haiku
<waddlesplash>
sure
HaikuUser has quit []
<Habbie>
(also want to note that i am not specifically eagerly waiting on any fix for that; i'm just happy to help)
<waddlesplash>
the above won't make any difference on x86_64 I don't think
<Habbie>
ack
<waddlesplash>
and based on dovsienko's log won't make a difference on 32-bit anyway
<Habbie>
dovsienko's log was 32 bit?
<Habbie>
oh the commit is for x86
<waddlesplash>
no, 64
<Habbie>
ok good
<Habbie>
otherwise i'd make a log for you this weekend :)
<waddlesplash>
but the log contains address ranges
<Habbie>
understood
<waddlesplash>
and based on the ranges, the above middle commit won't affect that hardware
<waddlesplash>
the first commit will, I think
<waddlesplash>
eh, maybe not, if this range ends below 4GB anyway
<Habbie>
i am able to change the amount of RAM in my AMD boxes btw, if that ever sounds interesting
imrahil has left #haiku [Disconnected: Hibernating too long]
imrahil has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
mattlacey has quit [Remote host closed the connection]
janking has joined #haiku
janking has quit []
<waddlesplash>
Skipp_OSX: at least the ones in the commit I commented on
<waddlesplash>
so, all of them probably
<Skipp_OSX>
well I mean the rest like IsRoot() Is Trash(), etc.
HaikuUser has joined #haiku
HaikuUser has quit []
<Skipp_OSX>
(I'm assuming yes)
ebruce613 has left #haiku [Error from remote client]
<waddlesplash>
yeah, we should invoke TargetModel()->... for clarity really
<waddlesplash>
or PoseView()->...
ebruce613 has joined #haiku
<Skipp_OSX>
I disagree, but ok
mmlr has quit [Quit: Vision[]: i've been blurred!]
imrahil has left #haiku [Disconnected: Hibernating too long]
<Skipp_OSX>
I used TargetModel() which calls PoseView()->TargetModel()...
mmu_man has quit [Ping timeout: 480 seconds]
ebruce613 has left #haiku [Error from remote client]