<Begasus>
just updating 32bit to new hrev (ready now PulkoMandy) :)
Begasus_32 has quit [Quit: Vision[]: i've been blurred!]
<Begasus>
ps, looks like we got some spammers at haikuports :/
<yann64>
:(
Begasus_32 has joined #haiku
mmu_man has joined #haiku
<Begasus>
ps PulkoMandy, another one down for flac13 :)
<Begasus>
now what was I up to yesterday :)
<Begasus>
ah libgit2
<Begasus>
so far no crash on 32bit when building +1 :D
Begasus has quit [Quit: Vision[]: i've been blurred!]
kinkinkijkin has joined #haiku
Begasus has joined #haiku
jmairboeck has joined #haiku
<Begasus>
Hi jmairboeck
<jmairboeck>
Hi Begasus
<jmairboeck>
you are faster with deleting the spam comments than I am reporting them :) Or did you report them also and they just show up as deleted to other users?
<Begasus>
already did 2 reports earlier ;)
<jmairboeck>
ok :)
<Begasus>
one did it again though, so not sure how fast GH is there :)
<jmairboeck>
there were 2 again, from both users as before, I reported the first one from the second time
<Begasus>
+1
<jmairboeck>
not sure how much it helps though, it seems these are bot accounts ...
<Begasus>
not much else we can do :(
<Begasus>
k, patch for libgit2 was adding Haiku here: +#if defined(GIT_RAND_GETENTROPY) && !defined(__HAIKU__)
<botifico>
[haikuports/haikuports] OscarL 49cc2e2 - xar: update to version 1.8.0.0~498. (#10995)
<Begasus>
got one mail back from GH jmairboeck :)
<jmairboeck>
me too :)
<nipos>
Good morning.I see that my patches from last weekend didn't have activity lately.Is there still something I need to improve before they can get merged?
<Begasus>
g'morning nipos
<nipos>
Good morning Begasus, any news about dark mode on the Buildmaster? :)
<botifico>
[haikuports/haikuports] Begasus 83adb01 - libgit2, revbump for openssl3 (#10997)
<Begasus>
no news from me nipos, haven't found my way around there, and no time to look into it :)
<Begasus>
still plenty to fix for openssl3
<nipos>
Hm ok. I see you're doing good progress on OpenSSL 👍
<Begasus>
hope it's good progress :D thanks :)
<nipos>
Sure it is. Depending on abandoned libraries would only cause more problems in the future. I'm happy to see the switch to latest OpenSSL 3
<Begasus>
ow PulkoMandy, see that error you mentioned on libgit2 at buildmasters: /boot/system/develop/headers/bsd/sys/uio.h:21:8: error: unknown type name 'inline'
<Begasus>
have to use older qtwebengine package on 32bit, so can't dump icu66 there yet
<Begasus>
weird, libgit2 builds fine here?
<Begasus>
static __inline ssize_t same local as the one mentioned at buildmaster
<Begasus>
checking something ...
<jmairboeck>
is libgit2 using an old C standard to build (or gcc2 even)?
<jmairboeck>
ah, I think buildmasters are just outdated
<Nephele>
it's a old english form for "you", Like in german you have "sie" and "du", in english only the formal variant "you" survived
vdamewood_ has joined #haiku
<Begasus>
my is fine then :)
<Begasus>
weird English :D
<Nephele>
so it's the informal one, but somehow makes shakespear sound very formal to me, even though the intent was to make characters speak informally to each other
<Nephele>
err, though spelling should be thou it seems ;D
<jmairboeck>
nephele: that is spelled "thou" (without the "gh") I think
<Begasus>
I'm just a simple guy you know? :P
<Begasus>
ah, like that
<Nephele>
jmairboeck: indeed you are correct :)
<Nephele>
Begasus: sometimes I just spit out random stuff I know :P
vdamewood has quit [Ping timeout: 480 seconds]
<Nephele>
I wonder why the "automatic" dark mode of wikipedia is not automatic
<Begasus>
:)
<Begasus>
nephele, they forgot the manual? ;)
<Nephele>
well the standard is "light"
<Nephele>
xD
<Nephele>
Maybe they should have called it "browser setting" instead
<Nephele>
maybe they are experimenting with this for a roll out
marzzbar has joined #haiku
<Nephele>
Their report button is kinda dumb. It sais "Report an issue with dark mode". I click on it, and it just goes "report received!"
<Nephele>
without me beeing able to specify what is wrong
<Nephele>
oh, by the way. link text colors may be fixed in webkit, i saw safari fixed it
<oanderso[m]>
tqh: I'm about to go to bed, but I'm on here :-)
<grepwood>
My usecase is fun and showcasing :)
<Nephele>
Showcase working drag and drop, that's way more impressive :P
<tqh>
oanderso[m], oh cool.
<Nephele>
but yeah, the long answer is that there is some wip port for some specific card, but it is not readily available. and we technically have acceleration for some really ancient cards, that doesn't correspond to modern opengl, so effectively useless now
<dovsienko>
Linux also uses Solaris partition code for its swap, which used to cause data loss in the past
<grepwood>
I plan to put Haiku on a computer originally intended as part of a 2000s themed arcade. The specific PC has the following specs: Xeon L3360, 16GB RAM, GeForce 750 Ti, SoundBlaster Audigy 2 ZS, Roland SC-88 Pro. I already have a pretty good idea what I can showcase on it so that people will want to spend their time at the arcade and use that PC. Sometimes I'm getting more technically inclined users and they respond very well to extra
<grepwood>
quirky things
<dovsienko>
grepwood: that overshoots 2000s hardware standards by 1-2 orders of magnitude
<grepwood>
dovsienko: yeah except Windows XP supports it very well
<grepwood>
the overkill hardware is meant to give 2000s games headroom for fps drops and people trying to really push the games to their limits
<x512[m]>
dovsienko: GPT GUID partition type too?
<dovsienko>
maybe get an old Celeron PC with 5.25" and 3.5" floppy drives, that may be fun to see for younger people
<grepwood>
dovsienko, for that I have a dual Pentium Tualatin setup with FreeDOS on it
<dovsienko>
x512[m]: not as far as I know
<dovsienko>
if you want people to get the sense of OS multitasking, get two old PCs with floppy drives, install Windows 95 on one and OS/2 on the other, and ask them to format a floppy and to do something else at the same time
<Nephele>
grepwood: for nvidia cards we usually don't even have modesetting drivers
<dovsienko>
Linux would also do if you don't have OS/2
<Nephele>
so if your bios supports a nice vesa resolution that's good. (or efi supports the native one), but other than that for nvidia cards there is no recourse
<Nephele>
but for (3d) gaming in generall Haiku is quite a poor choice at the moment
<Nephele>
I gues you can make it run scummvm games or other 2d stuff, that ought to work neat
<Nephele>
but even stuff like minetest on a modern cpu runs barely playable on haiku
lubos76 has quit [Quit: Vision[]: i've been blurred!]
<botifico>
[haikuports/haikuports] Begasus a1b2465 - khtml, revbump for openssl3 (#11004)
FreeFull has joined #haiku
kescher has quit [Quit: Bye]
kescher has joined #haiku
tqh has quit [Quit: Leaving]
<Begasus>
libetpan/clawsmail almost done PulkoMandy :)
<Begasus>
have to move to xcairo1.18
<Begasus>
launches fine, just stupid gtk keymap :P
<Begasus>
anyway will be for later
Begasus_32 has quit [Quit: Vision[]: i've been blurred!]
<x512[m]>
dovsienko: Windows 95 floppy formatting problem reason is not bad multitasking, but bad floppy drive driver and DOS compatibility. Windows 95 implements proper preemptive multitasking.
<dovsienko>
let's not argue about it
<dovsienko>
let's say if OS/2 was technologically worse than Windows 95, OS/2 would have died naturally rather than being legally smothered by Microsoft
<phschafft>
Windows 95 was right at the edge between hardware access was considered blocking the machine and not.
<phschafft>
in a world with little multitasking there was no need for that. and users also didn't had a concept of doing two things at once with one machine.
<phschafft>
don't want to say that some products where not objectivly better than others. just that it was an exciting time.
<dovsienko>
did you have the experience of burning a CD when CD burners were still 1-speed and it would take around 60 minutes to burn 650MB?
<phschafft>
dovsienko: I had no burner for... like... untill 2015? I think I used it first in 2018?
<phschafft>
and it was like *whrrrrrooom* and done.
<dovsienko>
so, around 1999-2000 on Linux you could just start the process and use the PC for other normal work, and on Windows you would have not to touch the PC for those 60 minutes
<phschafft>
I did that at my friends. just did like 'whatever /dev/cdrom' to copy without any buffer or something. and they were very confused how the cd was fine, not broken for some buffering problems.
<dovsienko>
so doing two things at once was a bit more common than it may seem
<dovsienko>
at least _wanting_ to do more than one thing at once, successfully
<dovsienko>
anyway, my scepticism about Windows motivated me to look not only at OS/2 and Linux, but also at BeOS, that's why many years later I decided to have a look at Haiku
<phschafft>
dovsienko: but in absolute time it was very quick.
<phschafft>
computers had been around for 30 years, and then in 10 years they *boom* git to what is basically the current era.
<dovsienko>
with a 8-speed or better CD writer, yes
<dovsienko>
a 36-speed writer does a CD in about 2 minutes
<phschafft>
multitasking desktops with the power to run many things in parallel. where watching a movie became a background thing you would not think about it's resource usage at all.
<dovsienko>
writing at the nominal CD speed (150KB/s) takes exactly as long as playing an audio CD
<dovsienko>
phschafft: there's even hardware acceleration for certain video codecs now, built into general-purpose CPUs
<phschafft>
and even for those without explicit support there is so much generic acceleration.
<dovsienko>
that said, even in 1999 I would expect my modest PC to play music and to access the Internet at the same time
<phschafft>
yes.
<phschafft>
and that is why stuff like Windows 95 are at the border of that. it's a bit blurry. but it's about there.
<dovsienko>
I remember trying to play an MP3 file on a 486DX, it barely kept up with real-time decoding
<phschafft>
then you had a bad decoder.
<phschafft>
I had a Ice Age VCD.
<phschafft>
and it worked on mx 486 on windows 3.11
<dovsienko>
it was still a very novel thing at the time, 1996 or 1997
<phschafft>
you should not move the mouse. but if you just let it play it had no frame drops.
<phschafft>
that is with microsoft's own media player!
gouchi has quit [Remote host closed the connection]
<phschafft>
(I think I had a later codec pack installed. but still the CPU is the same)
<dovsienko>
hard times -> strong people -> good times -> weak people -> hard times -> ...
<phschafft>
I think it's a little more complex than that.
Anarchos has joined #haiku
zardshard has left #haiku [Disconnected: Replaced by new connection]
zardshard has joined #haiku
zardshard has left #haiku [Disconnected: Replaced by new connection]
zardshard has joined #haiku
<x512[m]>
dovsienko: Windows 95 was more technologically advanced than OS/2.
<Habbie>
nostalgically it feels different ;)
<dovsienko>
arguably, it used batch files instead of Rexx for build-in shell scripting, but let's agree to disagree
<jmairboeck>
Begasus: I think I found a bug in fpc on 32 bit: there are two versions: fpc 3.2.0 and fpc_x86 3.2.2, but both link against the gcc2 version of libroot. Why is it called fpc_x86 then?
<jmairboeck>
according to listimage
ChaiTRex has quit [Remote host closed the connection]
ChaiTRex has joined #haiku
<PulkoMandy>
Freepascal has a long history of doing weird things with secondary architecture
<PulkoMandy>
It only needs libroot C APIs (not even libgcc) so in theory it can be linked with libraries from either ABI. But it would make sense to package it as secondary arch and consistently use gcc13 libraries if possible
<jmairboeck>
according to readelf, it links libroot.so and libnetwork.so, and listimage shows /boot/system/lib/libroot.so for both versions. That should be /boot/system/lib/x86/libroot.so for the "gcc13" version, correct?
<PulkoMandy>
Ah, yes, but that could be a runtime_loader problem
<PulkoMandy>
The elf file just says "libroot.so" and the runtime loader is supposed to detect the "architecture", normally it does that by detecting a gcc version string in the elf file I think?
<PulkoMandy>
Or something like that, which wouldn't be there in this case because it's not built with gcc
<jmairboeck>
so that would have to be fixed in runtime_loader to detect fpc?
<PulkoMandy>
either that, or fpc should include whatever is needed in its generated files for the existing detection to work
Anarchos has quit [Quit: Vision[]: i've been blurred!]
dby has joined #haiku
nosycat has joined #haiku
HaikuUser2 has joined #haiku
HaikuUser2 has quit []
bbjimmy has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
xet7 has joined #haiku
<nekobot>
[haiku/haiku] 62b6a4181d32 - mkfifo: Define in terms of mkfifoat().
<jmairboeck>
although the Pascal build would have been significantly faster than all this autoconf crap
<Begasus>
I had (have?) a script that did a build with pascal, but that didn't work with haikuporter for some reason (should be up at the forum somewhere I guess)
<jmairboeck>
I noticed that too
<jmairboeck>
It complained that it didn't find the system unit
<jmairboeck>
outside of haikuporter it worked fine
<jmairboeck>
that's what the SYSTEM make variable is for, it gets passed as -T to fpc in the Makefile
<Begasus>
ps, I think you would need fpc_source also?
<jmairboeck>
there is no dependency on it, I don't think I have that installed
<Begasus>
did it work before, nothing really changed regarding fpc in the last year(s) I think?
<jmairboeck>
no, it didn't work, at least today
<jmairboeck>
so in the end I switched back to the C version
<Begasus>
hmm come to think of it, when building Lazarus the last times I used fpcupdeluxe, that takes care of building fpc+lazarus
<Begasus>
too much wip atm to dive into texlive :)
<jmairboeck>
I'm just taking it apart now as much as reasonably possible, nothing more :)
<Begasus>
np, should dive into it when you are almost there :D (still haven't re-installed the 2024 version here on 64bit, for now keeping up with full-sync as much as I can
<botifico>
[haikuports/haikuports] Begasus eabbe8e - libetpan, revbump for openssl3 (#11005)
<jmairboeck>
good that you are taking care of the openssl 3 stuff :)
<grepwood>
nephele, considering I have a Roland SC-88 Pro connected directly to my sound card, would that work with ScummVM on Haiku?
<Begasus>
some should be removed there (deleted in the repo), but first tackling those I can :)
<hsp>
is it possible to install haiku on ext4 file system? I think not
<Begasus>
never tried hsp :)
<grepwood>
also as far as I can tell, the card supports vesa and the BIOS (non-EFI) has no problem with that... grub doesn't even have problems to set up anything from 640x480 to 1080p on that card
<hsp>
I'll try it in a vm later on
<jmairboeck>
Begasus: When you are finished with that, I have a comparatively minor thing for you: could you rebuild tuxpaint with libpaper 2? I updated libpaper because of this texlive thing, and tuxpaint was one of the (few) other things using that besides texlive.
<jmairboeck>
when all things are ported, we could remove the old version
<Begasus>
ah, would be a quicky jmairboeck :)
<Begasus>
providing nothing fails :)
<Begasus>
melonds failed build on 32bit, that will be for later also
<botifico>
[haikuports/haikuports] Begasus 07d59ab - claws_mail, revbump for openssl3 (#11006)
<Begasus>
k, version bumped for xcairo on this because latest bump on poppler_glib used that one
<PulkoMandy>
hsp: No, ext4 won't work, the OS relies on queries to access files by their extended attributes and that is not implemented on most filesystems. Also the bootloader only knows about bfs
<waddlesplash>
well, the bootloader also knows about FAT, but that won't help here
<waddlesplash>
it has some read support for a few other FSes but ext2/3/4 are not among them
<dovsienko>
hsp: what problem are you trying to solve?
<waddlesplash>
grepwood: depends what soundcard you have and what support you are asking about. If you just want audio out then that may work. If you want MIDI, not sure what soundcard drivers we have that support that; it's definitely supported via USB MIDI at least
<waddlesplash>
PulkoMandy: thanks to packagefs booting off volumes that don't support queries probably works now
<waddlesplash>
since packagefs provides its own queries
<waddlesplash>
but indeed ext4 won't work as the bootloader doesn't know it
<waddlesplash>
and our ext4 driver needs some more work anyway
<grepwood>
waddlesplash, SoundBlaster Audigy 2 ZS, with the extra MIDI DA-15 bracket
<PulkoMandy>
We have a driver for the "mpu401" port on soundcards, but I don't know if scummvm has code to talk with haiku midi kit
<Begasus>
heh, gettext biting me again :)
<waddlesplash>
grepwood: well, our Audigy driver has MIDI support it appears, but who knows if it works :)
<hsp>
dovsienko, was purely out of interest the question
<grepwood>
yeah it wouldn't be Haiku's problem alone PulkoMandy
<grepwood>
I'm struggling to make Sonic3K talk to the MIDI port
<grepwood>
on Windows XP x64
<Begasus>
jmairboeck, build is OK lib:libpaper >= 2.2.5
<nipos>
nephele: Can you maybe share a screenshot where I can see the issue https://dev.haiku-os.org/ticket/18968 or,even better,send a email to nipos@haikumail.online in which it happens,so I can try and fix it?
<Nephele>
I can't see the checkbox anymore in gerrit if a comment chain is resolved or not. Anyone else sees this? was there a gerrit update too? or only a webkit update?
<Nephele>
nipos: it's if an email has lines indented with "> " and the option colored quotes is enabled
<Nephele>
in the code those are just hardcoded colors, so should be quite easy to see. But i can make a screenshot if you still want one
<nekobot>
[haiku/haiku] 706d9ddd108b - packagefs: Use "pkgfs" in object cache names.
<nekobot>
[haiku/haiku] b60fa5fa051c - packagefs: Use only one object_cache for all TwoKeyAVLTreeNodes.
<nekobot>
[haiku/haiku] 4e88440d3f38 - packagefs: Use ClassCache for UnpackingDirectory and UnpackingLeafNode.
<PulkoMandy>
nephele: Itws probably something wrong in the rendertheme in webkit for rendering checkboxes
_-Caleb-_ has left #haiku [#haiku]
<dovsienko>
hsp: it would be best to try the happy path first to see how things are intended to work, then you can try changing one thing at a time to see what breaks
<Nephele>
PulkoMandy: you are probably correct. Damn :)
_-Caleb-_ has joined #haiku
<nipos>
nephele: You mean the green text then? https://upload.odirf.de/file/WTxM966ma881.jpeg?view I played around with getting a email that looks bad a few days ago already,but for me this is readable perfectly fine
<Nephele>
Yes, for example. This qoes severall layers deep
<Nephele>
goes*
<nipos>
And deeper layers are worse to read or does it keep this color?
<Nephele>
they are different colors
<nipos>
Okay,thanks,will have to send myself some additional mails then
<Nephele>
just send one with "> test" "> > test1" "> > > test3" etc
<Nephele>
the first layer is "okay" to read in dark mode the second layer is worse, the third layer is completely unreadable
<Nephele>
those are just hardcoded colors
<Nephele>
ideally they wouls be hsdl colors, and would adopt their lightness value from the text
_-Caleb-_ has left #haiku [#haiku]
<Nephele>
waddlesplash: is there already an api for copying lightness for the hsl test you made in appearence?
_-Caleb-_ has joined #haiku
<waddlesplash>
nephele: no, it's all private
<Nephele>
well, it is also quite incomplete. Was wondering if that part was done yet anyhow.
<nipos>
I may be able to use mix_color with B_DOCUMENT_TEXT_COLOR and the currently hardcoded value
<Begasus>
jmairboeck, now you tell me? :p
<Nephele>
Sure, we can get "okay" solutions with that, but perhaps we should start moving hsl into libshared so we can fix those issues with it to get some experience with that api
<Nephele>
see how much we need in practice, what can be improved
<botifico>
[haikuports/haikuports] Begasus 17ccad2 - tuxpaint_config, revbump for libpaper2 (#11008)
<Nephele>
nipos: for the colors changing on size changing but not for color updates i'm not sure. maybe there is some invalidation missing? maybe another dev has a better explanation of why that behaves in that way
<Habbie>
i built a HS colorwheel earlier this week ;) (but this is not on Haiku) https://imgur.com/a/KFLxZYs
<nipos>
No,the point is that it doesn't change
<nipos>
Otherwise it would be easy,yes
<Nephele>
it does not change on font sizes?
<Nephele>
Oh, then i misunderstood
<nipos>
->SetFontAndColor is called when changing the size,but the color still stays the same,only the size changes
<Nephele>
in that case you just have to tell it the new color when a new message about the color is received with SetFontAndColor
<Nephele>
yeah, SetFontAndColor has default arguments for arg2 and arg3
<Nephele>
it passes in color = null, which the TextView basically treats as "whatever we were doing before"
<nipos>
Yeah,but I never gave it any color in the first place,so why does it not use the HighUIColor?
<Nephele>
Text is styleable, you can tell it to draw some chunks in red and some in orange if you want
<nipos>
Currently reading the API specs if there is a alternative to SetFontAndColor that doesn't set the color and just uses the default
<Nephele>
It got the current (correct) color when getting created the first time on it's own
<Nephele>
but after that the text just keeps the color it had
<Nephele>
it's not set to a UI color, but just a specific color
<Nephele>
you basically told it to keep the current color by adding = null for the color var in SetFontAndColor
<Nephele>
when you get a message about a color update check if it is the correct color, and then use that with SetFontAndColor to tell it to use a new color
<Nephele>
(correct color constant, as in, the one you care about for this text)
<botifico>
[haikuports/haikuports] Begasus 99e39fb - claws_mail, revbump, set poppler_glib version (#11009)
yann64 has quit [Quit: Vision[]: i've been blurred!]
<Nephele>
nipos: the BView does, but the text color is seperate from that
<Nephele>
iirc
mmu_man has quit [Ping timeout: 480 seconds]
<Nephele>
Skipp_OSX: do you have input for that? (live updating of colors with text in a BTextView)
<Skipp_OSX>
apparently not since it seems like Appearance has the same bug...
<Skipp_OSX>
It's probably just a bug somewhere, should be calling SetUIColor instead of SetHighColor or whatever, but no, no input for that.
<Nephele>
not too sure about that, since the color can be set explicitly. But maybe it should just follow the UI color if there never was passed a color into SetFontAndColor
<Skipp_OSX>
you could look at how DeskCalc's CalcView does it I suppose to get an idea.
<Skipp_OSX>
with color drops, your favorite :)
<Nephele>
DeskCalc was one of the first things i started working on for fixing dark mode :P
<Nephele>
not that i managed to fix it back then
mmu_man has joined #haiku
<nipos>
I nearly have it,thanks for your help
<Nephele>
better not announce it too loudly, or you'll be sucker into fixing pref/appearence too! /s
<nipos>
It does now change colors when changing the size.Making it also listen to color change messages should be trivial then
<nipos>
I'll do it,it's basically the same code
<Nephele>
You can check deskcalc CalcView.cpp:294 roughly if you want on how to get the color from the message
<Nephele>
in theory you can also call ui_color there, but that seems silly to me :)
<nipos>
The function for redrawing does already call ui_color,I only need to call that function
<Nephele>
yes... but ui_color asks the server for the color again, where it has already send it in the BMessage, and we should only redraw if its actually the color we want and not some random other color
<Nephele>
otherwise with the automatic theme (in theory) you'd be redrawing like 38 times
<Nephele>
or actually it might check private data of the application instead? not too sure :)
<jhj>
Have you guys looked at replacing zlib with zlib-ng? It's source and binary compatible with zlib and actually considerably faster. A few major Linux distributions have adopted it (Fedora, next RHEL)...
<Nephele>
I haven't heard of it personally. I hope it does not need another compiler ;)
<jhj>
No, it should build fine with gcc or clang :)
<Nephele>
in that case, since haikuports depends on libraries, and not on ports, generally. You should be able to make a port for it and users would be able to install either
HaikuUser2 has quit []
<jhj>
I implements the hot paths in assembly and autodetects the best version to use at runtime.
<Habbie>
nephele, doesn't anything in haiku itself use zlib?
<Nephele>
Habbie: probably? But i highly doubt we have it imported intree
<jhj>
nephele: They'd conflict so users would have to pick.
<Habbie>
nephele, ah right
<Nephele>
jhj: yes
<Habbie>
nephele, was thinking too BSD there :)
<Nephele>
Habbie: hmm? what does BSD do differently there?
<jhj>
It can be installed side by side and used with LD_PRELOAD but bobody would want the hassle
<Nephele>
I guess it is kind of strange that haiku depends on haikuports too
<Habbie>
so, starting one step back, if you look at debian or redhat, everything is a package, there's strictly speaking no 'in-tree' at all
<Habbie>
but in the BSDs, there's a 'src' or 'world' tree that has all the binaries, and a bunch of libs, several of which are imported/forked
<Habbie>
-and- a ports system
<Habbie>
haiku is somewhere in between these two i guess
<Nephele>
jhj: well, for a desktop OS i think it does not matter too much if you can co-install them. You can if you want, if you install it into a different tree ($HOME/config vs /system for example)
<jhj>
nephele: You could eventually supercede one with the other after a period of testing I guess
<Nephele>
Having both available in ports might be neat, we can pick one we want, but it can be replaced if needed
<jhj>
It's a pretty good advantage on low end hardware
<Nephele>
Well, starting with a port would be a good first step
<nipos>
nephele: I added a B_COLORS_UPDATED case to MessageReceived,but it seems it never gets called.Is it possible it's filtered somewhere before reaching the FontSelectionView?
<Nephele>
it should be easy to replace the zlib version installed then, and show that it works
<nipos>
I did it similar to how AboutSystem does it
<jhj>
nephele: sounds like a fun project for me this weekend after I do other things :)
<Nephele>
nipos: sure, if it's handeled somewhere else and the default: case does not call it's parent's classes MessageReceived
<Nephele>
I had severall issues in Renga where this was the case, and as a result windows could not be screenshotted for example .-.
<PulkoMandy>
jhj: I think we tried zlib-ng a few years ago and we didn't end up using it. But I don't remember why and if we had problems with it. So it's probably time to try again
<nipos>
Hm,I'll first try it in the Appearance prefs then,maybe it works better there
<jhj>
PulkoMandy: Fedora had both for awhile and just recently zlib-ng superceded zlib.
<jhj>
So thats at least a promising sign it might be worthwhile to look at
<jhj>
It tested about 2X as fast on SSE2 (and 4X as fast on an AVX2 machine) so the speedup is nice.
<jhj>
That was using it standalone, I didnt try to replace the "system" zlib yet
<Nephele>
that is cool indeed
<Nephele>
> Zlib compatible API with support for dual-linking
<Nephele>
What is dual-linking?
<jhj>
I see gdb works in Haiku now, that's awesome guys!
<botifico>
[haikuports/haikuports] Begasus 0fa9eca - openimageio2.5, revbump for openssl3 (#11010)
<nipos>
I want to make it work now
<jhj>
I dont know what they mean by dual linking
<Nephele>
it seems they also have a zlib incompatible api. Do the other OS use that? is there an advantage to it or only "easier" to write for?
<nipos>
It's half-way done already and the root cause identified,so I think I should get it to work today
<jhj>
nephele: I think its faster somehow
<Nephele>
the root cause is object orientation! /s
<jhj>
their own minizip thing uses it and its faster
<Nephele>
jhj: hmm, well it's possible that it is faster with another api depending on how it is constructed
<jhj>
I think the main advantage is easier to use though
<Nephele>
But that is then more maintenance if both are to be maintained
<Nephele>
dunno though
<Nephele>
let's see how the compatible mode fares on haiku
<jhj>
nephele: there might be a concern if you require bit reproducibility anywhere
<jhj>
then you'd need a flag day for building ports or what not
<jhj>
needing to switch all builders at once
<Habbie>
i'd hope ports don't depend on -compressed- output generated locally to be stable
<Nephele>
reproducible builds? I think that is a nice goal *eventually* but right now uhm, i doubt there is much work in that direction
<Nephele>
needed man-power and such
<jhj>
Habbie: GitHub produces release tarballs on demand
<Habbie>
oh yes, i'm well aware of github breaking checksums on those once or maybe twice
<jhj>
Thry tried upgrading to zlib-ng and it broke all the checksums once
<Habbie>
ah, didn't recall it was -ng
<Habbie>
i vaguely recall the actual change was in git itself
<jhj>
One time it was, another time was something else but similar
<nipos>
nephele: I made live updating work,but now my whole UI except the mouse pointer hangs.Will have to reboot and try again then
<jhj>
I also believe Debian has multiple builders and notices if a build isnt reproducible but also if a package itself is different
<Nephele>
well, it's not live if it's hanging now is it? /s
<jhj>
Anyway, I guess not a problem for Haiku currently
<Nephele>
we have like one builder per arch
<nipos>
It updated three times or so before it hanged.Maybe it's just an accident.
Maylay has quit [Quit: Pipe Terminated]
xet7 has quit [Remote host closed the connection]
_-Caleb-_ has left #haiku [#haiku]
_-Caleb-_ has joined #haiku
B2IA has quit [Ping timeout: 480 seconds]
HaikuUser2 has joined #haiku
HaikuUser2 is now known as agent_J
<nipos>
Fine,the preview does now properly live-update,but the font size BSpinner doesn't...
<nipos>
And I fail to find documentation for BSpinner
MrSunshine_ has quit [Quit: Leaving]
B2IA has joined #haiku
<Nephele>
some stuff simply ins't documented :/ I usuaully try to improve this when i discover some, but writing docs also takes time
<Begasus>
done for today, need investigating on the other 2 openimageio build errors
<Begasus>
cu peeps!
Begasus has quit [Quit: Vision[]: i've been blurred!]
mmu_man has quit [Ping timeout: 480 seconds]
HaikuUser2 has joined #haiku
HaikuUser2 has quit []
mmu_man has joined #haiku
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
<nipos>
I can't even find Spinner.h in the be/interface header directory and Genio can't find the header file either,but the compiler does and it works.Now that's really weird
<Skipp_OSX>
headers/private/interface I believe
agent_J has quit [Remote host closed the connection]
<Skipp_OSX>
Haiku addition so it is private, part of GSOC 2014
erysdren has joined #haiku
<nipos>
Thanks,just found that myself :)
Anarchos has joined #haiku
<Skipp_OSX>
I hope you're not implementing time/date spinners...
jpelczar has joined #haiku
<nipos>
No
<Skipp_OSX>
I mean, it's a good thing to have but low priority
<Skipp_OSX>
ok then carry on
<nipos>
I only want to fix the color not live-updating for the font size spinners
<Skipp_OSX>
ah ok that makes sense
* Anarchos
is compiling haiku for 8 hours ... and still compiling in the buildtools...
<Skipp_OSX>
you're compiling one or more versions of gcc, but that is a long time. Shouldn't take that long unless your computer is slow.
* Nephele
remembers having a laptop taking 3 days to compile the linux kernel
* phschafft
... never compiled Linux...
<Nephele>
phschafft: yeah, me neither! i have a compiler for that!
<Skipp_OSX>
I sadly, have compiled Linux
<Anarchos>
Skipp_OSX my computer is slow :)
<Skipp_OSX>
ok then that is normal bc gcc is large
<scantysnax>
i built gentoo on a sinle ultrasparcII 300MHz. it took several days.
<Anarchos>
Skipp_OSX i know. and only 1Gb RAM doesn't help here
<Nephele>
I once tried building gentoo with the documentation flag, which of course wasn't possible
<Skipp_OSX>
generally you would cross-compile in that case, compile for one target to another but I understand why you wouldn't want to do that.
<Skipp_OSX>
this explains so much
<nipos>
The only method on BSpinner related to colors is private,funny.I could access the underlying TextView that is public and manipulate that directly,but that would be rather hacky I guess and it's probably better to fix BSpinner to automatically adjust its colors
<Nephele>
Yes, in that case BSpinner should be fixed instead of every view that uses it
<phschafft>
hahaha.
<scantysnax>
what does a BSpinner look like?
<Skipp_OSX>
I can tell you how to fix properly but you might not like it...
<Nephele>
scantysnax: the textbox in font preferences with plus and minus button
<nipos>
I never compiled any major OS,only SerenityOS but that's also years ago and was rather fast :)
<Skipp_OSX>
Instead of inheriting from BControl, inherit from BView and BInvoker (multiple-inheritance) then implement the SetValue method yourself to take a color
<nipos>
Skipp_OSX: Yes?
<Skipp_OSX>
wait... nvm, that is a different problem...
<Nephele>
huh?
<scantysnax>
nephele: thanks.
<Skipp_OSX>
well, it's probably a good idea anyway, and would allow you to make that part non-private...
<nipos>
The problem is that the underlying BTextView doesn't automatically adjust it's colors
<Skipp_OSX>
yeah different problem, wouldn't fix, sry
<Skipp_OSX>
yeah BTextView is a fickle beast
<nipos>
And the BTextView not auto-updating may be wanted,since it supports text with different formatting options.I wonder if BTextView is the right thing to use here,why not a simple text input box?
<Skipp_OSX>
because BStringView does not wrap
<Skipp_OSX>
annoyingly...
<Nephele>
Skipp_OSX: the question was about the spinner input boxy
<Nephele>
box*
<nipos>
...but the BSpinner doesn't need to wrap?
<Skipp_OSX>
oh right, confusing with web positive bug
<Skipp_OSX>
no it's a text box because it has a text input and label
<Nephele>
appearence prefs already has a better number entry style for the date picker
<Skipp_OSX>
it's a textview with buttons tacked on
<Skipp_OSX>
well it's a text control with buttons tacked on which is a text view with label tacked on
<nipos>
I already fixed the font previews to auto-update and wrap using TextView and manually adjusting the colors when a message is received.And it does indeed make sense to use TextView here.But why does the Spinner need a TextView internally and not just a simpler sort of a text input box that does live-update colors without manual intervention?
<Nephele>
we don't have a simpler sort of text input box
<Skipp_OSX>
we don't have anything simpler...
<Skipp_OSX>
not that supports the features we need here, editing
<Nephele>
and yes, in my opinion this is overkill. The appearence prefs date input shows how numbers can be entered in a better way for such a UI
<nipos>
Then all other text input boxes also can't live-update text colors and I just didn't notice it?Or is there any difference between them?
<Skipp_OSX>
BColorControl also uses an internal BTextView for the rgb values, it's not uncommon thing to do.
<Skipp_OSX>
There may be shenanigans going on but check any color control on the system, Backgrounds has one
<Nephele>
nipos: no they can, but i think either (if in a control) this is done by the control, or done by the application
<Skipp_OSX>
as far as I know they live update...
<nipos>
Oof,I thought I understood what was going on,but turns out I don't
<Nephele>
Skipp_OSX: Is BColorControl the one in appearence? that is in generall just pretty bad for color picking
<Nephele>
not that the color values help there tbh
<Nephele>
err text inputs. anyhow
<Skipp_OSX>
yes
<nipos>
Simple text inputs which are technically also BTextView *do* auto-update,but BSpinners don't
<Nephele>
maybe we can make a simpler number input control like in datetime, for such things
<Nephele>
nipos: rengas text entry does not auto-update either
<Skipp_OSX>
yeah it must be Bpinner shenanigans, color controls work fine
B2IA has quit [Quit: Vision[]: i've been blurred!]
<Nephele>
I think with the design of BTextView now this is missing code in BSpinner to react to the color update message and update the thing internally
<nipos>
Interesting.Currently chatting on Movim in LibreWolf,otherwise I would have noticed...
<nipos>
The test input in the keymap preferences does auto-update,that's where I tested
<Nephele>
Without looking at the code that's not a good test
<Nephele>
:P
<nipos>
I wanted to know if text inputs generally don't live-update.Now I know that some do.That's good.
<nipos>
I'll first upload the improvements for WebPositive and Appearance,then I'll have a look at the BSpinners not updating,which is a different thing.Trying not to mess up too much with my local repo at the same time.
<Nephele>
by the way Skipp_OSX, you tend to rebase changes and then reupload them to gerrit. Just wanted to let you know it is a tad annoying, since I will get an email for each rebase (or noop change) like that
<Nephele>
not sure if we can tell gerrit to ignore these types of changes, since having to have changes on different branches locally is also quite annoying
<Skipp_OSX>
I was afraid about the multiple emails...
<Skipp_OSX>
can you take yourself off the shortcuts_refactor relation chain?
<Skipp_OSX>
yeah I have big stacks of changes one depending on the next and Gerrit is annoying and haiku-format bot is annoying with emails.
<Nephele>
I can probably take myself of changes sure, but if you have other patches locally I've reviewed I will get emails for those, since it always considers those a rebase :)
jmairboeck has quit [Quit: Konversation terminated!]
<Nephele>
format bot should not email IMO, and build bot only on failed builds
<Skipp_OSX>
yeah we should turn off email for haiku-format bot
<Skipp_OSX>
I have to update again so probably take yourself off now
<Skipp_OSX>
once other updates go through or I'll put them on the stack otherwise
<Nephele>
it's fine. This is not a big problem, but a systemic one of gerrit
<Skipp_OSX>
yeah but you can remove yourself from cc list so at least you don't have to see it.
<Nephele>
nipos: unrelated question but why is the text for the a-Z sentence different between those two applications?
<nipos>
I honestly can't answer that. The english version seems to use the same text in both applications. I haven't done the translations
<Nephele>
Skipp_OSX: in the new changeset from nipos he explicitly handed the BMessage down to the FontSelectionView from SettingsWindow
<Nephele>
can you take a look at that? I feel like that shouldn't be needed
<Skipp_OSX>
I believe I have a 1 line fix for the spinner bug, are you ready?
<nipos>
Yes
<Skipp_OSX>
ok, 2 line... hold testing
<Nephele>
This won't become that famous 3 line diff i hope?
<nipos>
nephele: I also thought this shouldn't be needed, but not doing it results in the FontSelectionView never receiving a B_COLORS_UPDATED message
<nipos>
Now after having done some more work on Haiku,I finally created a pull request to get added to the contributor list as you suggested a week or so ago,nephele.Wanted to wait a bit,not directly after the first patch
Chain|Q has quit [Quit: [.life.support.system.failure.]]
<nekobot>
[haiku/haiku] c08e54426ed8 - libroot_debug: Use overcommitting and MADV_FREE in guarded heap.
<nekobot>
[haiku/haiku] be1d1fa3abdc - libroot_debug: Cleanup: Factor out duplicated code.
<Nephele>
nipos: put it into beta5 too
<Nephele>
;)
<nipos>
Thanks :)
<nipos>
For the WebPositive patch....I already filter that the _UpdateFontPreview function is only called if the right color setting is changed.With the current code,the same function can also handle changes of font size or font family.Using a color argument makes it more complicated,as there is no color argument known in the case of size or font family changes.
<Nephele>
the thing is, right now you are getting the rgb_color in the MessageReceived call. But you are discarding it, and instead asking ui_color for it directly
Chain|Q has quit [Quit: [.life.support.system.failure.]]
Chain|Q has joined #haiku
<nipos>
But I only get it if the color changes.If the size changes,I don't get it.With the current code,I can keep the universal update function.
<Nephele>
why make a universal function if it's not universal? :P
<nipos>
It is.It handles changes of the preview text font in every case.And it works perfectly fine.What's the issue with the ui_color call?
<Nephele>
you are asking for the color again. and that seems quite racy considering you already got the color. The thing is that updating colors, and updating the size (to me) are two independent operations, and don't relate to each other
<Nephele>
so maybe i'm just not understanding why this would be done like this
<Nephele>
instead of a one-line that calls SetFontAndColor(mycurrentfont, b_font_all, thecolorIgot)
Maylay has joined #haiku
<Nephele>
if you did *not* get a color update then asking for the ui_color and then sending it is superflous, as you would just get the same color again
gouchi has quit [Remote host closed the connection]
<Skipp_OSX>
there is no race condition, all superfluous
<nipos>
If I do not get a color update and call SetFontAndColor without any color argument,doesn't that reset it to black?
<Skipp_OSX>
standard annoyance dealing with BTextView
<Nephele>
nipos: no
<Nephele>
it just keeps whatever it had
<nipos>
"If the color pointer is NULL or not specified, the text is drawn in black. "
<Skipp_OSX>
yeah bc default color I guess
<nipos>
I didn't try it,but that's from the documentation
<Skipp_OSX>
but you get the color, then set it as needed, so I don't understand the bug here.
<Nephele>
SetFontAndColor(&fFont) should not modify the color iirc
<Skipp_OSX>
what?
<Skipp_OSX>
it's called SetFontAndColor() of course it sets the color
<Nephele>
What are you basing that on? There is no color passed. and there is no SetFont for that class
<Nephele>
If it *always* sets the color you would not be able to set font faces for parts of the string without manually keeping track of the colors
<Skipp_OSX>
it sets the color or it doesn't?
erysdren has quit [Quit: Konversation terminated!]
<Nephele>
Skipp_OSX: for example in stylededit you can make a string that is red a second one that is blue, Select both and choose "italic"
<Skipp_OSX>
You can set the parts you want to update with the second parameter
<Nephele>
if the api would set black if you did not pass a color in stylededit would have to have an internal array of those colors to reaply them
<Nephele>
yes and the default is B_FONT_ALL, no?
<Nephele>
not the color?
<Skipp_OSX>
right
<Skipp_OSX>
BTextView is old and crufty but that is not WebPositive's problem.
<Nephele>
hmm, but B_FONT_ALL is from "enum set_font_mask"
<Nephele>
so that makes no sense for the color
<Nephele>
but i looked at the code before (of textview internals) and it seemed like it would just ignore if you pass in NULL as a color
<Skipp_OSX>
which is why it's SetFontAndColor() but we're only concerned with the color here
<Skipp_OSX>
sure that makes sense if you only want to set the font style and not change the color, e.g. italics.
<Nephele>
enum with each entry beeing 2^N... sounds like there should be a bitfiel macro, or other language feature, heh
<botifico>
[haikuports/haikuports] pulkomandy b796aef - HaikuWebKit: fix 32 bit build
<Nephele>
Skipp_OSX: the update function is called now from two places, one when the size changes and the other when the color does
<Skipp_OSX>
heyo
<Skipp_OSX>
sure
<Nephele>
but when the size changes the color is uneeded, and when the color changes the color is known
<Skipp_OSX>
yeah the same function is used in two different ways in 2 places.
<Skipp_OSX>
It's a crufty function, you should be able to set either the font or the color but you can't, you either set the font or set the font and the color.
<Nephele>
the current font is tracked as a member of the class, so that's not that tragic
<Skipp_OSX>
so we get the font from the textview, set it again with the color.
<Nephele>
yep
<Nephele>
I think that should just be done in MessageReceived directly where the color was just passed in
<Skipp_OSX>
he put the cruft into a convenience function.
<Skipp_OSX>
I put the cruft into a convenience function in Appearance and he copied my cruft convenience function more than likely.
<Nephele>
The preview function was only a oneline to set the font, it should still do that, color is independent imo
<Skipp_OSX>
but BTextView makes you do it that way.
<Nephele>
You don't have to pass a color
<Skipp_OSX>
can you show me the code you're talking about?
<Skipp_OSX>
I am thinking of in WebPositive where we are passing a color...
<Nephele>
MessageReceived gets the color, then calls this function without the color and it gets re-acquired with ui_color
<Nephele>
I think the oneline of SetFontAndColor should be done in MessageReceived, setting the size is not needed
<Nephele>
and in the function to set the size the color should not be needed to be passed to SetFontAndColor
<Nephele>
afaik anyway. And I would be quite suprised if not setting it made it black
<Skipp_OSX>
It is setting the color to the current B_PANEL_TEXT_COLOR
<Nephele>
Yes, but either it is already set like that. Or it is from messagereceived and just got the color
<Nephele>
anyhow, i'm just gonna cherry-pick this and test this out to proove it
<Skipp_OSX>
ok I'm still confused.
<Skipp_OSX>
It calls it once on Constructor, once when it receives a color change message.... that seems normal to me.
CRThaze_ has joined #haiku
<Skipp_OSX>
Sometimes things aren't setup right and you have to call in AttachedToWindow() that could be the bug I suppose.
<Skipp_OSX>
the constructor is good enough most of the time, so I doubt it, would need to be some special circumstance of depending on something else.
CRThaze has quit [Ping timeout: 480 seconds]
<Skipp_OSX>
yeah it gets the same color twice, oh well
tuaris has joined #haiku
<Nephele>
I changed the commit, if it works i can show you what i mean
<Nephele>
testing now
<Skipp_OSX>
ok
BlueSky71 has quit [Ping timeout: 480 seconds]
<Nephele>
Don't have it compiled and running yet, but i think this should work: http://0x0.st/Xv6c.diff
dby has quit [Quit: Vision[]: i've been blurred!]
<Nephele>
nipos: seperate issue, if you increase the font size the box around it changes, and if i then change it back now the box is smaller than before
<nipos>
I can't reproduce that.It goes back to the original size here.
<nipos>
Do you have the issue only with WebPositive or also with Appearance preferences,which are using more or less the same code?
BlueSky71 has joined #haiku
<Nephele>
only webpositive
B2IA has quit [Quit: Vision[]: i've been blurred!]
<nipos>
Is it a big difference or only one or two pixels?
<Nephele>
big yeah
<nipos>
Sorry,I really can't reproduce it.Opening WebPositive,put my finger on the screen at the bottom border of a text preview box,make it a lot bigger,bring it back to 14 and the border is exactly where I have my finger
<Nephele>
I think i'm stupid and passing the pointer instead of the value or something, the text is colored randomly for me on each launch
<nipos>
Did you try compiling my original version?Everything works fine for me.
<Nephele>
ah, i've been using uninitialized memory, nice